Patent Application
20060069535
The present invention relates to methods and apparatus which convert data, as well as to methods and apparatus for providing warning signals. The present invention is further directed to methods and apparatus which receive data from one or more sensor and/or from one or more source of information, and predict, as early as possible, emergencies or other situations which require action.
There exist a variety of systems which gather data and attempt to identify emergencies or other situations which require action. Frequently, such data is obtained from sensors, databases or the like, which may produce an enormous amount of data. However, it is inefficient and in some cases impossible to process each piece of information substantially simultaneously and in real-time from hours of recorded data. Also, simple calculations of average values do not always give an accurate picture.
In accordance with one aspect of the present invention, there is provided a computer-readable medium having computer-executable components for converting data relating to a specific system, comprising: a data evaluation component which generates converted data information based on evaluation of data; a general report generating component which receives data information from the data evaluation component and expert information, and which generates general reports based on the expert information and the data information; a system-specific report generating component which receives information pertaining to the specific system and expert information, and which generates system-specific reports based on the expert information, the information pertaining to the specific system and data information received from the data evaluation component; and a decision-making component which makes decisions based on data information from the data evaluation component, the general reports and the system-specific reports.
The systems which can be monitored according to the present invention can include generally any system for which there is a desire to monitor (over long periods and/or dynamically), and for which information which is useful in performing such monitoring can be obtained by collecting data as to one or more parameter which can be measured, detected or otherwise sensed. For instance, systems which can be monitored according to the present invention include, for instance, humans (e.g., medical monitoring, body-building, exercising, mental abilities), the earth (e.g., global warming, weather patterns, air quality), the solar system, roadways (e.g., traffic flow, deterioration), polling (e.g., political elections, television programming), etc.
The apparatus preferably further comprises one or more sensors which collect data relating to one or more parameters, respectively, of or relating to the system. The methods and apparatuses according to the present invention are used to evaluate that data (1) to provide early warnings when there is a high likelihood of an emergency situation arising in the near future, (2) to advise that an emergency situation is present, and/or (3) to provide other analyses of the system or the data.
The data evaluation component generates data information which includes analyses of the data in specific formats, and provides such information to the general report generating component, the system-specific report generating component and the decision-making component.
The general report generating component generates general reports based on (1) expert information relating to the type of system being evaluated and (2) data information received from the data evaluation component.
The system-specific report generating component generates system-specific reports based on (1) expert information relating to the type of system being evaluated, (2) data information received from the data evaluation component, and (3) information pertaining to the specific system (as opposed to the type of system) being evaluated.
The decision-making component makes decisions based on (1) data information from the data evaluation component, (2) the general reports and (3) the system-specific reports.
The result is that the system-specific reports may be shifted from the general reports as a result of the information regarding the specific system. For example, in cases where the system is a human patient who is being monitored medically, based on expert information that elderly persons are more prone to high blood pressure, and specific system information that the patient is elderly, the system-specific report generating component can be configured to attribute less significance to a minor increase in the patient's blood pressure.
In one aspect of the present invention, the data evaluation component takes an initial set of data values (obtained from at least one sensor), and removes all data values which exceed a maximum possible value or which are below a minimum possible value (i.e., to remove incorrect values, e.g., values which can only have been obtained as a result of a sensor error or the like). The data evaluation component preferably determines whether a significant number of the data values in the initial set are outside of an “acceptable” range. If a significant number of the data values in the initial set are within the acceptable range, the data evaluation component preferably generates a “basic record”, which identifies a basic range (preferably within the acceptable range and which preferably encompasses fewer values than the acceptable range) within which a significant number of the data values in the initial set fall. The basic record is thus a baseline for the specific system being analyzed, against which subsequent “current records” can be compared, to help to see whether the data reflects a trending away from the original condition of the system, and possible movement toward a changed circumstance (and possibly an emergency situation). The data evaluation component preferably receives subsequent sets of data values, and examines whether significant quantities of data values are in specific ranges defined outside of the basic range and/or outside of the acceptable range. By evaluating the quantity of data values in a variety of such ranges established by the data evaluation component based on the data received, the data evaluation component selects from among different algorithms for generating data information in specific formats (examples of which are described in detail below), each format being useful in evaluating trends in the data values.
The present invention further provides methods for converting data into any of various reports, as well as components which are useful in converting data. The present invention further provides computer-readable media comprising computer instructions which, when executed by a computer, perform methods of converting data into any of various reports. The present invention further provides computer-readable media having computer-executable components as described herein. The present invention further provides methods and apparatus which provide warning signals regarding conditions and/or potential emergencies.
The invention may be more fully understood with reference to the accompanying drawings and the following detailed description of the invention.
The present invention provides a solution by generating specific records and images, showing basic ranges for statistical values. In accordance with the present invention, a potential problem is detected by comparing new readings (statistical values) with previously recorded readings to generate information which is easily understood and of a more manageable quantity.
In accordance with the present invention, in order to detect potential problems, raw data incoming from one or more sensors is converted and evaluated to form statistical values. The first step in this process is to establish the framework of analysis by identifying the basic and acceptable ranges of statistical values. The “acceptable range” of statistical values is the range between (and including) minimum and maximum values at which the system can typically function normally (for example, where the system is a patient, an acceptable range for the patient's heart rate, i.e., one of many possible parameters being monitored, may be 60 to 100 beats per minute). The “basic range” (or normal range) of statistical values is a range between the minimum and maximum values at which a particular system functions normally and without indication of a potential problem. Generally, the basic range is narrower than the acceptable range. The basic range may vary for different systems, but it must be determined for every system.
In accordance with one aspect of the present invention, there is provided a computer-readable medium having computer-executable components for converting data relating to a specific system, comprising: a data evaluation component which generates converted data information based on evaluation of data; a general report generating component which receives data information from the data evaluation component and expert information, and which generates general reports based on the expert information and the data information; a system-specific report generating component which receives information pertaining to the specific system and expert information, and which generates system-specific reports based on the expert information, the information pertaining to the specific system and data information received from the data evaluation component; and a decision-making component which makes decisions based on data information from the data evaluation component, the general reports and the system-specific reports.
Preferably, the data evaluation component, the general report generating component, the system-specific report generating component and the decision-making component are included on a single computer-readable medium (e.g., a computer-readable medium contained in a single computer or network), but alternatively, the data evaluation component, the general report generating component, the system-specific report generating component and the decision-making component can each be included on different computer-readable media, e.g., different computer-readable media on different computers, processors or devices, or can be included on one or more computer-readable media in any desired combinations (e.g., the data evaluation component and the general report generating component on a single computer-readable medium, the system-specific report generating component on a separate computer-readable medium and the decision-making component on a separate computer-readable medium).
Preferably, the data evaluation component, the general report generating component, the system-specific report generating component and the decision-making component are in the form of software stored on a computer-readable medium. Preferably, the general report generating component, the system-specific report generating component and the decision-making component are contained in a single software program, but alternatively, the data evaluation component, the general report generating component, the system-specific report generating component and the decision-making component can be contained in different software programs, or can be contained in one or more software programs in any desired combinations (e.g., the data evaluation component and the general report generating component on a single program, and the system-specific report generating component and the decision-making component on a separate program).
As noted above, the at least one sensor collects data relating to at least one parameter, respectively, of or relating to the system. The data evaluation component generates data information which includes analyses of the data in specific formats which are sent to the general report generating component, the system-specific report generating component and the decision-making component. The general report generating component generates general reports based on (1) expert information relating to the type of system being evaluated and (2) data information received from the data evaluation component. The system-specific report generating component generates system-specific reports based on (1) expert information relating to the type of system being evaluated and (2) data information received from the data evaluation component, and (3) information pertaining to the specific system (as opposed to the type of system) being evaluated. As a result, the system-specific reports may be shifted from the general reports, due to the fact that the system-specific reports factor in information pertaining to the specific system, as well as expert information relating to the type of system.
Each of the components described herein performs a useful function. Accordingly, the present invention is directed to not only the apparatuses described herein which comprise a data evaluation component; a general report generating component; a system-specific report generating component; and a decision-making component, but also to each of the individual components in such apparatuses, e.g., the invention is also directed to data evaluation components. Furthermore, the present invention is directed to each of the individual method steps carried out by these components, and to all combinations of such method steps, as well as means for performing each of these individual method steps, and all combinations of such means for performing these method steps.
As noted above, in one aspect of the present invention, the data evaluation component takes an initial set of data values, removes all data values which exceed a maximum possible value or which are below a minimum possible value, and preferably determines whether a significant number of the data values in the initial set are outside of an “acceptable” range. If a significant number of the data values in the initial set are within the acceptable range, the data evaluation component preferably generates a “basic record”, which identifies a basic range (preferably within the acceptable range and which preferably encompasses fewer values than the acceptable range) within which a significant number of the data values in the initial set fall. The basic record is thus a baseline for the specific system being analyzed, against which subsequent “current records” can be compared, to help to see whether the data reflects a trending away from the original condition of the system, and possible movement toward a changed circumstance (and possibly an emergency situation). The data evaluation component preferably receives subsequent sets of data values, and examines whether significant quantities of data values are in specific ranges defined outside of the basic range and/or outside of the acceptable range. By evaluating the quantity of data values in a variety of such ranges established by the data evaluation based on the data received, the data evaluation component selects different algorithms for generating data information in specific formats (examples of which are described in detail below), each format being useful in evaluating trends in the data values.
The general report generating component, the system-specific report generating component and the decision-making component each contain arrays and/or networks of logic. For example, where the system is a human patient being medically monitored, the general report generating component might include logic to the effect that when the patient is undergoing heart surgery, if the patient's blood pressure decreases, there is a high probability that the patient is going into or heading toward cardiogenic shock (which has a 95% mortality rate); or, where a patient is presenting with non-septic shock, if the patient shows a rapid SaO2 decrease, there is a high probability that the patient is in anaphylactic shock, rather than neurogenic shock. Where particular parameters are related to a large number of different logic patterns (e.g., high blood pressure suggests increased probabilities of different conditions in a variety of respective circumstances), the networks of logic can be inter-related to a large degree.
Below is a discussion of a specific example of an embodiment according to this aspect of the present invention. In the discussion of this embodiment which follows, the data evaluation component which receives sensor data from the at least one sensor and which generates data information based on evaluation of the sensor data in the present embodiment is the “DataAgent”; the general report generating component in the present embodiment is the “GeneralNET”; the system-specific report generating component is the “PersonalNET”; and the decision-making component is the “ArbiterAgent”.
Preferably, DataAgent, GeneralNET, PersonalNET and ArbiterAgent are included on a single computer-readable medium (e.g., on a single computer or network), but alternatively, DataAgent, GeneralNET, PersonalNET and ArbiterAgent can each be included on different computer-readable media on different computers, processors or devices, or can be included on one or more computer-readable media on different computers, processors or other devices in any desired combinations (e.g., DataAgent and GeneralNET on a single computer-readable medium, PersonalNET on a separate computer-readable medium and ArbiterAgent on a separate computer-readable medium).
Preferably, DataAgent, GeneralNET, PersonalNET and ArbiterAgent are contained in a single software program, but alternatively, commands for the functioning of DataAgent, GeneralNET, PersonalNET and ArbiterAgent can be contained in different software programs, or can be contained in one or more software programs in any desired combinations (e.g., the commands for DataAgent and GeneralNET on a single program, and the commands for PersonalNET and ArbiterAgent on a separate program).
An example of an apparatus corresponding to this specific embodiment is an inference engine to support decision-making in an intensive care unit (ICU). The present invention covers a variety of embodiments in any field of use where data needs to be evaluated, and is limited neither to the field of medical analysis nor to the specific handling of data as described in connection with the specific embodiment described below.
The inference engine of this example according to this embodiment helps an ICU staff to respond quickly to critical situations, predicts emergencies (e.g., recognizes pre-shock conditions and requires a doctor's attention at an early stage), and/or generates alarms to require an action (while minimizing any false alarms). That is, based on data from sensors, the inference engine can indicate a patient's condition and report probabilities of having one or more common types of emergency conditions (e.g., shocks). If probabilities are high, then the system may require an alarm. Similarly, based on tendencies of changes in readings, the system can predict a possible emergency condition which can occur in the near future.
This embodiment includes the following components:
The inference engine of this example corresponding to this embodiment receives substantially real-time data from one or more sensors, general data consisting of basic information about normal values for monitoring statistics (for example, blood pressure or heart rate), and patient data comprising information about the patient's history, the patient's current condition, etc. Optionally, the inference engine may also receive manually entered data (this data component is optional, i.e., the system is able to make decisions when there is no manual data).
The structure of the inference engine according to this example corresponding to this embodiment is depicted in
DataAgent receives information from one or more sensors, and/or from one or more data files, and places information in specific respective formats for receipt by GeneralNET, by PersonalNet and by ArbiterAgent.
DataAgent detects minor to severe increases or decreases of the incoming values from the one or more sensors and evaluates the data to model situations which can potentially occur. DataAgent interfaces with ArbiterAgent, GeneralNET and PersonalNET, the one or more sensors, a database with general information (referred to herein as “General Database”) such as acceptable ranges for many parameters (e.g., where the system is a patient, expert medical information), and a database with information regarding the particular system (referred to herein as “Personal Database”), e.g., where the system is a patient, medical information regarding the patient.
As discussed below in more detail, in this embodiment, DataAgent generates information in two types of modes, a real mode and a tendency mode.
In the real mode of this embodiment, DataAgent receives information from sensors, compares it with information in General Database and/or information in Personal Database, and generates a real mode record and a real mode output. In the real mode, the status is characterized as “normal” if a statistical value is within the acceptable range; the status is characterized as “high” if the statistical value is higher than the maximum value in the acceptable range; and the status is declared “low” if the statistical value is lower than the minimum value in the acceptable range.
In the tendency mode of this embodiment, DataAgent collects information from the sensors during a period of time (every T seconds) and compares it to basic records. Basic records, described below in more detail, function as baseline information for comparing later-obtained data. Based on such comparisons, DataAgent creates current records, decreasing current records and/or increasing current records (described below in more detail). If significant changes occur and a potential problem is detected, then DataAgent creates respective status records for each of GeneralNET and PersonalNET, and generates a status/color code record for ArbiterAgent.
In this embodiment, the tendency mode can be turned on or off, but the real mode is always on so that DataAgent always sends real data. The DataAgent processes data every T seconds.
The following discussion details analysis of data for a single parameter—other parameters for a system may be handled in an analogous way.
In this example according to this embodiment, when a new patient arrives at an ICU, as discussed below, in most cases (i.e., unless a significant number of data values for the parameter being monitored are and remain outside an acceptable range) a basic record is created with respect to the real-time data received from each of one or more sensors (e.g., where blood pressure, heart rate, temperature and CO sensors are each collecting data, a blood pressure basic record will be created, a heart rate basic record will be created, a temperature basic record will be created and a CO basic record will be created). Each of these records includes data set forth in a specific format, discussed in more detail below. In addition, a basic image may be created, which includes all of the basic records (e.g., where blood pressure, heart rate, temperature and CO sensors are each collecting data, the basic image includes the blood pressure basic record, the heart rate basic record, the temperature basic record and the CO basic record). The basic image includes its data set forth in a specific format, as discussed in more detail below. As discussed in more detail below, in specific circumstances (e.g., when the patient is no longer presenting with a particular symptom), the basic image (or one or more basic records) for the patient can be updated (i.e., replaced with a new basic image generated using current data).
In this example corresponding to this embodiment, a MIN/MAX/Low/High/Acceptable detection routine is performed for each parameter (e.g., medical parameters) being monitored and/or evaluated. The MIN/MAX/Low/High/Acceptable detection routine(s) are preferably performed when a patient first arrives at an ICU, but may instead be performed anytime thereafter. For each parameter being monitored or evaluated (e.g., heart rate, blood pressure, etc.), a set of data values is received from the sensor for that parameter. Each data value is the value of the parameter detected by the sensor during a sub-interval (e.g., one second) within an initial time interval (e.g., five minutes). Each sub-interval within the initial time interval is a period of time such that the sub-intervals taken together sequentially comprise an entirety of the initial time interval (e.g., 300 sequential one second sub-intervals within an initial time interval of five minutes).
Then, each of the data values (among the set of data values for this parameter) is assigned to one category selected from among categories characterized as “MIN”, “Low”, “Acceptable”, “High” and “MAX”. Each data value (i.e., if any) which is less than a minimum possible value is characterized as “MIN”. Each data value which is greater than a maximum possible value is characterized as “MAX”. Each data value which is within an acceptable range of values is characterized as “Acceptable”. Each data value which is greater than or equal to the minimum possible value and less than the lowest value in the acceptable range is characterized as “Low”. Each data value which is less than or equal to the maximum possible value and greater than the highest value in the acceptable range is characterized as “High”.
The effect of assigning data values less than a minimum possible value to “MIN” and assigning data values greater than a maximum possible value to “MAX” according to the present embodiment is to eliminate such readings because they are clearly erroneous (e.g., a negative heart rate or a heart rate which is larger than any possible human heart rate). This exclusion is done to eliminate data which is clearly incorrect, e.g., due to sensor error.
In accordance with a preferred aspect according to the present invention, if the quantity or percentage of data values (from among the set of data values) which are assigned to “MIN” (the “MIN value”) for the parameter exceeds a “replace sensor-min” value (e.g., 10% of the total quantity of signals), a signal is generated which indicates that the sensor which generated the readings for that parameter should be replaced. “Replace sensor-min” values may differ depending on the particular parameter. Similarly, in a preferred aspect, if the quantity or percentage of data values (from among the set of data values) which are assigned to “MAX” (the “MAX value”) for the parameter exceeds a “replace sensor-max” value, a signal is generated which indicates that the sensor which generated the readings for that parameter should be replaced. “Replace sensor-max” values also may differ depending on the particular parameter.
The acceptable range of values described above is defined as a range of values between and including a lowest acceptable value and a highest acceptable value. The lowest acceptable value is a reference value which is considered by experts in the appropriate field to be the minimum normally acceptable value, (e.g., in the case of heart rate, 60 beats per minute). The highest acceptable value is a reference value which is considered by experts to be the maximum normally acceptable value, (e.g., in the case of heart rate, 100 beats per minute).
Next, a determination is made as to whether either of the categories “Low” and “High” has the most data values. If “Low” has more data values than each of “Acceptable” and “High”, a color code indicator is characterized as “red”, a status indicator is characterized as “decreasing”, and another MIN/MAX/Low/High/Acceptable detection routine is then performed.
If “High” has at least as many data values as “Low”, and more data values than “Acceptable”, the color code indicator is characterized as “red”, the status indicator is characterized as “increasing” and another MIN/MAX/Low/High/Acceptable detection routine is then performed.
Otherwise (i.e., unless “Low” has more data values than each of “High” and “Acceptable”, or “High” has more data values than “Acceptable” and at least as many data values as “Low”), a basic record creation routine is then performed. In the basic record creation routine, the range of values within the “Acceptable” category for the parameter is divided into a number (e.g., 30, 40 or 60) of range segments. Preferably, the range segments each cover ranges of substantially equal size—for example, if the parameter is heart rate and acceptable values are within the range of from 60 to 100 (beats per minute), and the number of range segments is 41, each range segment covers a range of one value (i.e., 60, 61, 62, 63, etc.); if the parameter is blood pressure and acceptable values are within the range of from 100 to 120, and the number of range segments is 40, each range segment covers a range of about 0.5 (e.g., 100.0 to 100.5, 100.6 to 101.0, 101.1 to 101.5, etc.). Then, for each of the range segments, a quantity of the data values which fall within the range segment is determined. Then a range segment having a maximum quantity of the data values is identified and characterized as Rmax. Rmax is assigned to a final range category. Then, a next range segment analysis is performed.
In a next range segment analysis, a next lower range segment is the range segment having values just below the lowest value within the final range, and a next higher range segment is the range segment having values just above the highest value in the final range (if all range segments have already been included in the final range, the next range segment analysis is concluded and, as described below, a final range comparison is performed next). If the next lower range segment has a greater quantity of data values than the quantity of data values in the next higher range segment (or if there are no remaining range segments having values higher than the final range), a determination is made as to whether the next lower range segment has a quantity of data values which is at least a minimum ratio (e.g., 0.7 or any other desired ratio) of the quantity of data values in Rmax; if so, the next lower range segment is added to the final range category and another next range segment analysis is performed next. If the next lower range segment does not have a greater quantity of data values than the quantity of data values in the next higher range segment (or if there are no remaining range segments having values lower than the final range), a determination is made as to whether the next higher range segment has a quantity of data values which is at least the minimum ratio of the quantity of data values in Rmax; if so, the next higher range segment is added to the final range category and another next range segment analysis is performed next.
If the next lower range segment and the next higher range segment each have a quantity of data values which is less than the minimum ratio of the quantity of data values in Rmax, or if there is no next lower range segment and the next higher range segment has a quantity of data values which is less than the minimum ratio of the quantity of data values in Rmax, or if there is no next higher range segment and the next lower range segment has a quantity of data values which is less than the minimum ratio of the quantity of data values in Rmax, or if all range segments have already been included in the final range, a final range comparison routine is then performed.
In the final range comparison routine, a determination is made as to which of the range segments have a quantity of data values which is at least the minimum ratio of the quantity of data values in Rmax. Then a determination is made as to whether every range segment which has a quantity of data values which is at least the minimum ratio of the quantity of data values in Rmax is included in the final range. If not, the number of range segments is decreased by one and another basic record creation routine is performed next.
An example of an embodiment of an algorithm for determining the segments included in the “final range” is depicted in
If every range segment which has a quantity of data values which is at least the minimum ratio of the quantity of data values in Rmax is included in the final range, a determination is made as to the quantity of data values (from among the set of data values) falling within a Low range (the “Low value”), the quantity of data values falling within an X range (the “X value”), the quantity of data values falling within a Y range (the “Y value”), the quantity of data values falling within a Z range (the “Z value”) and the quantity of data values falling within a High range (the “High value”). The Low range includes all values which are not less than the minimum possible value and which are less than the lowest value in the acceptable range. The X range includes all values which are not less than the lowest value in the acceptable range and which are less than the lowest value in the final range. The Y range includes all values which are within the final range. The Z range includes all values which are greater than the highest value in the final range and which are not greater than the highest value in the acceptable range. The High range includes all values which are not greater than the maximum possible value and which are greater than the highest value in the acceptable range.
Optionally, in addition, a determination may be made as to the quantity of data values assigned to the “MIN” category (the MIN value) and the quantity of data values assigned to the “MAX” category (the MAX value).
For each set of data values relating to a specific parameter, a basic record is created. Each basic record contains the following components:
Optionally, the basic record may also include:
A similar basic record is created for each other parameter being detected. The basic records for the different parameters together make up a basic image.
After creating a basic record for a parameter, next, a current record creation routine is performed for that parameter. In the current record creation routine, a fresh set of data values for the parameter is received from the sensor for that parameter. As above, the following discussion focuses on the analysis of data for a single parameter—each of the other parameters can be handled in an analogous way. Each data value among the fresh set of data values is the value of the parameter detected by the sensor during a sub-interval (e.g., one second) within a current time interval (e.g., one minute). Each current time interval preferably follows immediately after the end of the previous time interval (i.e., immediately after a previous current time interval or, where there have been no previous current time intervals, immediately after the initial time interval). Each sub-interval within a current time interval is a period of time such that the sub-intervals taken together sequentially comprise an entirety of the current time interval (e.g., 60 sequential one second sub-intervals within a current time interval of one minute).
Then, a Low value, an X value, a Y value, a Z value and a High value are determined. The Low value is the quantity of data values within the fresh set of data values which fall within the Low range. The X value is the quantity of data values within the fresh set of data values which fall within the X range. The Y value is the quantity of data values within the fresh set of data values which fall within the Y range. The Z value is the quantity of data values within the fresh set of data values which fall within the Z range. The High value is the quantity of data values within the fresh set of data values which fall within the High range.
Optionally, in addition, a determination may be made as to the quantity of data values assigned to the “MIN” category (the MIN value) and the quantity of data values assigned to the “MAX” category (the MAX value).
If the Low value is greater than each of the X value, the Y value, the Z value and the High value, the color code indicator is characterized as “red”, the status indicator is characterized as “decreasing” and a decreasing current record creation routine (described below) is then performed.
If the High value is greater than or equal to the Low value, and is greater than each of the X value, the Y value and the Z value, the color code indicator is characterized as “red”, the status indicator is characterized as “increasing” and an increasing current record creation routine (described below) is then performed.
If the X value is greater than or equal to the Low value and greater than or equal to the High value, and is greater than each of the Y value and the Z value, the color code indicator is characterized as “yellow”, the status indicator is characterized as “decreasing” and a decreasing current record creation routine is then performed.
If the Z value is greater than or equal to the Low value, greater than or equal to the X value, greater than or equal to the High value, and greater than the Y value, the color code indicator is characterized as “yellow”, the status indicator is characterized as “increasing” and an increasing current record creation routine is then performed.
If the Y value is greater than or equal to the Low value, greater than or equal to the X value, greater than or equal to the Z value and greater than or equal to the High value, the color code indicator is characterized as “green”, the status indicator is characterized as “normal” and another current record creation routine is performed next.
For each fresh set of data values relating to a specific parameter, a current record is created. Each current record contains the following components:
Optionally, the current record may also include:
Optionally, the current record may also include:
In performing a decreasing current record creation routine (referred to above), a new fresh set of data values for the parameter is received from the sensor for that parameter. As above, the following discussion focuses on the analysis of data for a single parameter—each of the other parameters can be handled in an analogous way. Each data value among the new fresh set of data values is the value of the parameter detected by the sensor during a sub-interval (e.g., one second) within a current time interval (e.g., one minute). Each current time interval preferably follows immediately after the end of the previous current time interval. Each sub-interval within a current time interval is a period of time such that the sub-intervals taken together sequentially comprise an entirety of the current time interval (e.g., 60 sequential one second sub-intervals within a current time interval of one minute).
Then, a new Low value, an X1 value, an X2 value, an X3 value, a new Y value, a new Z value and a new High value are determined. The new Low value is the quantity of data values within the new fresh set of data values which fall within the Low range. The X1 value is the quantity of data values within the new fresh set of data values which fall within a first portion of the X range. The X2 value is the quantity of data values within the new fresh set of data values which fall within a second portion of the X range, the second portion of the X range including values which are greater than values in the first portion of the X range. The X3 value is the quantity of data values within the new fresh set of data values which fall within a third portion of the X range, the third portion of the X range including values which are greater than values in the second portion of the X range. In a preferred representative example, the third portion of the X range is the highest approximately 50% of the values within the X range, the first portion of the X range is the lowest approximately 30% of the values within the X range, and the second portion of the X range is the remaining approximately 20% of the values within the X range. The new Y value is the quantity of data values within the new fresh set of data values which fall within the Y range. The new Z value is the quantity of data values within the new fresh set of data values which fall within the Z range. The new High value is the quantity of data values within the new fresh set of data values which fall within the High range.
Optionally, in addition, a determination may be made as to the quantity of data values in the new fresh set of data values assigned to the “MIN” category (the MIN value) and the quantity of data values in the new fresh set of data values assigned to the “MAX” category (the MAX value).
Then, if the new High value is greater than each of the new Low value, the X1 value, the X2 value, the X3 value, the new Y value and the new Z value, the color code indicator is characterized as “red”, the status indicator is characterized as “increasing” and an increasing current record creation routine is performed next.
If the new Low value is greater than or equal to the new High value, and is greater than each of the X1 value, the X2 value, the X3 value, the new Y value and the new Z value, the color code indicator is characterized as “red”, the status indicator is characterized as “decreasing” and another decreasing current record creation routine is performed next.
If the new Z value is greater than or equal to the new High value and greater than or equal to the new Low value, and is greater than each of the X1 value, the X2 value, the X3 value and the new Y value, the color code indicator is characterized as “yellow”, the status indicator is characterized as “increasing” and an increasing current record creation routine is performed next.
If the X1 value is greater than or equal to the new Low value, greater than or equal to the new High value and greater than or equal to the new Z value, and is greater than each of the X2 value, the X3 value and the new Y value, the color code indicator is characterized as “orange”, the status indicator is characterized as “decreasing” and another decreasing current record creation routine is performed next.
If the X2 value is greater than or equal to the new Low value, greater than or equal to the new High value, greater than or equal to the new Z value and greater than or equal to the X1 value, and is greater than each of the X3 value and the new Y value, the color code indicator is characterized as “yellow”, the status indicator is characterized as “decreasing” and another decreasing current record creation routine is performed next.
If the X3 value is greater than or equal to the new Low value, greater than or equal to the new High value, greater than or equal to the new Z value, greater than or equal to the X1 value, greater than or equal to the X2 value, and greater than the new Y value, the color code indicator is characterized as “yellow”, the status indicator is characterized as “decreasing” and another decreasing current record creation routine is performed next.
If the new Y value is greater than or equal to the new Low value, greater than or equal to the X1 value, greater than or equal to the X2 value, greater than or equal to the X3 value, greater than or equal to the new Z value and greater than or equal to the new High value, the color code indicator is characterized as “green”, the status indicator is characterized as “normal” and another current record creation routine is performed next.
For each new fresh set of data values relating to a specific parameter, a decreasing current record is created. Each decreasing current record contains the following components:
Optionally, the decreasing current record may also include:
Optionally, the decreasing current record may also include:
In performing an increasing current record creation routine (referred to above), a current fresh set of data values for the parameter is received from the sensor for that parameter. As above, the following discussion focuses on the analysis of data for a single parameter—each of the other parameters can be handled in an analogous way. Each data value among the current fresh set of data values is the value of the parameter detected by the sensor during a sub-interval (e.g., one second) within a current time interval (e.g., one minute). Each current time interval preferably follows immediately after the end of the previous current time interval. Each sub-interval within a current time interval is a period of time such that the sub-intervals taken together sequentially comprise an entirety of the current time interval (e.g., 60 sequential one second sub-intervals within a current time interval of one minute).
Then, a current Low value, a current X value, a current Y value, a Z1 value, a Z2 value, a Z3 value and a current High value are determined. The current Low value is the quantity of data values within the current fresh set of data values which fall within the Low range. The current X value is the quantity of data values within the current fresh set of data values which fall within the X range. The current Y value is the quantity of data values within the current fresh set of data values which fall within the Y range. The Z1 value is the quantity of data values within the current fresh set of data values which fall within a first portion of the Z range. The Z2 value is the quantity of data values within the current fresh set of data values which fall within a second portion of the Z range, the second portion of the Z range including values which are greater than values in the first portion of the Z range. The Z3 value is the quantity of data values within the current fresh set of data values which fall within a third portion of the Z range, the third portion of the Z range including values which are greater than values in the second portion of the Z range. In a preferred representative example, the first portion of the Z range is the lowest approximately 50% of the values within the Z range, the third portion of the Z range is the highest approximately 30% of the values within the Z range, and the second portion of the Z range is the remaining approximately 20% of the values within the Z range. The current High value is the quantity of data values within the current fresh set of data values which fall within the High range.
Optionally, in addition, a determination may be made as to the quantity of data values assigned to the “MIN” category (the MIN value) and the quantity of data values assigned to the “MAX” category (the MAX value).
Then, if the current Low value is greater than each of the current X value, the current Y value, the Z1 value, the Z2 value, the Z3 value and the current High value, the color code indicator is characterized as “red”, the status indicator is characterized as “decreasing” and a decreasing current record creation routine is performed next.
If the current High value is greater than or equal to the current Low value, and is greater than each of the current X value, the current Y value, the Z1 value, the Z2 value and the Z3 value, the color code indicator is characterized as “red”, the status indicator is characterized as “increasing” and another increasing current record creation routine is performed next.
If the current X value is greater than or equal to the current Low value and greater than or equal to the current High value, and is greater than each of the current Y value, the Z1 value, the Z2 value and the Z3 value, the color code indicator is characterized as “yellow”, the status indicator is characterized as “decreasing” and a decreasing current record creation routine is performed next.
If the Z3 value is greater than or equal to the current Low value, greater than or equal to the current High value and greater than or equal to the current X value, and is greater than each of the current Y value, the Z1 value and the Z2 value, the color code indicator is characterized as “orange”, the status indicator is characterized as “increasing” and another increasing current record creation routine is performed next.
If the Z2 value is greater than or equal to the current Low value, greater than or equal to the current High value, greater than or equal to the current X value and greater than or equal to the Z3 value, and is greater than each of the current Y value and the Z1 value, the color code indicator is characterized as “yellow”, the status indicator is characterized as “increasing” and another increasing current record creation routine is performed next.
If the Z1 value is greater than or equal to the current Low value, greater than or equal to the current High value, greater than or equal to the current X value, greater than or equal to the Z3 value, greater than or equal to the Z2 value, and greater than the current Y value, the color code indicator is characterized as “yellow”, the status indicator is characterized as “increasing” and another increasing current record creation routine is performed next.
If the current Y value is greater than or equal to the current Low value, greater than or equal to the current X value, greater than or equal to the Z1 value, greater than or equal to the Z2 value, greater than or equal to the Z3 value and greater than or equal to the current High value, the color code indicator is characterized as “green”, the status indicator is characterized as “normal” and another current record creation routine is performed next.
For each current fresh set of data values relating to a specific parameter, an increasing current record is created. Each increasing current record contains the following components:
Optionally, the increasing current record may also include:
Optionally, the increasing record may also include:
In the above discussion of this embodiment, various choices have been made regarding how the embodiment proceeds in the event that two or more values are equal (e.g., in the event that the X value and the Z value are equal, according to the discussion of the embodiment set forth above, the Z value controls, i.e., the status indicator is characterized as “increasing” and an increasing current record creation routine is performed next). However, in accordance with the present invention, the hierarchy as to which value controls in situations where any two or more values are equal can be set up in any other way (e.g., in the event that the X value and the Z value are equal, the X value can control, i.e., such that the status indicator would be characterized as “decreasing” and a decreasing current record creation routine would be performed next).
In accordance with the present embodiment, it is not necessary that any or all types of current records (i.e., including “current records”, “decreasing current records” and “increasing current records”) be retained after real mode output and optionally also tendency mode output (discussed in more detail below) are sent to the ArbiterAgent, the PersonalNET and the GeneralNET. For example, upon initiating a next routine, the previous current record, decreasing current record or increasing current record can optionally be discarded.
The current records, increasing current records and decreasing current records for all the different parameters being monitored or evaluated together make up a current image.
Below is a discussion of a second specific example of an embodiment according to this aspect of the present invention, this second example being referred to below as the second embodiment.
The second embodiment is similar to the first embodiment described above, except that:
In the second embodiment as described above:
For example, in one preferred aspect of the second embodiment:
In a second preferred aspect of the second embodiment,
The architecture of DataAgent according to the embodiments described above is depicted in
In accordance with a preferred aspect of the present invention, DataAgent provides the option of automatically updating the basic record for a particular parameter when the color code indicator for that parameter has been yellow for a particular number of consecutive times (e.g., 30 times), which number may vary depending on the particular parameter, and/or when the color code indicator for that parameter has been orange for a particular number of consecutive times (which number may vary depending on the particular parameter). In addition, the number of consecutive times which prompts the basic record to be updated can differ depending on whether the yellow or orange color code is caused by elevated values (i.e., where the status indicator is “increasing”) or depressed values (i.e., where the status indicator is “decreasing”). In order to update the basic record, the basic record creation sub-routine is performed using fresh data from the sensor or sensors.
In accordance with a further preferred aspect of the present invention, DataAgent provides the option of automatically updating the basic record for a particular parameter when the values for that parameter have been within a range of a particular size (i.e., in which the highest value differs from the lowest value by not more than a particular amount), such that the basic range (and/or the acceptable range) can be modified so as to encompass such values. For example, even if the values have been consistently of quantities at which the color code is red, the basic range can be modified so as to encompass such values. Thus, where the acceptable range for heart rate is from 60 to 100 beats per minute, and the patient values have consistently been about 103, the acceptable range and the basic range for the heart rate parameter for the patient can be set such that they encompass 103 beats per minute.
Preferably, in addition, for each new fresh set of data values and for each current fresh set of data values (i.e., whenever a decreasing current record or an increasing current record is created), DataAgent also creates a temporary record.
For each new fresh set of data values, i.e., when DataAgent is performing a decreasing current record creation routine, the temporary record which is created is a decreasing temporary record. A decreasing temporary record includes the parameter's name (e.g., heart rate) or identifier (e.g., a code or ID number), an N value, the X1 value, the X2 value, the X3 value and a C value. The X1 value, the X2 value and the X3 value are the same as those described above in connection with the decreasing current record creation routine. In a decreasing temporary record, the N value is equal to the new Low value determined in the decreasing current record creation routine. In a decreasing temporary record, the C value is equal to the sum of the new Y value, the new Z value and the new High value determined in the decreasing current record creation routine.
For each current fresh set of data values, i.e., when DataAgent is performing an increasing current record creation routine, the temporary record which is created is an increasing temporary record. An increasing temporary record includes the parameter's name (e.g., heart rate) or identifier (e.g., a code or ID number), an N value, the Z1 value, the Z2 value, the Z3 value and a C value. The Z1 value, the Z2 value and the Z3 value are the same as those described above in connection with the increasing current record creation routine. In an increasing temporary record, the N value is equal to the sum of the new Low value, the new X value and the new Y value determined in the increasing current record creation routine. In an increasing temporary record, the C value is equal to the new High value determined in the increasing current record creation routine.
Upon returning to a green color code, preferably, any and all temporary records are discarded.
In each of the embodiments described above, for each current record, each decreasing current record and each increasing current record, DataAgent also generates a real mode record. The real mode record for a particular parameter includes a reading (“high”, “normal” or “low”, and/or an identifier which signifies one of such readings, e.g., 0 for normal, 1 for high or increased and 2 for low or decreased), a color code, as well as the parameter's name (e.g., heart rate) or identifier (e.g., a code or ID number). If the color code indicator for a parameter is red and the status is “increasing”, the parameter is given a “high” reading, and the color code is red. If the color code for the parameter is red and the status is “decreasing”, the parameter is given a “low” reading, and the color code is red. Otherwise (i.e., if the color code indicator is green, yellow or orange), the parameter is given a “normal” reading, and the color code is green. The real mode record is sent to ArbiterAgent, preferably along with a mode value indicating that it is a real mode record.
In addition, DataAgent preferably also generates a real mode output which is preferably sent to PersonalNET and GeneralNET, the real mode output preferably including the parameter's name (e.g., heart rate) or identifier and the reading (“high”, “low” or “normal”-determined in the same way as described above with regard to the real mode record).
In each of the embodiments described above, for each current record, each decreasing current record and each increasing current record, DataAgent preferably also generates a status/color code record. The status/color code record for a particular parameter preferably includes a status value (“increasing”, “decreasing” or “normal”), as well as the color code indicator (green, yellow, orange or red) and the parameter's name (e.g., heart rate) or identifier.
The status/color code record is sent to ArbiterAgent (preferably along with a tendency mode value indicating that it is a tendency mode reading).
In addition, one or more status records are preferably generated, which are preferably sent to GeneralNET and to PersonalNET. Preferably, a first status record includes (1) the parameter's name or identifier, and (2) a reading of “normal” if the color code indicator is green or yellow, a reading of “high” if the color code indicator is orange or red and the status is “increasing”, or a reading of “low” if the color code indicator is orange or red and the status is “decreasing”. The first status record is preferably used by GeneralNET and by PersonalNET to generate “orange” reports. Preferably, a second status record includes (1) the parameter's name or identifier, and (2) a reading of “normal” if the color code indicator is green, a reading of “high” if the color code indicator is yellow, orange or red and the status is “increasing”, or a reading of “low” if the color code indicator is yellow, orange or red and the status is “decreasing”. The second status record is preferably used by GeneralNET and by PersonalNET to generate “yellow” reports.
The color codes can be interpreted as follows:
In accordance with a preferred aspect of the present invention, a tracking record can optionally be produced. A tracking record can include information as to the sequence of the ranges over time which had the maximum quantity of data values. For example, a tracking record might read “Y range, Y range, Y range, Y range, Z range, Z range, Z range, Z range, Z1 range, Z2 range, Z2 range, Z2 range, Z2 range, Z2 range, Z2 range, Z2 range, Z1 range, Z1 range, Z1 range, Z2 range, Z2 range, Z2 range, Z3 range, Z3 range, Z3 range, High range, High range, High range, Z3 range, Z3 range, Z2 range, Z2 range, Z1 range, Z1 range, Y range, Y range, Y range, Y range, Y range, Y range, Y range, Y range, Y range, Y range, etc.” Alternatively or additionally, the tracking record (or records) can include any other information, e.g., the quantity of data values in the range which had the maximum quantity of data values, and/or the quantity of data values in any other or all of the ranges.
The GeneralNET component of the inference engine according to the embodiments described above is designed based on experts' knowledge of the system being evaluated or monitored. For example, where the system is a human whose medical condition is being monitored, expert data can include information based on knowledge from the medical domain and preferably the use of a Bayesian network frame. Input to GeneralNET can be in the form of expert data contained in a case file or a separate formatted case. GeneralNET receives and evaluates the expert data, as well as real mode outputs and status records generated by DataAgent to generate expert reports, which show probabilities of a number of possible emergency conditions. For example, GeneralNET might recognize a pattern, e.g., of particular readings going up while other readings are going down, which is characteristic of one or more specific problems, and generate a probability as to whether that problem or problems exist. GeneralNET sends the expert reports to ArbiterAgent.
In the case where GeneralNET is in the form of a Bayesian network, the Bayesian network is set up in any suitable format, a variety of which are well known to those of skill in the art. For example, in accordance with one format, each parameter or condition has associated with it a table of values, and the table can further include one or more columns which each indicate whether a particular condition exists, and a plurality of probability columns. Each row of each probability column indicates the probability that the parameter or condition to which the table applies will fall within a corresponding range (for example, the ranges can include “high”, “normal”, and “low”) for that column when the condition or each of the conditions in that row are either present or not present as indicated by the responses in the respective columns. For example, a representative table is shown below for a parameter:
The first row indicates that where condition X is present and condition Y is not present, the probability of the parameter (e.g., heart rate, blood pressure, etc.) being “high” is 5%, the probability of the parameter being “normal” is 5%, and the probability of the parameter being “low” is 90%. As described above, Data Agent can provide information as to whether any particular parameter (e.g., heart rate) is within any particular range, e.g., “high”, “normal” and “low”. In addition, when operating in an “orange” mode, the parameter can be considered to be:
The PersonalNET component of the embodiments described above is designed based on expert knowledge and the particular system's integrated case. PersonalNET is developed preferably on the base of a Bayesian network frame. PersonalNET receives expert data (described above) and situation data, as well as real mode outputs and status records (as described above) from DataAgent, and it generates expert/situation reports. Situation data includes information which pertains to the specific system being monitored or evaluated. For example, where the system is a patient whose medical condition is being monitored, when the patient arrives at the ICU, PersonalNET learns about that patient and creates a probabilities table for that patient, e.g., by taking the following steps:
Input to PersonalNET can be in the form of a case file or a separate formatted case.
PersonalNET generates a report which is based on knowledge about the system (e.g., knowledge about the patient's condition when the patient arrives at the ICU) and PersonalNET evaluates the real mode outputs and the status records. The expert/situation report from PersonalNET shows the probabilities of possible emergency conditions. PersonalNET sends the expert/situation report to the ArbiterAgent.
When PersonalNET is in the form of a Bayesian network, the Bayesian network can be set up in any suitable format, and can function in a manner analogous to the manner described above with respect to when GeneralNET is in the form of a Bayesian network.
In the embodiments described above, PersonalNET can be updated by the request of ArbiterAgent. There are two methods to update PersonalNET:
PersonalNET receives a current status record as a new case and updates its probabilities table; or PersonalNET unlearns a previous case and then learns a new case. Preferably, PersonalNET is updated only when the system has significantly changed, or when a condition of the system has significantly changed. For example, if the system is a patient and the patient arrived in the ICU with Hypovolemia and currently has no signs of Hypovolemia, the patient's condition has changed, and PersonalNET can be updated for that patient.
The ArbiterAgent of the embodiments described above makes decisions based on real mode records and status/color code records from said DataAgent, expert reports (from GeneralNET) and expert/situation reports (from PersonalNET). ArbiterAgent performs the following functions:
ArbiterAgent of the inference engine according to the embodiments described above makes decisions and generates final reports based on the following information:
In the real mode of the embodiments described above, ArbiterAgent makes a final decision regarding each possible condition based on reports from GeneralNET and PersonalNET and the real mode records received from DataAgent. A separate method of decision-making is included in ArbiterAgent for each type of condition (e.g., shock). These methods are based on expert information, e.g., expertise of medical doctors, and are preferably adjusted after testing.
In the tendency mode of the embodiments described above, ArbiterAgent makes a final decision based on reports from GeneralNET and PersonalNET and the status/color code record from DataAgent. In this case, the problem is more complex and the severity of changes in readings becomes a very important factor. With active participation of domain experts, the patterns for each emergency condition are developed. Decisions are made based on comparing a summary of tendencies of changes for all readings and the pattern for a particular emergency condition.
In the embodiments described above, ArbiterAgent generates a final report, which includes:
In the embodiments described above, ArbiterAgent preferably requires an alarm when:
Abnormal values of the readings show an increase of the severity of the problem.
To avoid false alarms, a summary of probabilities for all readings is preferably considered before an alarm is requested. When abnormal values of the readings continue to increase/decrease, the severity of the problem may be increasing and an alarm may therefore be required. However, if the value of the readings is abnormal, but steady, an alarm may be required once and a report indicating the problem can be generated continuously—but no further alarm is required.
In the embodiments described above, ArbiterAgent preferably requires an update of PersonalNet when the following information is received:
In the embodiments described above, generally, for a system operating in only the real mode (i.e., tendency mode is OFF), ArbiterAgent will make a decision to update PersonalNET for each emergency condition according to Table 1 shown below:
Where:
“Negative” means a low probability of having a particular emergency condition.
“Positive” means a low probability of having a particular emergency condition.
In the embodiments described above, ArbiterAgent preferably creates such tables for each emergency condition and requests update knowledge about each emergency condition separately. However, if the system supports tendency mode, then the results are also considered according to the modified table (Table 2) shown below:
Where:
“Negative” means low probability of having particular emergency condition.
“Positive” means low probability of having particular emergency condition.
“*” means regardless of the value of probability in particular case.
In the embodiments described above, ArbiterAgent preferably creates the tables described above for each emergency condition. For system supported tendency mode, ArbiterAgent requires updated information regarding each emergency condition separately according the table described above.
While the embodiments described above refer to specific examples where situation data is medical data regarding a patient, the present invention can be applied to any kind of situation data, e.g., data from sensors which monitor properties of or constituents in any system or element, and in which the expert data correspondingly is information relating to such system or element.
While the GeneralNET and PersonalNET components of the embodiments described above are Bayesian networks (which are conventionally known learning networks), either or both of these components can instead be any other learning network, e.g., a virtual vector machine, a neural network, a discrimination function, a decision tree, an expectation-maximization on mixtures of Gaussians, a probability distribution function (PDF), estimation through moment computation, PDF estimation through histograms, etc.
In addition, while the basic records, current records, decreasing current records, increasing current records, status/color code records, status records, real mode records and real mode outputs described above include specific data in specific respective formats, the types of data contained in each record and/or image can vary and can be in any desired format.
In addition, the time interval over which data is collected (and/or sub-intervals reflecting the period of time between each piece of data for a sensor) for generating a basic record can be selected from among a wide range of time periods, and/or can differ from one another for different parameters.
Also, the duration of time over which data is collected (and/or sub-intervals reflecting the period of time between each piece of data for a sensor) for generating current records, decreasing current records or increasing current records can be selected from among a wide range of time periods and/or can differ from one another.
Further, data (e.g., status, color code and/or readings) can be grouped differently for sending to other components (e.g., DataAgent can send PersonalNET a status/color code record which is similar to the status/color code record sent to ArbiterAgent).
Also, the number of range segments can be selected from among a wide range of quantities; in addition, it is not necessary that each of the segments have an equal range from lowest value to highest value.
The “minimum ratio” used in the basic record creation routine (i.e., the “minimum ratio of the quantity of data values in Rmax”) can be selected from within a wide range of quantities.
The percentages used in the decreasing current record creation routine (i.e., 30% for the first portion, 20% for the second portion and 50% for the third portion) and in the increasing current record creation routine (i.e., 50% for the first portion, 20% for the second portion and 30% for the third portion) can each vary from what is described in the above embodiments.
Also, the number of consecutive times for the color code to be yellow or the number of consecutive times for the color code to be orange for the option of automatically updating the basic record can be selected from among a wide range of values.
Any two or more functions can be conducted simultaneously, and/or any function can be conducted in a series of steps.
This application claims the benefit of U.S. Provisional Patent Application No. 60/578,277, filed Jun. 9, 2004, the entirety of which is incorporated herein by reference. This application claims the benefit of U.S. Provisional Patent Application No. 60/619,725, filed Oct. 18, 2004, the entirety of which is incorporated herein by reference. This application claims the benefit of U.S. Provisional Patent Application No. 60/620,191, filed Oct. 19, 2004, the entirety of which is incorporated herein by reference. This application claims the benefit of U.S. Provisional Patent Application No. 60/636,775, filed Dec. 15, 2004, the entirety of which is incorporated herein by reference. This application claims the benefit of U.S. Provisional Patent Application No. 60/642,379, filed Jan. 7, 2005, the entirety of which is incorporated herein by reference.
| Number | Date | Country | |
|---|---|---|---|
| 60578277 | Jun 2004 | US | |
| 60619725 | Oct 2004 | US | |
| 60620191 | Oct 2004 | US | |
| 60636775 | Dec 2004 | US | |
| 60642379 | Jan 2005 | US |
