Patent Application
20090228298
The present disclosure relates to the field of physiological data analysis. More specifically the present disclosure relates to the detection and analysis of morphological features of physiological data.
Automatic and semi-automatic analysis of physiological data are important tools used in both medical and clinical research applications. In automatic analysis, one or more algorithms are applied to the physiological data to produce a computer generated interpretation and/or analysis of the physiological data. Semi-automatic analysis similarly applies one or more algorithms to the physiological data to produce a computer generated interpretation or analysis of the physiological data, but the computer generated interpretation is then presented to a clinician who reviews the interpretation and edits them on a computer screen according to the clinician's own review of the data and judgment in an interactive process.
Typically, analysis of physiological data can be performed by looking at either interval related features of the physiological data (i.e. the timing between features or events in the physiological data) or the morphology of the features in the physiological data (i.e. the shape or geometry of the features in the physiological data). Most automatic and semi-automatic physiological data analysis applications focus on interval related physiological data characteristics as these characteristics are easier to identify and quantify as opposed to the feature morphologies that are more subjective in detection and analysis. While algorithms exist for the detection and description of feature morphologies, these algorithms often produce outputs that consist of a prohibitively large number of parameters and typically express each of these parameters as a continuous value such as an integer or floating point value.
Therefore, the sheer number of morphological parameters and the continuous nature of the expression of each of these parameters make it difficult to use a semi-automatic physiological data analysis technique for the analysis of data feature morphology.
A system for the interactive analysis of morphological features of physiological data between computerized algorithms and review physicians is disclosed herein. In one embodiment, the system includes a morphological segment detection module that receives physiological data from a physiological data source and applies at least one morphological segment of the physiological data. The system further includes a segment feature rating module that applies at least one algorithm to the at least one identified morphological segment to identify at least one segment feature and produce a rating of the severity of the at least one segment feature.
Also disclosed herein is a method of analyzing physiological data morphology. The method includes the steps of receiving physiological data and identifying at least one morphological segment of the physiological data. The method further includes the steps of identifying at least one feature of each identified morphological segment and determining a feature rating for each identified feature.
a and 2b depict embodiments of the presentation of morphological features of a P-wave segment of electrocardiographic (ECG) data;
a and 3b depict morphological features of a QRS segment of ECG data;
a and 4b depict morphological features of a T-wave segment of ECG data;
a is a flow chart depicting the steps in an embodiment of a method of analyzing physiological data morphology;
b is a flow chart depicting the steps in an embodiment of a sub-method of analyzing physiological data morphology to allow clinician review and editing;
c is a flow chart depicting the steps in an embodiment of a sub-method of analyzing physiological data morphology to compare sets of ECG data; and
d is a flow chart depicting the steps in an embodiment of a sub-method of analyzing physiological data morphology to perform data mining analysis.
The detection and analysis of morphological features of physiological data is an important tool in both medical diagnosis and clinical research applications. One such application is the analysis of electrocardiographic data (ECG) which will herein be used in an exemplary manner; however, it should be understood that other types of physiological data such as, but not limited to, electromyography (EMG) and electroencephalography (EEG) may be aided by embodiments of the system and method as disclosed herein.
The ECG data from the ECG data source 12 is sent to a morphological segment detection module 18. The morphological segment detection module 18 receives the ECG data and applies at least one algorithm to the ECG data. The results of the application of the at least one algorithm is to identify at least one morphological segment of the ECG data. The morphological segments that may be identified in the ECG data may include the P-wave, the QRS complex, the ST interval, the T-wave, or the U-wave. It is understood that alternative embodiments analyzing other physiological data may detect different morphological segments intrinsic to the physiological data being analyzed. The algorithms applied by the morphological segment detection module 18 may include a series of morphology descriptors that identify each of the ECG segments. These descriptors may be used in conjunction with pattern recognition techniques to identify each of the segments.
Some embodiments herein disclosed may utilize one or more computers that apply one or more algorithms as disclosed herein to process data. The technical effect of these algorithms applied by at least one computer is to identify the morphological segments and segment features exhibited by the data and produce a rating of the identified segment features to simplify a clinician's review and editing of a computer determined analysis of a physiological signal.
The ECG data with the detected segments is then sent to a segment feature rating module 20. The segment feature rating module 20 applies at least one algorithm to at least one identified morphological segment of the ECG data. The application of the at least one algorithm to the at least one identified morphological segment produces a rating of the severity at least one segment feature. Each of the identified morphological segments may be broken into a number of segment features which may be used to describe the morphological segment. Each of the features may reflect a potential segment morphology that may be clinically relevant. A fuzzy clustering technique may be used to quantify the existence of the features in the morphological segment. These feature ratings may be quantified into discrete severity levels such as to produce a rating of the severity of any detected segment features.
In an embodiment, the discrete severity levels may include four levels represented by the numbers 0, 1, 2, and 3. These severity level ratings may coincide with no, moderate, obvious, and severe ratings for the existence of a particular segment feature. The severity levels for each feature are generated from statistical analysis of the baseline distribution for these segment features. The baseline distribution may be acquired from a large pool of ECG data as a part of one or more databases. From this baseline distribution, clustering and/or fuzzy logic grouping techniques may be applied to generate the discrete severity levels.
Embodiments of the physiological data analysis system 10 disclosed herein may include specific elements directed towards particular applications facilitated by the identification of a discrete severity level for identified segment features performed by the morphological segment detection module 18 and the segment feature rating module 20.
One embodiment of the system 10 may include a clinician review and editing sub-system 22 in which the ECG data with rated segment features is sent to an ECG display 24. The ECG data and the identified discrete severity levels for each identified segment feature are presented to a clinician.
a and 2b show an exemplary embodiment of the presentation of ECG data and the segment feature ratings as may be presented by the ECG display 24.
A segment feature rating region 38 of the GUI 30 includes indications of a plurality of segment features that may be identified within the morphological segment. An exemplary listing of the segment features may include, but is not limited to, Missing 40, Biphasic 42; Sharp 44; Long PR; and Short PR 48. The segment feature rating region 38 also includes a plurality of discrete levels 50 within which the segment features are rated. The discrete levels 50 may include “+” for moderate levels; “++” for obvious features; and “+++” for very severe features. In this fashion, each of the segment features may be indicated as being present or not present, and if they are present, then a discrete level of the severity of the feature is similarly presented.
In
The clinician is able to review each of the identified morphological segments for the ECG data. In one embodiment, this is performed by selecting a variety of tabs 32 that are each associated with a different morphological segment.
In
b depicts still further ECG data 70 with the QRS complex 72 highlighted. In this example, the QRS complex 72 exhibits a “very severe” Q-wave feature. It is indicated as such in the segment feature rating region 38 by highlighting the circle associated with the “very severe” segment feature rating. As described with respect to
Additionally,
In
b depicts still further ECG data 98 with the T-wave 96 highlighted. In this example, the T-wave 96 has been identified by the morphological feature analysis algorithm to exhibit “obvious” notch 82, flatness 84, and unsymmetrical 86 features as well as the same “moderate” U feature 88 found in ECG data 94. However, a comparison of the ECG data 94 and the ECG data 98 yields that the T-wave 95 appears to be very different from T-wave 96. In fact, the clinician, upon viewing the ECG 98 as presented by the GUI 30, may determine that the T-wave 96 of the ECG data 98 exhibits a “very severe” unsymmetrical feature as opposed to the computer determined “obvious” level of the unsymmetrical feature 86. The clinician may then select the T-wave segment 96 and change the unsymmetrical feature 86 rating level to that which the clinician determines to be more proper.
Similarly, upon a review of the ECG data 94 in comparison to the ECG data 98, the clinician may decide that T-wave 95 only presents a “moderate” unsymmetrical feature 86. The clinician may at that time choose to select the T-wave 95 segment and change the segment feature rating for the unsymmetrical feature 86 to identify that feature as being only “moderate”. Any clinician modifications that have been made may be saved to the ECG database 28 such that they may be available at a later time and add a remote location to a later reviewing clinician.
By presenting the morphological feature analysis as a plurality of discrete levels for each of the predetermined clinically relevant morphological features, the clinician's review of the ECG data is focused on those features. This helps the clinician to distill the multitudes of morphological feature data that may be produced by an automated system; and, therefore, enable the clinician to effectively interject his or her own clinical opinion into the automated morphological feature analysis result. This combination of both automated and clinician analysis of the ECG data thus yields a more accurate morphological feature analysis, capitalizing on the strengths of automated systems as well as clinician review and modification of those results.
It is understood that the clinician may select any of the tabs 32 of the GUI 30 to navigate to each of the other morphological segments, including the ST segment and the T-U segment. Upon selection of these alternative segment tabs, a similar segment feature rating region 38 would be brought up that includes segment features that are associated within or particular to the selected morphological segment. Also, the selected morphological segment would be highlighted on the display of ECG data below the segment feature rating region 38.
The clinician review and editing sub-system 22 of the physiological data analysis system 10 gives the reviewing clinician the ability to review and modify an analysis or interpretation of ECG data performed by the application of algorithms to the ECG data, similar to that which is already available with respect to interval based physiological data analysis. This promotes improved quality in the final analysis of the ECG data, as the clinician is assisted by the algorithm analysis, but can adjust the output to account for algorithm identified false positives and modifications.
Referring back to
In one embodiment of the ECG comparison module 69, the ECG comparison module 69 compares each of the feature ratings between the first ECG data and the second ECG data to determine the similarity and differences between the first ECG data and the second ECG data.
In a still further embodiment, the comparison between the first ECG data and the second ECG data may be performed by using a distance measure method wherein a numerical value is given to each of the discrete segment feature levels and the difference between the levels for each of the segment features is found. In one simple distance measure method, each of the differences are squared and summed. The square root of this summation is indicative of the overall difference between the two ECG signals and may be easily implemented by the application of this algorithm. It is also understood that other methods and/or algorithms may be used to provide a comparison between the first ECG data and the second ECG data as well. These alternative methods and/or techniques are considered within the scope of the present disclosure.
The data mining sub-system 74 of the physiological data analysis system 10 uses the ECG data with the rated segment features from the segment feature rating module 20 to create an improved data mining system 74. An ECG database populator 76 receives the ECG data with the rated segment features from the segment feature rating module 20. The ECG database populator 76 sorts the ECG data by the segment feature and the rated level for each segment feature. This sorted ECG data is then stored in an ECG database 78 wherein the sorted ECG data may be stored as a lookup table wherein the ECG data is tabulated by each segment feature and its rating severity level. A morphology feature based database search engine can be built by first generating a morphology index server. A data mining module 80 may access the index sever to search a specific segment feature and/or segment feature level very fast. This can easily and quickly allow the retrieval of a very specific data set comprising all of the ECG data that exhibits a specified segment feature and/or specified feature level.
Thus, the data mining system 74 can improve upon previous data mining systems in that sets of morphology based segregated ECG data may be easily acquired to enhance the application of data mining techniques that may be applied to the obtained data sets.
It should be understood that in the present disclosure the term module has been used to describe components of the physiological data analysis system 10. In the present disclosure, the term module is used to refer to a logical component of a system that is implemented in either hardware, software, or firmware that receives an input and produces an output.
Also disclosed herein is a method of analyzing physiological data morphology, as depicted in
After at least one segment feature has been identified in step 104, a severity level for the identified segment features is determined at step 106. The severity level for the identified segment features may be represented by a discrete number of levels upon which the severity of the identified segment features are rated. The severity level for each of the identified segment features may be determined by the degree in which the identified segment feature deviates from a specified baseline norm for that particular segment feature. The baseline may be calculated from an analysis of exemplary physiological data.
The determined severity levels for the identified segment features of step 106 in combination with the first physiological data may be utilized in a variety of alternative sub-method applications as represented by reference point 200. These sub-methods may include clinician review and modification of the physiological data 210; serial comparison between the first physiological data and other physiological data 220; and data mining applications 230.
Referring to
Referring to
Lastly,
At step 126, a data set is retrieved from the database created in step 124 that includes physiological data of a specified segment feature and level. The organization and grouping of the physiological data in the database created in step 124 facilitates the retrieval of these highly specified data sets in step 126. The data set retrieved in step 126 may then be used in step 128 to build a morphology feature based index server. The morphology based index server may be constructed in the form of a look-up table that allows for the selection and/or sequential ordering of sets of ECG data based on any of the identified morphological segment features stored with each of the sets of ECG data. It is understood, however, that other strategies for data organization and index server structure may be utilized in connection with the morphology feature based index server.
Finally, data mining techniques are applied in step 130 using the index server created in step 128. The application of data mining techniques may be facilitated by the specialized data sets that may be easily retrieved from the index server built in step 128 due to the organization and the grouping of the physiological data by the identified segment feature and the segment feature levels in the index server. Thus, the data mining techniques applied in step 130 may result in faster and more accurate results due to the efficiencies gained through the use of the morphology feature based index server.
One particular field in which the system and method as disclosed herein may be of particular relevance may be in the field of pharmaceutical cardiac safety testing. As pharmaceutical cardiac safety testing requirements increase, these tests may require more sophisticated analysis techniques that look not only at ECG data interval timing but also at ECG morphology changes, since the inclusion of ECG morphology analysis may yield a higher correlation with severe drug induced arrhythmia than simply ECG interval measurements alone. Therefore, a technique wherein clinicians are able to review ECG data and a series of computer determined segment features, segment feature severity levels, check the computer determined levels for accuracy, and modify the determined levels with the clinician's own interpretation of the ECG data would be beneficial in that the resulting ECG data with human annotated computer derived segment feature levels would be more accurate then that determined by the computer or the clinician alone.
This written description uses examples to disclose the invention, including the best mode, and also to enable any person skilled in the art to make and use the invention. The patentable scope of the invention is defined by the claims, and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims if they have structural elements that do not differ from the literal language of the claims, or if they include equivalent elements with insubstantial differences form the literal languages of the claims.
Various alternatives and embodiments are contemplated as being with in the scope of the following claims, particularly pointing out and distinctly claiming the subject matter regarded as the invention.
