Medical system, control method, and non-transitory computer-readable storage medium

Abstract

The present disclosure relates to a medical system, a control method and a non-transitory computer readable storage medium. A medical system capable of providing symptom inquiry explanation and / or disease diagnosis explanation, the medical system comprises an interface and a processor. The interface is configured to receive an input state. The processor is coupled to the interface, and the processor is configured to execute a symptom checker based on a neural network model to select a current action from a plurality of candidate symptom inquiries and a plurality of candidate diseases predictions according to the input state. If the current action is a first symptom inquiry, the processor executes an explanatory module interacting with the symptom checker to generate a diagnosis tree to simulate a plurality of potential diagnosis paths, and to generate a symptom inquiry explanation about the first symptom inquiry according to the diagnosis tree. The explanatory module is capable of generating a symptom inquiry explanation or a post-mortem analysis explanation, which helps to improve the user's understanding and trust in the medical system.

Description

Technical Field

[0001] This disclosure relates to a medical system capable of generating symptom inquiries during computer-aided diagnosis, and more particularly to an artificial intelligence-based medical system capable of providing explanatory information regarding symptom inquiries and disease prediction. Background Technology

[0002] With technological advancements, the concept of computer-assisted medical systems has emerged, enabling healthcare institutions to assist in patient diagnosis. These systems may request relevant information from patients, then interact with them by asking symptom-related questions and collecting their responses. Upon completion of the diagnostic process, the system will provide a diagnosis or recommendation for potential illnesses (or appointment suggestions with relevant medical specialties) based on the patient's experience. Computer-assisted medical systems can also assist doctors in diagnosis, provide patient-related consultation services, or assist patients in self-diagnosis.

[0003] Most computer-assisted medical systems utilize artificial intelligence (including machine learning and / or neural network models) to predict potential diseases or provide related advice. However, AI-based technologies often fail to provide any explanation when providing symptom inquiries and final predictive results (diagnosis or advice). Therefore, patients or doctors may find it difficult to understand why these symptom inquiries are made during the diagnostic process. Without proper explanation, patients or doctors may feel confused or distrustful of AI-based diagnostic results. Summary of the Invention

[0004] One embodiment of this disclosure discloses a medical system including an interface and a processor. The interface is used to receive an input state. The processor is coupled to the interface and is used to: execute a symptom checker based on a neural network model to select a current action from a plurality of candidate symptom queries and a plurality of candidate disease predictions based on the input state; in response to the current action being a first symptom query, execute an interpretive module interacting with the symptom checker to generate a diagnostic tree to simulate a plurality of potential diagnostic paths, each of the plurality of potential diagnostic paths passing through the first symptom query and terminating at one of the disease predictions, the plurality of potential diagnostic paths encompassing a positive hypothesis and a negative hypothesis regarding the first symptom query; and generate a symptom query interpretation regarding the first symptom query based on the diagnostic tree.

[0005] In some embodiments, the symptom checker based on the neural network model is used to generate multiple current state values ​​of the multiple candidate symptom queries and the multiple candidate disease predictions according to the input state, and the symptom checker is used to select the current action according to the maximum value of the multiple current state values.

[0006] In some embodiments, the interpretive module generates the diagnostic tree by: generating a first simulated state containing the affirmative hypothesis of the inquiry into the first symptom and the input state; inputting the first simulated state into the symptom checker to select a first simulated action following the current action; and completing a potential diagnostic path in response to the first simulated action being a prediction of one of the diseases.

[0007] In some embodiments, the interpretive module generates the diagnostic tree by: generating a positive hypothesis and a negative hypothesis for the second symptom inquiry in response to the first simulated action being a second symptom inquiry; and bifurcating the positive hypothesis and the negative hypothesis for the second symptom inquiry into at least two potential diagnostic paths.

[0008] In some embodiments, the interpretive module generates the diagnostic tree by: generating a second simulation state containing the negative hypothesis of the inquiry into the first symptom and the input state; inputting the second simulation state into the symptom checker to select a second simulation action following the current action; and completing a potential diagnostic path in response to the second simulation action being a prediction of one of the diseases.

[0009] In some embodiments, the interpretive module generates the diagnostic tree by: generating a positive hypothesis and a negative hypothesis for the third symptom inquiry in response to the second simulated action being a third symptom inquiry; and bifurcating the positive hypothesis and the negative hypothesis for the third symptom inquiry into at least two potential diagnostic paths.

[0010] In some embodiments, the interpretation of the symptom inquiry regarding the first symptom is used to indicate that the first symptom inquiry can exclude or distinguish two sets of diseases.

[0011] In some embodiments, the input state includes multiple symptom responses corresponding to multiple past symptom queries along a past diagnostic path, and the processor is further configured to: in response to the current action being a disease prediction in a complete diagnostic process, execute the interpretive module interacting with the symptom checker to generate a post-analysis diagnostic tree based on the past diagnostic path, wherein the post-analysis diagnostic tree includes all potential diagnostic paths branching off from the multiple past symptom queries; and generate a post-analysis interpretation of the multiple past symptom queries and the disease prediction based on the post-analysis diagnostic tree.

[0012] In some embodiments, the interpretive module generates the post-analysis diagnostic tree by: calculating, based on the post-analysis diagnostic tree, a quantitative change in the number of multiple disease hypotheses considered by the neural network model before and after each of the multiple past symptom inquiries; and calculating multiple importance scores for each of the multiple past symptom inquiries based on the quantitative change.

[0013] Another embodiment of this disclosure discloses a control method comprising: receiving an input state; using a neural network model, selecting a current action from a plurality of candidate symptom queries and a plurality of candidate disease predictions based on the input state; in response to the current action being a first symptom query, generating a diagnostic tree to simulate a plurality of potential diagnostic paths, each of the plurality of potential diagnostic paths passing through the first symptom query and terminating at one of the disease predictions, the plurality of potential diagnostic paths encompassing a positive hypothesis and a negative hypothesis regarding the first symptom query; and generating a symptom query interpretation regarding the first symptom query based on the diagnostic tree.

[0014] Another embodiment of this disclosure discloses a non-transitory computer-readable medium comprising at least one instruction program, which is executed by a processor to implement the above-described control method.

[0015] In the example above, the explanatory module can generate explanations for symptom inquiries and post-analysis, allowing users to understand why the system raises a symptom inquiry and provides a disease prediction. In this context, users may be more receptive to the healthcare system's recommendations and develop greater trust in it during interactions. Attached Figure Description

[0016] This disclosure can be more fully understood by reading the following detailed description of the embodiments, in conjunction with the accompanying drawings: Figure 1 A schematic diagram of a medical system according to some embodiments of this disclosure is shown; Figure 2 Some embodiments according to this disclosure are shown. Figure 1 A schematic diagram of the interface and processor functions in the diagram; Figure 3 A flowchart of a control method according to some embodiments of this disclosure is shown; Figure 4 The illustration shows a schematic example of selecting a symptom query as the current action based on the input state, according to some embodiments. Figure 5 A schematic diagram of a diagnostic tree generated according to an interpretive module in some embodiments is shown; Figure 6A schematic diagram illustrating the first simulated action following the selection of a symptom inquiry as the current action based on a first simulated state, according to one of some embodiments, is shown. Figure 7 A schematic diagram illustrating the past diagnostic paths and past symptom inquiries contained in the input state in an illustrative example according to some embodiments is shown. Figure 8 A schematic diagram is shown of the post-analysis diagnostic tree generated by the interpretive module from past diagnostic paths; and Figure 9 The graph shows the change in the number of disease hypotheses considered by the neural network model before and after each past symptom inquiry, based on the diagnostic tree obtained from post-hoc analysis.

[0017] Symbol explanation: 120: Interface 140: Processor 142: Symptom Checker 142a: Neural Network Model 144: Explanatory Module INst: Input status INinfo: Medical Information INsym: Enter symptom status Cst: Status value Csym: Candidate Symptom Inquiry Cdp: Candidate Disease Prediction ACT: Current Action QRY: Symptom Inquiry DP: Disease Prediction EXP1: Symptom Inquiry and Explanation EXP2: Post-event analysis and explanation Est1: First simulation state Est2: Second Simulation State ACTp1: First Simulation Operation ACTp2: Second Simulation Operation DT: Diagnostic Tree Detailed Implementation

[0018] The following disclosure provides numerous different embodiments or examples for implementing various features of this disclosure. Elements and configurations in the specific examples are used in the following discussion to simplify this disclosure. Any examples discussed are for illustrative purposes only and do not in any way limit the scope or meaning of this disclosure or its examples. Where appropriate, the same reference numerals are used between figures and in corresponding text descriptions to represent the same or similar elements.

[0019] Please see Figure 1This illustrates a schematic diagram of a medical system 100 according to some embodiments of this disclosure. For example... Figure 1 As shown, the medical system 100 includes an interface 120, a processor 140, and a storage unit 160.

[0020] In some embodiments, the processor 140 is communicatively connected to the interface 120. The medical system 100 is used to interact with the user U1 through the interface 120. For example, during interaction with the user U1, the interface 120 may collect initial symptoms from the user U1 (as part of the input state INst), pose symptom queries (QRY) to the user U1, and collect corresponding symptom responses from the user U1 (as part of the input state INst). Based on the above interaction process, the medical system 100 can thereby analyze, diagnose, or predict diseases that the user U1 may encounter, thereby generating a disease prediction (DP) and feeding it back to the user U1.

[0021] In some embodiments, user U1 can be a patient, a patient's family member, a patient's friend, or a patient accompanied by a doctor. It should be noted that the medical system 100 is capable of generating a symptom inquiry explanation EXP1 for the symptom inquiry QRY. The symptom inquiry explanation EXP1 can be displayed simultaneously with the symptom inquiry QRY on the interface 120. For example, when the interface 120 displays the symptom inquiry QRY, such as "Does your ear hurt?", the interface 120 can also display the corresponding symptom inquiry explanation EXP1, such as "This symptom inquiry helps differentiate / rule out acute otitis media and influenza".

[0022] In some embodiments, the initial symptoms and symptom responses entered by the user Ul are collected via interface 120 and become the input symptom state INsym in the input state INst.

[0023] In some embodiments, interface 120 may further collect other medical information (INinfo) about user U1 (e.g., gender, weight, age, race, blood pressure, occupation, DNA report, test results, etc.) as another part of the input state (INst). This medical information (INinfo) also helps to generate appropriate symptom queries (QRY) and to arrive at accurate disease predictions (DP). For example, if user U1 is biologically male, inquiries and predictions about pregnancy can be ignored.

[0024] Please refer to the following: Figure 2 , Figure 2 Some embodiments according to this disclosure are shown. Figure 1 A functional block diagram of interface 120 and processor 140. (See diagram below.) Figure 2As shown, processor 140 is used to execute symptom checker 142 and interpretation module 144. Symptom checker 142 operates based on neural network model 142 to select a current action ACT from multiple candidate symptom queries Csym and multiple candidate disease predictions Cdp based on input state INst.

[0025] In some embodiments, the symptom checker 142 and the neural network model 142a are trained using machine learning algorithms or reinforcement learning algorithms, thereby enabling the symptom checker 142 to formulate inquiries (generating appropriate symptom queries QRY) and make diagnoses (generating correct disease prediction DP) based on limited patient information. In some embodiments, the medical system 100 employs a reinforcement learning (RL) framework to formulate inquiry and diagnosis strategies (e.g., using Markov decision processes). In some embodiments, the symptom checker 142 and the neural network model 142a are trained by the processor 140 using machine learning algorithms or reinforcement learning algorithms based on some training data (e.g., known medical records), wherein the post-training parameters of the neural network model 142a can be stored in the storage unit 160.

[0026] In some embodiments, the neural network model 142a is pre-trained based on some training data (e.g., known medical records). The processor 140 uses the neural network model 142a to generate a state value Cst and accordingly selects a series of multiple sequence actions from a set of candidate actions. In some embodiments, these sequence actions include multiple symptom inquiry actions, one or more medical examination actions (suitable for providing additional information to facilitate subsequent disease prediction or diagnosis), and a disease prediction action.

[0027] When the symptom checker 142 selects an appropriate action (e.g., appropriate symptom inquiries, appropriate medical examinations, or a correct disease prediction that matches known medical records in the training data), a corresponding reward is generated and provided to the neural network model 142a. In some embodiments, the training objective of the neural network model 142a is to maximize the cumulative reward obtained from the aforementioned sequence of actions. In some embodiments, the cumulative reward may be the sum of symptom abnormality rewards and / or correct / incorrect disease prediction rewards. In other words, the training of the neural network model 142a aims to enable it to generate appropriate symptom inquiries and make as accurate disease predictions as possible.

[0028] In some embodiments, the medical system 100 may be established by a computer, server, or data processing center. The processor 140 may be implemented by a central processing unit (CPU), graphics processing unit (GPU), tensor processing unit (TPU), application-specific integrated circuit (ASIC), or any equivalent computing unit. The interface 120 may include an output interface (e.g., a display panel for displaying information) and input devices (e.g., a touch panel, keyboard, microphone, scanner, or flash memory reader) for users to input text commands, send voice commands, or upload relevant data (e.g., images, medical records, or personal examination reports). Figure 1 As shown, storage unit 160 is coupled to processor 140. In some embodiments, storage unit 160 may be implemented by memory, flash memory, read-only memory, hard disk or any equivalent storage element.

[0029] like Figure 1 and Figure 2 As shown, user U1 can operate through interface 120. User U1 can see the information displayed on interface 120, and user U1 can input his / her instructions or information on interface 120. In one embodiment, interface 120 will display a notification to inquire about the symptoms experienced by user U1. Interface 120 is used to collect the input symptom status INsym responded by user U1. Interface 120 can also collect other medical information INinfo about user U1. Interface 120 transmits the collected input status INst (including input symptom status INsym and medical information INinfo) to the symptom checker 142 of processor 140.

[0030] Please refer to further information. Figure 3 , Figure 3 A flowchart of a control method 200 according to some embodiments of this disclosure is shown. Control method 200 is used to control... Figure 1 The medical system 100.

[0031] like Figure 1 , Figure 2 and Figure 3 As shown, in step S210, interface 120 is used to collect input status INst (including input symptom status INsym and medical information INinfo) and transmit input status INst to processor 140.

[0032] In step S220, processor 140 receives input state INst, and symptom checker 142 of processor 140 uses neural network model 142a to generate current state value Cst for each candidate symptom query Csym and each candidate disease prediction Cdp based on input state INst.

[0033] In some embodiments, the neural network model 142a may be pre-trained using machine learning algorithms or reinforcement learning algorithms based on training data. In some embodiments, the training data includes known medical records. The medical system 100 uses the known medical records in the training data to train the neural network model 142a. In one example, the training data may be obtained from data and statistics published on the website of the U.S. Centers for Disease Control and Prevention (CDC).

[0034] After training, the neural network model 142a is able to generate a state value Cst based on the content of the input state INst (including the input symptom state INsym and medical information INinfo).

[0035] Neural network model 142a evaluates and calculates the state value Cst based on the input state INst. Based on the functionality of neural network model 142a, if the input state INst contains insufficient information to predict the disease (e.g., only containing responses to two symptom queries, and insufficient evidence for disease prediction), then the corresponding candidate symptom query Csym is more likely to have a higher state value, while the candidate disease prediction Cdp tends to have a lower state value. On the other hand, if the input state INst contains sufficient information for disease prediction, then the candidate symptom query Csym tends to have a relatively lower state value, while the candidate disease prediction Cdp tends to have a higher state value.

[0036] In step S230, the symptom checker 142 selects the one with the maximum value from all the current state values ​​Cst of the candidate symptom query Csym and the candidate disease prediction Cdp as the current action ACT.

[0037] For example, if one of the candidate symptom queries Csym has the largest state value, the corresponding symptom query QRY will be selected as the current action ACT. On the other hand, if one of the candidate disease predictions Cdp has the largest state value, the corresponding disease prediction DP will be selected as the current action ACT.

[0038] In step S240, processor 140 determines whether the current action ACT is a symptom inquiry QRY or a disease prediction DP. If the current action ACT is a symptom inquiry QRY (i.e., the information of the input symptom state INsym of the input state INst is insufficient to give a reliable disease prediction DP at the current stage), then steps S250 and S260 are executed to generate a symptom inquiry interpretation EXP1 for the symptom inquiry QRY.

[0039] Further reference Figure 4 , Figure 4 This illustration shows a schematic example, according to some embodiments, of selecting the symptom query QRYs6 as the current action ACT based on the input state INst. Figure 4 In the illustrative example shown, assume the input symptom status INsym includes nine data bits s1-s9, corresponding to nine different symptom queries. Each bit in data bits s1 to s9 indicates whether user U1 has a corresponding symptom. For example, data bit s2 set to "1" indicates that user U1 has the symptom of "cough"; data bit s4 set to "-1" indicates that user U1 does not have the other symptom "headache". Other data bits s1, s3, and s5 to s9 set to "0" indicate that it has not yet been determined whether user U1 has the corresponding symptom (such as "stomach pain", "loss of appetite", "fever", "ear pain", "shortness of breath", etc.).

[0040] Currently, the input symptom state INsym in the input state INst includes only two confirmed answers, data bits s2 and s4, in all symptom data bits s1-s9. In this case, the symptom checker 142 selects the symptom query QRYs6 as the current action ACT. For example, the symptom query QRYs6 could be "Do you have ear pain?". In some embodiments, the symptom query QRYs6 will be displayed on the interface 120.

[0041] Meanwhile, the symptom queries QRYs6 and the input state INst are transmitted to the interpretive module 144. In step S250, the interpretive module 144 is able to interact with the symptom checker 142 and is used to generate a diagnostic tree DT, which is used to simulate multiple potential diagnostic paths starting from the input state INst and the symptom queries QRYs6.

[0042] Please refer to the following: Figure 5 This illustrates a schematic diagram of the diagnostic tree DT generated by the interpretive module 144 in some embodiments. The diagnostic tree DT is used to simulate all potential diagnostic paths starting from the input state INst and the symptom queries QRYs6. Regarding Figure 5Further details on how the diagnostic tree DT is generated in step S250 will be discussed in the following paragraphs.

[0043] like Figure 4 As shown, the interpretive module 144 generates a positive hypothesis PH for the symptom query QRYs6 and fills the positive hypothesis PH into the data bit s6 of the input symptom state INsym in the first simulation state Est1. In other words, the first simulation state Est1 includes the positive hypothesis PH "1", which replaces the original data bit s6 "0" of the input symptom state INsym in the original input state INst, while the other data bits in the first simulation state Est1 (data bits s1-s5 and s7-s9) are copied from the input state INst. In this way, the first simulation state Est1 can simulate the positive response input by the user U1, that is, the answer that he / she has indeed encountered the symptoms described in the symptom query QRYs6.

[0044] like Figure 4 As shown, the interpretive module 144 generates a negative hypothesis NH for the symptom query QRYs6 and fills the data bit s6 in the input symptom state INsym in the second simulation state Est2 with the negative hypothesis NH "-1". In other words, the second simulation state Est2 includes the negative hypothesis NH "-1", which replaces the original data bit s6 "0" of the input symptom state INsym in the original input state INst. The other data bits in the second simulation state Est2 (data bits s1-s5 and s7-s9) are copied from the input state INst. In this way, the second simulation state Est2 can simulate the negative response input by the user U1, that is, answer that he / she has not experienced the symptoms described in the symptom query QRYs6.

[0045] like Figure 5 As shown, the diagnostic tree DT branches into at least two potential diagnostic paths, corresponding to the affirmative hypothesis PH and the negative hypothesis NH of the symptom inquiry QRYs6, respectively. Figure 5 As shown, the potential diagnostic pathways include at least a first potential diagnostic pathway PATH1 and a second potential diagnostic pathway PATH2. The first potential diagnostic pathway PATH1 covers the affirmative hypothesis PH of symptom inquiry QRYs6. The second potential diagnostic pathway PATH2 covers the negative hypothesis NH of symptom inquiry QRYs6.

[0046] It is worth noting that, Figure 4 The input state INst (and the first simulation state Estl) can further include Figure 2 The medical information (INinfo) is included. For the sake of brevity, Figure 2The medical information INinfo shown is not displayed in Figure 4 among.

[0047] like Figure 2 and Figure 5 As shown, the interpretive module 144 inputs the first simulated state Est1 to the symptom checker 142. The symptom checker 142, based on the neural network model 142a, is able to select the first simulated action ACTp1 after the current action ACT (i.e., symptom inquiry QRYs6).

[0048] If the first simulated action ACTp1 selected based on the first simulated state Est1 is one of the candidate disease predictions (e.g., one of the disease predictions DPd1 to DPd6), then this potential diagnostic path ends here.

[0049] In this example, such as Figure 5 As shown, the first simulated action ACTp1 selected based on the first simulation state Est1 is not the disease prediction DP. Instead, the selected first simulated action ACTp1 is another symptom query QRYs9 related to data bit s9. Please refer to [further details needed]. Figure 6 , Figure 6 A schematic diagram is shown illustrating a first simulated action ACTp1 following the selection of symptom inquiry QRYs9 as the current action ACT based on a first simulated state Est1, according to one of some embodiments.

[0050] like Figure 6 As shown, the interpretive module 144 generates a positive hypothesis PH for the symptom query QRYs9 and fills the positive hypothesis PH into the data bits s9 of the third simulation state Est3. In other words, the third simulation state Est3 contains the positive hypothesis PH "1" replacing the original data bits s9 "0" in the first simulation state Est1.

[0051] like Figure 6 As shown, the interpretive module 144 generates a negative hypothesis NH for the symptom query QRYs9 and fills the negative hypothesis NH into the data bit s9 in the fourth simulation state Est4. In other words, the fourth simulation state Est4 contains the negative hypothesis NH "-1" replacing the original data bit s9 "0" in the first simulation state Est1.

[0052] like Figure 5 As shown, the diagnostic tree DT branches again at symptom query QRYs9 into at least two potential diagnostic paths, corresponding to the affirmative hypothesis PH and the negative hypothesis NH of symptom query QRYs9, respectively. Figure 5As shown, the potential diagnostic pathways include at least a first potential diagnostic pathway PATH1 and a third potential diagnostic pathway PATH3. The first potential diagnostic pathway PATH1 covers the affirmative hypothesis PH of the symptom inquiry QRYs9. The third potential diagnostic pathway PATH3 covers the negative hypothesis NH of the symptom inquiry QRYs9.

[0053] Similarly, the interpretability module 144 can again input the third simulation state Est3 into the symptom checker 142. The symptom checker 142, based on the neural network model 142a, can select another simulation action following the symptom queries QRYs9. Suppose that the symptom checker 142 selects disease prediction DPd1 after symptom queries QRYs9. In this case, the potential diagnostic path PATH1 ends at disease prediction DPd1. Figure 5 As shown, the potential diagnostic path PATH1 starts from the input state INst, goes through symptom inquiry QRYs6 (simulation with positive assumptions) and symptom inquiry QRYs9 (simulation with positive assumptions), and stops at disease prediction DPd1.

[0054] like Figure 5 As shown, the interpretability module 144 can further input a fourth simulation state Est4 into the symptom checker 142 and repeat the above process until each potential diagnostic path arrives at its respective disease prediction. Figure 5 As shown, the potential diagnostic path PATH3 starts from the input state INst, goes through symptom query QRYs6 (simulation with positive hypothesis), symptom query QRYs9 (simulation with negative hypothesis) and symptom query QRYs3 (simulation with negative hypothesis) and stops at disease prediction DPd2.

[0055] Similarly, the interpretive module 144 can further input the second simulated state Est2 into the symptom examiner 142 to generate a second simulated action ACTp2, thereby generating all potential diagnostic pathways under the second simulated state Est2. For example... Figure 5 As shown, the potential diagnostic path PATH2 starts from the input state INst, proceeds through symptom queries QRYs6 (simulation with negative hypothesis), symptom queries QRYs7 (simulation with negative hypothesis), and symptom queries QRYs8 (simulation with negative hypothesis), and stops at disease prediction DPd4.

[0056] Based on the above embodiments, the interpretability module 144 can interact with the symptom checker 142, and has generated, for example... Figure 5 The complete diagnostic tree (DT) is shown. In some embodiments, the diagnostic tree (DT) contains all potential diagnostic paths that branch off at each branch point (i.e., each symptom inquiry (QRY)).

[0057] When Figure 5After the diagnostic tree DT is generated, in step S260, the interpretability module 144 generates a symptom query interpretation EXP1 based on the diagnostic tree DT regarding the symptom queries QRYs6 (i.e., the current action ACT). In some embodiments, the symptom query interpretation EXP1 is used to indicate that the symptom queries QRYs6 can exclude or distinguish two disease sets.

[0058] In some embodiments, this can be achieved by combining the set of all disease predictions under the affirmative hypothesis of the symptom inquiry QRYs6. The set of all disease predictions under the negative hypothesis of QRYs6 for symptom inquiry. By comparing these findings, the two disease sets in EXP1, as described above, are derived from the symptom inquiry explanation. For example... Figure 5 In the illustrated embodiment, the set Includes disease prediction DPd1, DPd2, and DPd3; while the set This includes disease prediction DPd1, DPd3, DPd4, DPd5 and DPd6.

[0059] In this example, the set of diseases that the symptom inquiry QRYs6 can be used to identify or rule out includes... (Its representative belongs to the set) But it does not belong to a set. Disease prediction and collection (Its representative belongs to the set) But it does not belong to a set. Disease prediction.

[0060] In this illustrative example, the set Includes disease prediction DPd2; set This includes disease prediction DPd5, disease prediction DPd6, and disease prediction DPd6. In this case, the symptom query interpretation EXP1 for symptom query QRYs6 can point to two disease sets, which contain the set That is, {DPd2} and the set That is, {DPd4, DPd5, DPd6}. According to the Symptom Questioning Explanation EXP1, the user U1 can understand that the Symptom Questioning QRYs6 helps to rule out the possibility of disease predictions DPd4, DPd5, and DPd6 (if the answer to Symptom Questioning QRYs6 is "yes") or rule out the possibility of a disease prediction DPd2 (if the answer to Symptom Questioning QRYs6 is "no").

[0061] In this case, when the symptom inquiry QRYs6 is displayed on interface 120, the symptom inquiry explanation EXP1 related to symptom inquiry QRYs6 can be displayed on interface 120 together with symptom inquiry QRYs6.

[0062] Table 1 below is an example of the symptom inquiry QRYs6 and symptom inquiry explanation EXP1 used to display on interface 120.

[0063]

[0064] Table 1 As described in the foregoing embodiments, the interpretive module 144 can generate a symptom inquiry explanation EXPl for the current action ACT (i.e., symptom inquiry QRYs6), enabling the user Ul to know or understand why the system raised this symptom inquiry QRYs6 and why this inquiry is important. In this case, the user Ul may be more receptive to the suggestions of the medical system 100 in their interactions with the medical system 100 and develop greater trust in the medical system 100.

[0065] Please refer to it again. Figure 3 If the selected current action ACT is already a disease prediction DP in a complete diagnostic process (meaning that the input symptom state INsym of the input state INst is sufficient to make a disease prediction DP at present), then execute steps S270 and S280 to produce a post-hoc interpretation EXP2.

[0066] If the current action (ACT) is a disease prediction DP, the input state (INst) in this case may contain symptom responses related to all past symptom inquiries from past diagnostic pathways. The post-hoc analysis interpretation (EXP2) can be used to explain and support why the current disease prediction DP was chosen, and can also describe the importance score for each past symptom inquiry. Please refer to further details. Figure 7 , Figure 7 A schematic diagram of the past diagnostic path pPATH and its past symptom queries contained in the input state INst, according to some embodiments, is shown in an exemplary example.

[0067] In this illustrative example, based on the input state INst, the disease prediction DPdl is selected as the current action ACT. For example... Figure 7 As shown, before reaching the current input state INst and the current action ACT (i.e., disease prediction DPd1), the past diagnostic path pPATH starts from the past state pst0 and goes through multiple past symptom inquiries pQRYs4, pQRYs6, pQRYs7, pQRYs3 and pQRYs1.

[0068] In step S270, the interpretive module 144 (through interaction with the symptom checker 142) generates a post-analysis diagnostic tree based on the past diagnostic path pPATH. Please refer to [further details omitted]. Figure 8 This illustrates a schematic diagram of the post-analysis diagnostic tree DTp generated by the interpretive module 144 from the past diagnostic path pPATH. The post-analysis diagnostic tree DTp is generated by simulating all potential diagnostic paths branching off from all bifurcation points in the past diagnostic path pPATH, where the bifurcation points refer to the past symptom queries pQRYs4, pQRYs6, pQRYs7, pQRYs3, and pQRYs1 present in the past diagnostic path pPATH.

[0069] Regarding how it is generated Figure 8 The detailed implementation of the post-hoc analysis diagnostic tree DTp, which branches off from each of the past symptom queries pQRYs4, pQRYs6, pQRYs7, pQRYs3, and pQRYs1, is similar to the simulation generated in the previous embodiment. Figure 5 The diagram shows all potential diagnostic paths in the diagnostic tree DT. For example, a positive hypothesis for past symptom queries pQRYs4 is generated and fed back to the symptom checker 142 to simulate / compute a local portion SIMs4+ of the post-analysis diagnostic tree DTp. Similarly, another positive hypothesis for past symptom queries pQRYs6 is generated and fed back to the symptom checker 142 to simulate / compute a local portion SIMs6+ of the post-analysis diagnostic tree DTp. Other portions of the post-analysis diagnostic tree DTp can be simulated and generated in the same manner.

[0070] When such a situation occurs Figure 8 Following the post-analysis diagnostic tree DTp shown, in step S280, the interpretive module 144 can be used to generate a post-analysis interpretation EXP2 based on the aforementioned post-analysis diagnostic tree DTp regarding past symptom inquiries pQRYs4, pQRYs6, pQRYs7, pQRYs3, and pQRYs1, as well as disease prediction DPd1.

[0071] In some embodiments, the post-hoc analysis interpretation EXP2 is generated based on the importance scores of pQRYs4, pQRYs6, pQRYs7, pQRYs3, and pQRYs1 in the past symptom inquiry.

[0072] The interpretive module 144 calculates the change in the number of disease hypotheses considered by the neural network model 142a before and after each of the past symptom queries pQRYs4, pQRYs6, pQRYs7, pQRYs3, and pQRYs1, based on the post-analysis diagnostic tree DTp.

[0073] Please refer to the following: Figure 9 It shows a line graph of the change in the number of disease hypotheses considered before and after each past symptom inquiry pQRYs4, pQRYs6, pQRYs7, pQRYs3 and pQRYs1, based on the post-analysis diagnostic tree DTp.

[0074] Next, the interpretive module 144 calculated the importance scores for each of the past symptom inquiries pQRYs4, pQRYs6, pQRYs7, pQRYs3 and pQRYs1 based on the quantitative change VAR.

[0075] The importance scores for pQRYs4, pQRYs6, pQRYs7, pQRYs3, and pQRYs1 in the above past symptom inquiry can be calculated in the following ways:

[0076] The importance score mentioned above can provide information on which past symptom inquiries are of greater importance to the final disease prediction DPd1. In other words, user U1 can use the importance score to understand why the responses to the past symptom inquiries pQRYs4, pQRYs6, pQRYs7, pQRYs3, and pQRYs1 can support the final disease prediction DPd1.

[0077] In the above formula, The changes in disease hypotheses considered by the neural network model 142a before and after individual past symptom inquiries; This represents the amount of change in all disease hypotheses along the previous diagnostic path pPATH in the entire diagnostic process of neural network model 142a.

[0078] For example, the importance score of the past symptom inquiry pQRYs4 can be calculated as follows:

[0079] For example, the importance score of the past symptom inquiry pQRYs6 can be calculated as follows:

[0080] After calculating the importance scores for each of the past symptom queries (pQRYs4, pQRYs6, pQRYs7, pQRYs3, and pQRYs1), these past symptom queries can be ranked according to their respective importance scores and displayed in the post-hoc analysis and interpretation (EXP2). For example, past symptom query pQRYs4 can be marked as the most important; past symptom queries pQRYs6 and pQRYs3 can be marked as having the next highest importance (lower than the importance of past symptom query pQRYs4).

[0081] User U1, through post-hoc analysis interpreting the importance scores of past symptom inquiries pQRYs4, pQRYs6, pQRYs7, pQRYs3, and pQRYs1 presented in EXP2, may further understand why these past symptom inquiries pQRYs4, pQRYs6, pQRYs7, pQRYs3, and pQRYs1 were selected and presented, and understand the importance of these inquiries in determining the disease prediction DP. In this case, User U1 can have greater confidence in the disease prediction DP.

[0082] While this disclosure has revealed specific details regarding the above embodiments, these embodiments are not intended to limit this disclosure. Various alternatives and modifications can be made by those skilled in the art without departing from the principles and spirit of this disclosure. Therefore, the scope of protection of this disclosure is determined by the appended claims.

Claims

1. A medical system, characterized in that, Include: An interface for receiving an input status; and A processor, coupled to the interface, wherein the processor is used to: A symptom checker is executed based on a neural network model to select a current action from multiple candidate symptom queries and multiple candidate disease predictions based on the input state. In response to the current action being a first symptom inquiry, an interpretive module interacting with the symptom checker is executed to generate a diagnostic tree, thereby simulating multiple potential diagnostic paths, each passing through the first symptom inquiry and terminating at one of the disease predictions. These multiple potential diagnostic paths encompass both a positive hypothesis and a negative hypothesis regarding the first symptom inquiry. Based on the diagnostic tree, a symptom inquiry explanation is generated regarding the first symptom inquiry, wherein the symptom inquiry explanation is displayed on the interface simultaneously with the first symptom inquiry, and the symptom inquiry explanation is used to explain why the first symptom inquiry was raised.

2. The medical system of claim 1, wherein the symptom checker based on the neural network model is used to generate multiple current state values ​​of the multiple candidate symptom queries and the multiple candidate disease predictions according to the input state, and the symptom checker is used to select the current action according to a maximum value of the multiple current state values.

3. The medical system of claim 1, wherein the interpretive module generates the diagnostic tree by: The generation of a first simulated state includes the affirmative hypothesis regarding the inquiry into the first symptom and the input state; The first simulated state is input into the symptom checker to select a first simulated action following the current action; and In response to the prediction of one of the diseases by the first simulation action, a potential diagnostic pathway is completed.

4. The medical system of claim 3, wherein the interpretive module generates the diagnostic tree further by: In response to the first simulated action being a second symptom inquiry, a positive hypothesis and a negative hypothesis regarding the second symptom inquiry are generated; and The affirmative hypothesis that should be asked about the second symptom and the negative hypothesis that should be asked about the second symptom bifurcate into at least two potential diagnostic pathways.

5. The medical system of claim 1, wherein the interpretive module generates the diagnostic tree further by: A second simulated state is generated, which includes the negative hypothesis regarding the question about the first symptom and the input state; The second simulated state is input into the symptom checker to select a second simulated action following the current action; and In response to the second simulation action predicting one of the diseases, a potential diagnostic pathway is completed.

6. The medical system of claim 5, wherein the interpretive module generates the diagnostic tree further by: In response to the second simulated action being a third symptom inquiry, a positive hypothesis and a negative hypothesis regarding the third symptom inquiry are generated; and The affirmative hypothesis that should be asked about the third symptom and the negative hypothesis that should be asked about the third symptom bifurcate into at least two potential diagnostic pathways.

7. The medical system of claim 1, wherein the interpretation of the symptom inquiry regarding the first symptom inquiry is used to indicate that the first symptom inquiry can exclude or distinguish two disease sets.

8. The medical system of claim 1, wherein the input state includes multiple symptom responses corresponding to multiple past symptom inquiries along a past diagnostic path, the processor further being configured to: In response to the current action being a disease prediction within a complete diagnostic process, the interpretive module interacting with the symptom checker is executed to generate a post-analysis diagnostic tree based on the past diagnostic paths, wherein the post-analysis diagnostic tree includes all potential diagnostic paths branching off from the multiple past symptom inquiries; and Based on this post-hoc diagnostic tree, a post-hoc interpretation is generated regarding the multiple past symptom inquiries and the prediction of the disease.

9. The medical system of claim 8, wherein the interpretive module generates the post-analysis diagnostic tree by: Based on the post-hoc diagnostic tree, calculate the change in the number of disease hypotheses considered by the neural network model before and after each of the multiple past symptom inquiries; and Based on the change in this quantity, calculate multiple importance scores for each of the multiple past symptom inquiries.

10. A control method, characterized in that, Include: Receive an input status; Using a neural network model, based on the input state, a current action is selected from multiple candidate symptom queries and multiple candidate disease predictions; In response to the current action being a first symptom inquiry, a diagnostic tree is generated to simulate multiple potential diagnostic paths, each of which passes through the first symptom inquiry and terminates at one of the disease predictions. These multiple potential diagnostic paths encompass a positive hypothesis and a negative hypothesis regarding the first symptom inquiry. Based on the diagnostic tree, a symptom inquiry explanation is generated for the first symptom inquiry, wherein the symptom inquiry explanation is displayed on an interface simultaneously with the first symptom inquiry, and the symptom inquiry explanation is used to explain why the first symptom inquiry was asked.

11. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores at least one instruction program executed by a processor to implement a control method, the control method comprising: Receive an input status; Using a neural network model, based on the input state, a current action is selected from multiple candidate symptom queries and multiple candidate disease predictions; In response to the current action being a first symptom inquiry, a diagnostic tree is generated to simulate multiple potential diagnostic paths, each of which passes through the first symptom inquiry and terminates at one of the disease predictions. These multiple potential diagnostic paths encompass a positive hypothesis and a negative hypothesis regarding the first symptom inquiry. Based on the diagnostic tree, a symptom inquiry explanation is generated for the first symptom inquiry, wherein the symptom inquiry explanation is displayed on an interface simultaneously with the first symptom inquiry, and the symptom inquiry explanation is used to explain why the first symptom inquiry was asked.

Images