Field of the Invention
The present invention generally relates to health care diagnosis and treatment, and more particularly to a method of evaluating information to determine the relevant medical history of a patient.
Description of the Related Art
Over the years medicine has become an increasingly complex science. In other to properly treat a patient, it is accordingly important to understand as much as possible about the patient's medical history. Much of this information can be gleaned from electronic documents, but there is also often a trail of paper (hard copy) records that should be examined. These can include a multitude of notes, forms and publications from different authors over a wide range of time.
While experienced doctors are still the best at determining a proper diagnosis and crafting appropriate therapies and responses, computer-based intelligent advisors such as Watson Oncology Advisor and Watson Oncology Expert Advisor have been developed to assist with these functions. Physicians, oncologists, and these intelligent advisors need an accurate representation of a patient's history in order to understand a patient's current state and to develop treatment plans for the future with the highest likelihood of success.
The present invention in at least one embodiment is generally directed to building a patient's medical history by receiving electronic documents pertaining to the patient's past health care, applying natural language processing to identify at least one medical concept and a date associated with the medical concept for each document, grouping the electronic documents based on the associated dates into document clusters, determining a primary concept for each document cluster including performing an analysis which assigns confidence values to each of the documents in a given cluster and selects the concept in the document having the highest confidence value as the primary concept, and combining primary concepts from respective document clusters to generate a combined history. If the combined history is not feasible due to a conflict between primary concepts from different clusters, the electronic documents can be re-grouped into different document clusters, and the analysis repeated for the different document clusters. The grouping can be performed in such a way as to make at least one of the clusters have at least two medical concepts which are the same. The analysis may include determining that a particular cluster has a minimum predefined number of documents, with the primary concept for the particular cluster appearing in a majority of the documents in the particular cluster. The analysis may also include removing one or more documents from a particular cluster. In an illustrative implementation, the medical concepts include a therapy concept type, a treatment concept type, and a diagnosis concept type. The invention can further identify an inter-concept conflict among the primary concepts involving at least two of the concept types that are different, then receive guidelines pertaining to relationships between the different concept types, and resolve the conflict by applying the relationships to select a different primary concept for at least one of the document clusters and thereby generate a different combined history.
The above as well as additional objectives, features, and advantages in the various embodiments of the present invention will become apparent in the following detailed written description.
The present invention may be better understood, and its numerous objects, features, and advantages of its various embodiments made apparent to those skilled in the art by referencing the accompanying drawings.
The use of the same reference symbols in different drawings indicates similar or identical items.
In health care, to properly treat a patient it is important to understand their entire medical history, including their current/past ailments, current/past treatments, and responses to these treatments. This history is difficult to piece together, as it is generally recorded across various documents written years apart by different authors with different perspectives, goals, and terminology.
Generally a patient case file has a list of documents which contain many different concept types (e.g., therapies received, diagnoses, responses, etc.). Over time, a patient's care may generate a numerous amount of clinical notes, which may have complex interdependencies, duplications of information, or omissions of information. For example, one doctor's “protocol A” may be the same treatment as a different doctor's “Treatment X,” and both may include drugs B, C, and D. As such, the patient's clinical notes may be more confusing than helpful to a caregiver, especially a caregiver that is new to providing care to this patient. The documents are generally not evenly distributed over time, but for concept mining there are patterns that can be exploited in these time-patterns.
One approach to this concept mining is set forth in U.S. patent application Ser. No. 14/514,563 filed Oct. 15, 2014,which is hereby incorporated. In that system, a therapy history timeline is built using documents with drug start dates, combined with correlations from guidelines to determine drug regimens and cycles. However, error detection is only achieved by eliminating drug references that directly conflict with an implied regimen, and this approach lacks a robust conflict resolution mechanism. Furthermore, this approach only considers one concept at a time (e.g., just therapy history).
It would, therefore, be desirable to devise an improved method of building a patient's medical history from disparate information sources. It would be further advantageous if the method could more reliably detect and resolve history conflicts. The present invention achieves these goals by correlating additional information sources (to improve accuracy) and by considering additional methods for rejecting false history entries. This process is preferably carried out in two parts or processes. In the first process, concepts can be ingested from documents using natural language processing NLP), with a frequency/weighting mechanism to filter out low-quality concepts (scoring) like one-time mentions and documents that give conflicting information. A series of time-boxed windows can be used to determine the most probable concepts within that window, with the scoring to filter out less-likely concept instances. The window sizes can vary, for example based on frequency of documents and expected size of window (e.g., for a treatment regimen, a window might be 6-12 months, which is the average length of a regimen). In the second process, concepts can be correlated into a history, including inter-concept relations (not just intra-concept relations). For example, 10-12 drugs are used in 90% of lung cancer cases—thus therapy history can be used to infer diagnosis history, or vice-versa. A series of relationships and inferences can then be invoked to determine how to best combine several different intra-concept histories into a single inter-concept history by scoring each concept history not just on how coherent it is by itself, but how well it fits with other concepts.
These two parts of the preferred implementation can be run serially, first as intra-concept correlation and then as inter-concept correlation. However, they can also be run in parallel, just meaning that more potential inter-concept histories are built.
With reference now to the figures, and in particular with reference to
MC/HB 16 has an interface to peripheral component interconnect (PCI) Express links 20a, 20b, 20c. Each PCI Express (PCIe) link 20a, 20b is connected to a respective PCIe adaptor 22a, 22b, and each PCIe adaptor 22a, 22b is connected to a respective input/output (I/O) device 24a, 24b. MC/HB 16 may additionally have an interface to an I/O bus 26 which is connected to a switch (I/O fabric) 28. Switch 28 provides a fan-out for the I/O bus to a plurality of PCI links 20d, 20e, 20f. These PCI links are connected to more PCIe adaptors 22c, 22d, 22e which in turn support more I/O devices 24c, 24d, 24e. The I/O devices may include, without limitation, a keyboard, a graphical pointing device (mouse), a microphone, a display device, speakers, a permanent storage device (hard disk drive) or an array of such storage devices, an optical disk drive which receives an optical disk 25 (one example of a computer readable storage medium) such as a CD or DVD, and a network card. Each PCIe adaptor provides an interface between the PCI link and the respective I/O device. MC/HB 16 provides a low latency path through which processors 12a, 12b may access PCI devices mapped anywhere within bus memory or I/O address spaces. MC/HB 16 further provides a high bandwidth path to allow the PCI devices to access memory 18. Switch 28 may provide peer-to-peer communications between different endpoints and this data traffic does not need to be forwarded to MC/HB 16 if it does not involve cache-coherent memory transfers. Switch 28 is shown as a separate logical component but it could be integrated into MC/HB 16.
In this embodiment, PCI link 20c connects MC/HB 16 to a service processor interface 30 to allow communications between I/O device 24a and a service processor 32. Service processor 32 is connected to processors 12a, 12b via a JTAG interface 34, and uses an attention line 36 which interrupts the operation of processors 12a, 12b. Service processor 32 may have its own local memory 38, and is connected to read-only memory (ROM) 40 which stores various program instructions for system startup. Service processor 32 may also have access to a hardware operator panel 42 to provide system status and diagnostic information.
In alternative embodiments computer system 10 may include modifications of these hardware components or their interconnections, or additional components, so the depicted example should not be construed as implying any architectural limitations with respect to the present invention. The invention may further be implemented in an equivalent cloud computing network.
When computer system 10 is initially powered up, service processor 32 uses JTAG interface 34 to interrogate the system (host) processors 12a, 12b and MC/HB 16. After completing the interrogation, service processor 32 acquires an inventory and topology for computer system 10. Service processor 32 then executes various tests such as built-in-self-tests (BISTs), basic assurance tests (BATs), and memory tests on the components of computer system 10. Any error information for failures detected during the testing is reported by service processor 32 to operator panel 42. If a valid configuration of system resources is still possible after taking out any components found to be faulty during the testing then computer system 10 is allowed to proceed. Executable code is loaded into memory 18 and service processor 32 releases host processors 12a, 12b for execution of the program code, e.g., an operating system (OS) which is used to launch applications and in particular the medical history builder application of the present invention, results of which may be stored in a hard disk drive of the system (an I/O device 24). While host processors 12a, 12b are executing program code, service processor 32 may enter a mode of monitoring and reporting any operating parameters or errors, such as the cooling fan speed and operation, thermal sensors, power supply regulators, and recoverable and non-recoverable errors reported by any of processors 12a, 12b, memory 18, and MC/HB 16. Service processor 32 may take further action based on the type of errors or defined thresholds.
The present invention may be a system, a method, and/or a computer program product. The computer program product may include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present invention.
The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of the computer readable storage medium includes the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.
Computer readable program instructions described herein can be downloaded to respective computing/processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and/or a wireless network. The network may comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and/or edge servers. A network adapter card or network interface in each computing/processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing/processing device.
Computer readable program instructions for carrying out operations of the present invention may be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like, and conventional procedural programming languages, such as the “C” programming language or similar programming languages. The computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) may execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present invention.
Aspects of the present invention are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer readable program instructions.
These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks. These computer readable program instructions may also be stored in a computer readable storage medium that can direct a computer, a programmable data processing apparatus, and/or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture including instructions which implement aspects of the function/act specified in the flowchart and/or block diagram block or blocks.
The computer readable program instructions may be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computer, other programmable apparatus, or other device implement the functions/acts specified in the flowchart and/or block diagram block or blocks.
The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts or carry out combinations of special purpose hardware and computer instructions.
Computer system 10 carries out program instructions for a medical history build process that uses novel correlation techniques to provide an improved patient profile. Accordingly, a program embodying the invention may include conventional aspects of various medical history tools, and these details will become apparent to those skilled in the art upon reference to this disclosure.
Referring now to
The documents can be ingested by computer system 10 using natural language processing (NLP). NLP is a known science which enables computers to derive meaning from human or natural language input. In some NLP methodologies, a text annotator program searches text in documents and analyze it relative to a defined set of tags. The front-end NLP can include identification of a lexical answer type and a focus, and creation of a common analysis structure. Lexical answer type, focus and common analysis structure are known features of the prior art. Those skilled in the art will appreciate that the present invention may be applied to other analysis techniques which can parse a natural language document which includes medical terminology.
In accordance with one implementation of the present invention,
Clustering of the documents can be performed by computer system 10 based on a variety of factors. For example, for building a therapy history date range, computer system can use the length of an average treatment (say, 6-12 months). Other time windows are possible, both longer and shorter. A domain expert could manually set the ideal cluster date range as an input to computer system 10, or a range can be inferred from supporting data about the concepts themselves. Sliding time windows are also possible, so a single document (history element) may be included in two different time clusters. Ideally, a cluster is formed so that at least one concept appears twice in that cluster (in the treatment example, two instances of the same therapy), so computer system 10 may adjust the cluster date range within predefined constraints to accommodate this goal. For example, computer system 10 may use a default cluster range of 6-12 months but if no concept appears twice in a cluster with this basis then the range might be adjusted to 3-15 months. If a document has a date range but no specific date, any reasonable date can be used such as the midpoint of the date range, but if the range is too wide (beyond some predetermined range like two years) then it can be omitted entirely.
Once the documents have been clustered, computer system 10 can perform an analysis to determine the most likely concept within a date group. Dominant concepts from each cluster can then be selected to produce a probable concept history, as seen in
The analysis used to determine the most appropriate concept in a cluster can again be performed by computer system 10 based on a variety of factors. For example, for a reasonably large cluster (i.e., having some minimum predefined number of documents N), if a concept appears in the majority of the documents that is the most probable concept for that cluster. If a clear favorite is not found according to such base criteria, the cluster can be culled, such as by removing concepts appearing only once in a cluster, or removing documents that support multiple candidate clusters. Computer system 10 can assign a confidence value for the favored concept within a cluster; for example, the confidence value could be the number of documents supporting the concept in the cluster divided by the number of total documents in the cluster. The best answers from the clusters are then combined into the probable concept history.
In some embodiments, this probable concept history is just a candidate or proposed history, and can be rejected. Computer system 10 can perform a further analysis to determine if a particular combined history is feasible. For example, with a therapy history, if regimens represented by the primary concepts are not spaced far enough apart in time according to relevant guidelines, then the proposed history can be rejected. For a diagnosis history, it would be possible that a diagnosis could progress from myelodysplastic syndromes (MDS) to acute myeloid leukemia (AML), but the diagnosis would never progress from AML to MDS. Another false diagnosis history could show a primary cancer first as lung cancer, one month later as breast cancer, and two weeks later as lung cancer, as the primary diagnosis would never change that fast.
If a candidate history is found to be unfeasible, the analysis can be repeated with a different set of clusters. For example, small clusters (i.e., one document, or below some predefined threshold) can be culled from the timeline, although exceptions to this rule can be made such as when the cluster is the most recent. Also, for history elements that generate the invalid history, their date cluster can be expanded or contracted. The process can be repeated until the best combined history is generated. Multiple candidate histories can be considered feasible; in such a case the one with the highest combined confidence values can be selected, or other criteria can be used to pick the best concept history.
The history building process can be understood with regard to two further examples. According to the first example, computer system 10 is trying to decide whether a patient's treatment history includes all of AB, CD, EF, GH, or some combination thereof, based on eight documents in the patient history. Document 1 indicates that the patient was treated with drug A in June of 2000. Document 2 states that the patient continued regimen AB with drug B in July of 2000. Document 3 suggests that a previous treatment was unsuccessful, and as of April 2002 (in the future) a new drug C will be administered. Document 4 asserts that if this treatment does not work, a new regimen EF will be given to the patient in June of 2002. Document 5 indicates that a doctor continued treatment by giving drug D in June of 2002. Document 6 notes that, in June of 2002, the patient complained that regimen CD is an even worse than regimen AB, and asked about switching to regimen EF. Document 7 shows that the patient started regimen GH in January of 2004. Finally, Document 8 indicates that the patient completed regimen GH in July of 2004 and achieved complete remission of symptoms. From these documents, computer system 10 can detect five possible regimens received: AB, CD, EF, AB (again), and GH. From guidelines provided to computer system (see U.S. patent application Ser. No. 14/514,563), it is known that only one of AB/CD/EF was actually given in 2002 since they conflict, even though there is evidence for all three. While these guidelines maintain that only one of the three treatments is possible, the prior art does not have any mechanism for picking the correct one. Systems such as that disclosed in U.S. patent application Ser. No. 14/514,563 are forced to simply make a random selection among AB/CD/EF. The present invention uses additional analysis to select the most appropriate history element. Computer system 10 will rank Document 6 as low quality since it is not recent (over two years old, with 25% of the documents newer than this), and regimen AB is a one-time mention within the cluster. Regimen EF is mentioned twice, however one mention is in the low-quality Document 6. Regimen CD is mentioned three times (including the low-quality Document 6). From this scoring, the patient received CD in 2002, not AB or EF. The complete concept history is therefore AB in 2000, CD in 2002, and GH in 2004.
According to the second example, the same patient has the same eight documents with a new Document 9 which indicates that the patient relapsed in late 2004 and immediately started on regimen IJ Even though IJ is a one-time mention, it is the most recent document and it should therefore be probable that IJ is part of the therapy history (noting also, it does not conflict with guidelines)
In the foregoing examples, a probable concept history is still not as complete of a solution as desired. Intra-concept history can generate several conflicting histories, especially if there are sparse numbers of documents supporting multiple hypotheses. Further analysis can be used to combine different concept histories into a coherent whole.
In order to better correlate the therapies with the diagnoses, computer system 10 can ingest a set of guidelines 60 seen in
The inter-concept guidelines can include a vary of relational bases to resolve low-confidence individual concept histories. A relationship may indicate how often one concept leads to a different concept (e.g., 90% of the mentions of a given therapy are related to a particular diagnosis). A relationship may indicate how often one concept progression influences a different concept progression (e.g., a history of regimen AB followed by CD and then EF typically happens when the disease metastasizes, and a secondary diagnosis is likely around the beginning of regimen EF). A relationship may indicate how one concept occurring means another concept should never occur (e.g., when a “failed treatment response” is found on Date X, a different regimen should be seen before and after Date X). The same therapy appearing some time span (say, at least 6 months) after the first occurrence of the therapy can indicate a recurrence of the disease.
In a further example, lung cancer guidelines indicate 10-12 drugs that are commonly used in 90% of cases. Referring back to the first text example above, it is presumed for this further example that regimen CD and regimen EF were similarly weighted for the 2002 history entry. From the guideline examination, computer system 10 finds that regimen CD correlates most strongly to lung cancer and regimen EF correlates most strongly to breast cancer. If the diagnosis history for the patient suggests lung cancer from 1999-2007 and then melanoma from 2010 onward, computer system 10 will conclude that regimen CD was most likely administered to this patient in 2002.
Additional cognition could be provided in the conflict resolution mechanism. For example, if the documents suggest an inconsistent timeline of therapies with diagnoses and there were two equally weighted choices, both yielding a similarly consistent final result, there are two approaches that could be implemented. First, it could be assumed that the diagnoses were correct in which case the interpretation of the therapies would be adjusted. Conversely, it could be assumed that the therapies were correct in which case the interpretation of the diagnoses would be adjusted. Machine-learning protocols could be used to identify over time which choice was the best based on the attributes of the patient case. It may be that most of the time when there are conflicts with lung cancer as the diagnosis, it is the therapies that are wrong, but for some rare cancer type (ex: ear cancer) it's the therapies that are usually right and the cancers usually wrong. Since there are so many possible combinations of therapy/diagnosis/response, a machine-learning implementation could help fine-tune the conflict resolution.
The invention may be further understood with reference to the charts of
The inter-concept correlation process 90 of
The present invention thereby allows a cognitive system to more accurately piece together a patient's medical history documents, and provides a robust resolution mechanism for intra-concept conflicts. Inter-concept correlations also increase the likelihood of developing a more coherent combined patient history.
Although the invention has been described with reference to specific embodiments, this description is not meant to be construed in a limiting sense. Various modifications of the disclosed embodiments, as well as alternative embodiments of the invention, will become apparent to persons skilled in the art upon reference to the description of the invention. For example, while the invention has been disclosed in conjunction with examples pertaining to cancer diagnoses and treatments, it is more generally applicable to any medical conditions, including mental health diagnoses. It is therefore contemplated that such modifications can be made without departing from the spirit or scope of the present invention as defined in the appended claims.