TWI894564B - Methods, devices, and non-transitory computer storage medium of matching clinical trials - Google Patents

Abstract

Disclosed are methods, devices and the non-transitory computer storage media of matching clinical trials. The present disclosure provides a method of matching clinical trials. The method comprises: obtaining a first data set from a pathology report; obtaining a second data set of a clinical trial; determining whether the first data set and the second data set are matched with respect to a first set of fields; determining a relevance value between the first data set and the second data set with respect to a second set of fields when the first data set and the second data set are matched with respect to the first set of fields; and determining the clinical trial as recommended when the relevance value exceeds a threshold.

Description

Methods, devices, and non-transitory computer storage media for matching clinical trials

This disclosure relates to a method for matching clinical trials, a device for matching clinical trials, and a related non-transitory computer storage medium. Specifically, this disclosure relates to a method for matching patients to clinical trials based on their pathology reports, and a related device and non-transitory computer storage medium.

Patient pathology reports, especially those for cancer patients, contain a wealth of information and can be complex and complex. Surgeons and primary care physicians can spend considerable time understanding a patient's condition and identifying appropriate clinical trials. Computers can help reduce this wasted time and thus increase overall efficiency.

This publication analyzes a patient's pathology report and identifies appropriate clinical trials. A pathology report may contain a diagnosis determined by microscopic examination of cells and tissues. Pathology reports can be used for patients with lung cancer. Key information can be summarized from a confusing and complex pathology report. This information may include feature categories: basic pathology description, tumor characteristics, histology description, immunohistochemistry (IHC) information, genetic test results, and pathological TNM (tumor, node, and metastasis) staging. This publication further summarizes multiple pathology reports for a single patient. This disclosure can further provide the ability to collect data from a large number of clinical trials and compare characteristics obtained from pathology reports with clinical trials to determine appropriate clinical trials for patients. These clinical trials can serve as a reference for surgeons and physicians.

Embodiments of the present disclosure provide a method for matching clinical trials. The method includes: obtaining a first data set from a pathology report; obtaining a second data set of a clinical trial; determining whether the first data set and the second data set match with respect to a first set of fields; when the first data set and the second data set match with respect to the first set of fields, determining a correlation value between the first data set and the second data set with respect to the second set of fields; and when the correlation value exceeds a threshold value, determining that the clinical trial is recommended.

Another embodiment of the present disclosure provides a device for matching clinical trials. The device includes a processor and a memory coupled to the processor. The processor executes computer-readable instructions stored in the memory to perform operations, wherein the operations include: obtaining a first data set from a pathology report; obtaining a second data set of a clinical trial; determining a correlation value between the first data set and the second data set with respect to a first set of fields; and determining that the clinical trial is recommended when the correlation value exceeds a threshold value.

Another embodiment of the present disclosure provides a non-transitory computer storage medium. The non-transitory computer storage medium has program instructions stored thereon. When executed by a processor, the program instructions cause a set of operations to be performed. The operations include: obtaining a first data set from a pathology report; obtaining a second data set from a clinical trial; determining whether the first data set and the second data set match with respect to a first set of fields; when the first data set and the second data set match with respect to the first set of fields, determining a correlation value between the first data set and the second data set with respect to the second set of fields; and determining that the clinical trial is recommended when the correlation value exceeds a threshold value.

11: Pathology report

12: Pre-trained model

13: Pathological characteristics

14: Demographics

15: Clinical Trial Matching System

30: Methods

40: Methods

41: Methods

50: Methods

151: Operation

152: Operation

153: Operation

154: Operation

301: Operation

302: Operation

303: Operation

304: Operation

305: Operation

401: Operation

402: Operation

403: Operation

404: Operation

405: Operation

406: Operation

411: Operation

412: Operation

501: Operation

502: Operation

503: Operation

504: Operation

505: Operation

601: Data

602: Data

700: Computer System

710: Computing device

711: Processor

712: Input/Output Interface

713: Communication Interface

714: Memory

720:Database

1521: Steps

1522: Steps

To illustrate how the advantages and features of the present disclosure may be obtained, the description of the present disclosure is presented with reference to specific embodiments thereof, which are illustrated in the accompanying drawings. These drawings depict only exemplary embodiments of the present disclosure and, therefore, are not to be considered limiting of its scope.

FIG1 is a schematic diagram of a system for matching clinical trials according to some embodiments of the present disclosure.

FIG2 illustrates a flow chart of secondary condition matching in the clinical trial matching system of FIG1 according to some embodiments of the present disclosure.

FIG3 illustrates a flow chart of a method for matching clinical trials according to some embodiments of the present disclosure.

FIG4A illustrates a flow chart of a method for pre-training a model for extracting features from pathology reports according to some embodiments of the present disclosure.

FIG4B illustrates a flow chart of a method for extracting features from a pathology report according to some embodiments of the present disclosure.

FIG5 shows a flow chart of a method for collecting clinical trial data according to some embodiments of the present disclosure.

FIG6 is a schematic diagram illustrating a representation of a clinical trial matching system according to some embodiments of the present disclosure.

FIG7 is a schematic diagram illustrating a computer system according to some embodiments of the present disclosure.

The following disclosure provides numerous different embodiments or examples for implementing various features of the presented subject matter. Specific examples of operations, components, and configurations are described below to simplify this disclosure. However, these are merely examples and are not intended to be limiting. For example, a description of a first operation being performed before or after a second operation may include embodiments in which the first and second operations are performed together, and may also include embodiments in which additional operations are performed between the first and second operations. For example, in the following description, a description of a first feature being formed above, on, or within a second feature may include embodiments in which the first and second features are formed in direct contact, and may also include embodiments in which an additional feature is formed between the first and second features, such that the first and second features are not in direct contact. Furthermore, this disclosure may refer to repeated reference numbers and/or letters throughout the various examples. This repetition is for the purpose of simplicity and clarity and does not in itself indicate a relationship between the various embodiments and/or configurations discussed.

For ease of description, time-relative terms, such as "before," "before," "after," "after," and the like, may be used herein to describe the relationship of one operation or feature to another operation or feature as depicted in the figures. Time-relative terms are intended to encompass different sequences of operations depicted in the figures. Additionally, for ease of description, spatially-relative terms, such as "beneath," "below," "lower," "above," "upper," and the like, may be used herein to describe the relationship of one element or feature to another element or feature as depicted in the figures. Spatially-relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. The device may be otherwise oriented (rotated 90 degrees or at other orientations), and the spatially relative descriptors used herein should be interpreted accordingly. For ease of description, relative terms for connection may be used herein, such as "connect," "connected," "connection," "coupled," "coupled," "in communication," and the like, to describe an operational connection, coupling, or linking between two elements or features. Relative terms for connection are intended to encompass different types of connections, couplings, or links between devices or components. Devices or components may be connected, coupled, or linked to each other directly or indirectly, such as through another set of components. Devices or components may be connected, coupled, or linked to each other using wired and/or wireless means.

As used herein, the singular terms "a," "an," and "the" may include plural referents unless the context clearly indicates otherwise. For example, a reference to a device may include a plurality of devices unless the context clearly indicates otherwise. The terms "include" and "comprising" may indicate the presence of described features, integers, steps, operations, elements, and/or components, but may not exclude the presence of a combination of one or more of those features, integers, steps, operations, elements, and/or components. The term "and/or" may include any and all combinations of one or more of the listed items.

Additionally, quantities, ratios, and other numerical values are sometimes presented herein in a range format. It should be understood that such range format is used for convenience and brevity and should be construed flexibly to include not only the values specifically designated as limits of the range, but also all individual values or sub-ranges encompassed within that range, as if each value and sub-range were specifically designated.

The nature and uses of the embodiments are discussed in detail below. However, it should be understood that the present disclosure provides many applicable inventive concepts that can be embodied in a wide variety of specific contexts. The specific embodiments discussed are merely illustrative of specific ways to embody and use the present disclosure and are not intended to limit its scope.

To match a patient's pathology report (e.g., a lung cancer pathology report) with a clinical trial, this disclosure provides a method for extracting pathology features from the report. In some embodiments of this disclosure, the pathology report may include pathology features in several categories. Exemplary pathology features are listed in Table 1. Categories may include: basic description, tumor detection, histology (tumor information), IHC information, genetic testing (results), and TNM staging. In other embodiments, the data representation of the patient's report may be represented by the pathology features shown in Table 1.

This disclosure provides a clinical trial matching system that analyzes pathology reports, including pathological characteristics and demographic data (i.e., patient information). The clinical trial matching system determines the similarity and correlation between the pathology report and clinical trials, and subsequently identifies recommended clinical trials for the patient. These recommended clinical trials serve as a reference for surgeons and physicians, allowing them to provide patients with more treatment options based on the recommended clinical trials. Using the clinical trial matching system, surgeons and physicians can quickly and accurately find appropriate clinical trials, saving the considerable time required to manually search for clinical trials.

FIG1 is a schematic diagram of a system for matching clinical trials according to some embodiments of the present disclosure.

Referring to FIG1 , a patient's pathology report 11 may be provided. The pathology report 11 may be input into a pre-trained model 12 to extract one or more pathology features 13. The pathology features 13 may be provided to a clinical trial matching system 15.

The pre-trained model 12 can perform classification and/or sequence labeling tasks to extract or obtain pathological features 13. Pathological features 13 may include information related to EGFR, ALK, ROS1, KRAS, BRAF, RET, NTRK, MET, P53, and Her2, as well as information related to surgery (e.g., surgical procedure), histology, tumor size, stage (e.g., pathological stage), and PDL1. Through the classification task, information related to EGFR, ALK, ROS1, KRAS, BRAF, RET, NTRK, MET, P53, Her2, etc. can be extracted or obtained. Regarding the sequence labeling task, information related to surgery, histology, tumor size, stage, PDL1, etc. can be extracted or obtained.

In some embodiments, the patient's demographic data 14 may be provided to the clinical trial matching system 15. The demographic data 14 may include data or information related to age, sex, smoking, lymph node metastasis, distant metastasis, CNS metastasis, bone metastasis, wild-type, anti-angiogenesis, platinum, EGFR TKI, ALK inhibitor, PD-1/PD-L1 inhibitor, CTLA-4 inhibitor, radiation therapy, cisplatin/carboplatin, chemotherapy, systemic therapy, disease status, ECOG PS, etc. The demographic data 14 may be extracted or obtained through a classification task or a sequence labeling task of a pre-trained model 12. When the demographic data 14 is stored in a format that can be directly used by the clinical trial matching system 15 (e.g., computer-processable data), the demographic data 14 can be obtained by accessing the relevant database.

The clinical trial matching system 15 can analyze a patient's pathological characteristics 13 and demographic data 14 and identify one or more suitable clinical trials for the patient. In some embodiments, the clinical trial matching system 15 can be coupled to a clinical trial database (not shown), allowing the clinical trial matching system 15 to compare the pathological characteristics 13 and demographic data 14 with clinical trials. Thus, a recommended clinical trial can be found for the patient.

Referring to FIG. 1 , the clinical trial matching system 15 includes operations 151, 152, 153, and 154. A patient's pathological features 13 and demographic data 14 may be input into the clinical trial matching system 15 and compared with various clinical trials. In operation 151, a determination is made as to whether one or more key criteria (or fields) in the pathological features 13 and demographic data 14 match the inclusion criteria of the clinical trial. In some embodiments, one or more key criteria in the pathological features 13 and demographic data 14 fully match the inclusion criteria of the clinical trial. These key criteria (or fields) may include at least one of the following: estimated glomerular filtration rate (EFGR), surgical procedure, histology, pathological stage, age, sex, or smoking status. In some embodiments, at operation 151, when the patient's condition set values or information fully match those of the condition sets described in the inclusion criteria, the process proceeds to operation 152. The condition sets may include EGFR conditions, surgical conditions, histological conditions, pathological stage conditions, age conditions, gender conditions, and smoking conditions.

When one or more primary conditions are fully matched, the process proceeds to operation 152. On the other hand, when one or more primary conditions are not fully matched, the process proceeds to operation 154.

In operation 152, a determination is made as to whether one or more secondary conditions (or fields) in the pathology feature 13 and demographic data 14 match the clinical trial. In some embodiments, one or more secondary conditions in the pathology feature 13 and demographic data 14 partially match. Unlike primary conditions, secondary conditions may not necessarily fully match the clinical trial. In some embodiments, secondary condition matching is determined by correlation between secondary conditions in the pathology report and secondary conditions in the clinical trial. In some embodiments, when the correlation value is greater than a threshold value, it is determined that one or more secondary conditions match the clinical trial. Details of correlation determination are discussed in FIG. 2 .

When one or more secondary conditions are matched, the process proceeds to operation 153. On the other hand, when one or more secondary conditions are not matched, the process proceeds to operation 154.

In operation 153, when a patient's pathological characteristics 13 and demographic data 14 match a clinical trial (i.e., through operations 151 and 152), a clinical trial is recommended for the patient. The clinical trial matching system 15 may then execute another process to determine whether another clinical trial matches the same patient's pathological characteristics 13 and demographic data 14.

In operation 154, if the patient's pathological characteristics 13 and demographic data 14 do not match a clinical trial (i.e., they fail operations 151 or 152), no clinical trial is recommended for the patient. The clinical trial matching system 15 may then execute another process to determine whether another clinical trial matches the same patient's pathological characteristics 13 and demographic data 14.

Using the Clinical Trial Matching System 15, doctors can easily find relevant ongoing clinical trials for their patients. The Clinical Trial Matching System 15 can screen clinical trials and thus help surgeons and physicians recommend appropriate clinical trials for their patients, giving them more treatment options.

FIG2 illustrates a flowchart of secondary condition matching in the clinical trial matching system 15 of FIG1 according to some embodiments of the present disclosure. Referring to FIG2 , the secondary condition matching operation 152 may include two steps 1521 and 1522.

In some embodiments, the secondary condition match in operation 152 is determined by the relevance of the pathology report and the clinical trial with respect to the secondary condition. In step 1521, a relevance value S _d is determined between the pathology report and the clinical trial with respect to the secondary condition (field). In some embodiments, the pathology report may include the patient's pathological characteristics 13 and demographic data 14. In some embodiments, the relevance value S _d may be determined based on the BM25 algorithm, which is a ranking function used to estimate the relevance of a document to a given search query.

In some embodiments, the correlation value S _d between the pathology report and the clinical trial d can be calculated by Equation 1: Where Q represents all queries (e.g., several secondary conditions); q represents an individual query (e.g., a secondary condition) _; Wq represents an individual weight assigned to the individual query q ; represents the respective inverse document frequency (IDF) of each keyword t ; df _t represents the number of clinical trials including each keyword t ; N represents the number of clinical trials in the clinical trial database; tf _td represents the number of occurrences of keyword t in clinical trial d ; L _d represents the length of clinical trial d ; L _avg represents the average length of all clinical trials in the clinical trial database; k ₁ is a constant used to normalize the frequency range of keywords in the document (for example, k ₁ can be in the range of 1.2 to 2.0, preferably 1.2 or 1.5) ; k ₃ is a constant used to correct the frequency range of keywords in the query (for example, k ₃ may be in the range of 1.2 to 2.0, preferably 1.2 or 1.5); and b is a constant (for example, b may be 0.75 or 0.5).

In some embodiments, a separate relevance value may be obtained from a query (e.g., a field in the second set of fields or a condition in the secondary condition), and the relevance value _Sd is the sum of the separate relevance values. Equation 1 may include the separate weight Wq _, the separate inverse document frequency (IDF), the similarity between the separate keyword t and the separate query q , and the weight of the separate keyword t . The separate IDF may be expressed as The similarity between each keyword t and each query q can be expressed as In some embodiments, due to the use of Laplace smoothing, the similarity equation includes (k ₁ +1). The weight of each keyword t can be expressed as In some embodiments, due to the use of Laplace smoothing, the weight of each keyword t comprises (k ₃ +1).

In some embodiments, each relevance value may be associated with a respective weight Wq _, which is assigned to each query by a clinician. Specifically, each relevance value may be proportional to the respective _weight Wq . The clinician may determine the importance or relevance of each query (or condition) and then assign an appropriate weight to the query (or condition).

In some embodiments, IDF is a numerical statistic that purports to reflect the importance of a word/term to documents in a collection or corpus. IDF is a weight that indicates how commonly a word/term is used. The more frequently a word/term is used in documents in a collection or corpus, the lower its IDF score. The lower the IDF score, the less important the word/term becomes. For example, the term "the" appears in almost all English text and therefore would have a very low IDF score because the term contains very little "topic" information.

In some embodiments, the individual relevance values may be associated with the individual IDF of the individual keywords. For example, the individual relevance values may be proportional to the individual IDF. Thus, the fewer times a individual keyword appears across clinical trials in the clinical trial database, the greater its individual IDF will be, and thus the greater its individual relevance value will be.

Using Equation 1, the correlation value S _d between the pathology report and the clinical trial for the secondary condition (field) can be determined.

In step 1522, when the correlation value _Sd exceeds the threshold value K, the pathology report (or the corresponding pathology features 13 and demographic data 14) is determined to match the clinical trial. Returning to FIG1 , when the pathology features 13 and demographic data 14 match the clinical trial at operation 152, the clinical trial is then recommended to the corresponding patient at operation 153. The correlation value _Sd can be compared with the threshold value K to determine whether the clinical trial matches the pathology report. When the correlation value _Sd of the clinical trial exceeds the threshold value K, the clinical trial is recommended to the patient (i.e., proceeding to operation 153 in FIG1 ). On the other hand, when the correlation value S _d between the pathology report and the clinical trial is less than the threshold value K, the pathology report (or the corresponding pathology features 13 and demographic data 14) is determined to be unmatched with the clinical trial. Returning to FIG. 1 , when the pathology features 13 and demographic data 14 do not match the clinical trial at operation 152 , the clinical trial is not recommended to the corresponding patient at operation 154 .

FIG3 illustrates a flow chart of a method 30 for matching clinical trials according to some embodiments of the present disclosure.

In operation 301, a first dataset may be obtained from a pathology report. In some embodiments, the first dataset may include the patient's pathological features 13 and demographic data 14 as discussed in FIG. 1 . For example, the first dataset (e.g., demographic data 14) may be obtained from a pathology report. In another embodiment, the first dataset (e.g., pathological features 13) may be obtained from the pathology report via a pre-trained model 12.

In operation 302, a second dataset of a clinical trial may be obtained. In some embodiments, the second dataset of the clinical trial may be obtained from a clinical database.

In operation 303, a determination may be made as to whether the first dataset and the second dataset match with respect to a first set of fields. In some embodiments, operation 303 may correspond to operation 151 in FIG. 1 . The first set of fields may include one or more of the following: estimated glomerular filtration rate (EFGR), surgical procedure, histology, pathological stage, age, sex, or smoking.

In operation 304, when the first data set and the second data set match with respect to the first set of fields, a correlation value between the first data set and the second data set with respect to the second set of fields may be determined. In some embodiments, operation 304 may correspond to operation 152 in FIG. 1 . The second set of columns may include one or more of the following: ALK, ROS1, KRAS, BRAF, RET, NTRK, MET, P53, Her2, tumor size, tumor maximum diameter, programmed death ligand 1 (PD-L1), lymph node metastasis, distant metastasis, CNS metastasis, bone metastasis, wild-type, anti-angiogenesis, platinum, EGFR TKI, ALK inhibitor, PD-1/PD-L1 inhibitor, CTLA-4 inhibitor, radiation therapy, cisplatin/carboplatin, chemotherapy, systemic therapy, disease status, or Eastern Cooperative Oncology Group Performance Status (ECOG PS).

In operation 305, when the correlation value exceeds a critical value, the clinical test may be determined to be recommended. When the correlation value of the clinical test exceeds the critical value, it indicates that the clinical test is relevant to the patient and, therefore, the clinical test may be recommended for the patient.

FIG4A illustrates a flow chart of a method 40 for pre-training a model for extracting features from pathology reports according to some embodiments of the present disclosure. In some embodiments, the model can be pre-trained by inputting unlabeled word content, such as pathology reports of several patients.

In operation 401, the content of the pathology report may be divided into a plurality of sequences according to a predetermined length. In some embodiments, each of the sequences may include a plurality of sentences.

In operation 402, a classification token may be added to the beginning of each of a plurality of sequences. In some embodiments, the sequence of a pathology report may be one or more paragraphs that can be identified by a clinician. The classification token may represent a vector of the entire sequence.

In operation 403, a sentence separator may be added between two consecutive sentences. In some embodiments, the sentence separator may be used to identify different sentences. In some embodiments, each of the two consecutive sentences may include one or more sentences identified by a clinician.

In operation 404, pre-processing may be performed on the content of the pathology report to obtain token embeddings, sentence embeddings, and position embeddings. In some embodiments, the token embeddings may be a value representation of the content. The sentence embeddings may be a value representation of the sentence. The position embeddings may be a position representation of the content.

In operation 405, the token embeddings, sentence embeddings, and position embeddings may be aggregated into pre-processed content. In some embodiments, the token embeddings, sentence embeddings, and position embeddings may be aggregated, the aggregated content may be copied into three copies, and a multi-head self-attention algorithm may then be applied to the three copies to obtain pre-processed content containing a representation vector for each word.

In operation 406, the model may be trained by performing masked language modeling and/or next sentence prediction on the pre-processed content to obtain a pre-trained model.

In some embodiments, a masked language model can easily predict target terms in a multi-layered context. The masked language model allows portions of the input term (e.g., some phrasing of a particular term) to be simply randomly masked, and then the masked term can be predicted. In some embodiments, the input term can be converted into tokens for analysis.

In some embodiments, to train a model that understands sentence relationships, next sentence prediction can be used to train the model. In some embodiments, next sentence prediction is a task that can be easily generated from a corpus/database. Specifically, when sentences A and B are selected for each relevant example, 50% of the time, B is the actual next sentence after A, and 50% of the time, it is a random sentence from the corpus. After training with a large amount of input data, the accuracy of the pre-trained model's next sentence prediction can be increased.

In some embodiments, the pre-trained model can be a Bidirectional Encoder Representations from Transformers (BERT) model. The pre-trained model can be pre-trained by inputting clinical pathology reports from one or more hospitals.

FIG4B illustrates a flow chart of a method 41 for extracting features from a pathology report according to some embodiments of the present disclosure. In some embodiments, method 41 may be performed using a pre-trained model trained according to method 40 in FIG4A .

In operation 411, a classification task may be performed on the pathology report using a pre-trained model, resulting in at least one status value. In some embodiments, the classification task involves determining whether the pathology report includes a specific field. Therefore, the answer/result of the classification task will be yes or no (1 or 0). In other words, the result of the classification task is a status value.

In some embodiments, at least one status value in the pathology report may include a status value for a field (or condition) including: EGFR, ALK, ROS1, KRAS, BRAF, RET, NTRK, MET, P53, or Her2. In some embodiments, at least the status value may be included in the first dataset discussed in FIG. 3 .

For example, the status value for the EGFR field can include 2 ^<4> + 1 possible values, such as mutation status and unknown status for exons 18, 19, 20, and 21. For the ALK, ROS1, KRAS, BRAF, RET, NTRK, MET, P53, and Her2 fields, the status value can be positive, negative, or unknown.

In operation 412, a sequence labeling task may be performed on the pathology report using a pre-trained model to obtain at least one description. In some embodiments, the sequence labeling task may determine specific terms of different categories. Therefore, the answer/result of the sequence labeling task will be a description.

In some embodiments, at least one description in the pathology report includes a description of a field (or condition) including: procedure (or surgical procedure), histology, tumor size, stage (or pathological stage), or PDL1. In some embodiments, at least the description may be included in the first dataset discussed in FIG. 3 .

For example, the description for the procedure field might be "Video-assisted thoracic surgery (VATS) lobectomy." The description for the histology field might be "Poorly differentiated non-small cell carcinoma." The description for the tumor size field might be "0.6 × 0.4 × 0.3 cm," and the description for the maximum tumor diameter might be "0.6 cm." The description for the stage field might be "pStageIVA."

FIG5 illustrates a flow chart of a method 50 for collecting clinical trials according to some embodiments of the present disclosure. A device performing method 50 can be used to update a clinical trial database, which can be coupled with the clinical trial matching system in FIG1 .

In operation 501, one or more keywords may be searched on one or more online clinical trial databases to obtain one or more query results. The one or more online clinical trial databases may be government public clinical trial databases (e.g., clinicaltrials.gov and www1.cde.org.tw/ct_taiwan). In some embodiments, the keywords may be diseases/diagnoses and/or stages. In some embodiments, the diseases/diagnoses may include NSCLC (non-small cell lung cancer), non-small cell lung cancer, lung adenocarcinoma, non-squamous, squamous cell carcinoma, non-squamous non-small cell lung cancer, squamous cell lung cancer, large cell lung cancer, etc. In some embodiments, the stage may include progressive, stage IIIB, stage IIIC, stage IV, metastatic, etc. For example, the keywords may be NSCLC advanced stage, NSCLC stage IIIB, etc.

In operation 502, one or more parameters of each of the query results may be recorded.

In operation 503, a website link for each of the query results may be constructed based on one or more parameters.

In operation 504, data for one or more fields of the query results may be collected. In some embodiments, the fields may be columns of interest in the query results. For example, fields may include keywords, clinical trial/program ID, clinical trial/project title, applicant, sponsor, expected clinical trial start date, actual clinical trial start date, expected clinical trial end date, actual clinical trial end date, clinical trial inclusion criteria, and clinical trial exclusion criteria, trial hospital, trial location (e.g., state or country), expected number of trials in Taiwan, expected number of trials worldwide, last updated date of the clinical trial, and website link (i.e., URL) of the clinical trial.

At operation 505, data from one or more fields of the query result may be stored in a clinical trial database. In some embodiments, the clinical trial database may be coupled with a clinical trial matching system, enabling the system to compare pathological characteristics and demographic data with clinical trials. Thus, recommended clinical trials may be found for the patient.

FIG6 is a schematic diagram illustrating a representation of a clinical trial matching system according to some embodiments of the present disclosure. Referring to FIG6 , the clinical trial matching system may include a demographic block, a gene/transfer block, a treatment/drug block, and a pathology information block.

Data in the demographic block, such as age, sex, smoking, and ECOG PS, can be obtained from demographic data 14 in Figure 1. For example, the age may be 50. The sex may be male. The patient may have a smoking habit. The ECOG PS score may be 3, which can range from 0 to 5.

In some embodiments, data 601 in a gene/transfer block can be obtained from a pathology report via a pre-trained model. That is, data 601 can be obtained from pathological features 13 in FIG. 1 . On the other hand, data other than data 601 in a gene/transfer block can be obtained from demographic data 14 in FIG. 1 . In some embodiments, each of the data in a gene/transfer block can be a status value. For the EGFR field, it displays the number of exons (exons) with "unknown" or mutations; the status value "18, 19" indicates that exons 18 and 19 have mutations. For other fields in data 601, the status value can be P (positive), N (negative), or U (unknown). For example, lymph node metastasis could be Yes (i.e., lymph node metastasis has occurred).

In the Gene/Metastasis section, the status values for "Wild Type," "Lymph Node Metastasis," "Distant Metastasis," "CNS Metastasis," and "Bone Metastasis" can be either Yes or No.

The data in the treatment/drug block can be obtained from demographic data 14 in Figure 1. In some embodiments, each of the data in the treatment/drug block can be a status value, which can be either yes or no. For example, radiation therapy can be yes (i.e., radiation therapy has been administered).

In some embodiments, data 602 in the pathology information block can be obtained from pathology reports via a pre-trained model. Specifically, data 602 can be obtained from pathology features 13 in FIG. 1 . Alternatively, data other than data 602 in the pathology information block can be obtained from demographic data 14 in FIG. 1 .

FIG7 is a schematic diagram illustrating a computer system according to some embodiments of the present disclosure.

Referring to FIG. 7 , an example of a computer system 700 capable of performing one or more operations of the disclosed methods is shown. In at least some embodiments of the present disclosure, the computer system 700 includes a computing device 710 and a database 720. The computing device 710 can be a server computer, a client computer, a personal computer (PC), a tablet PC, a set-top box (STB), a personal digital assistant (PDA), a cellular phone, or a smartphone. The computing device 710 includes a processor 711, an input/output interface 712, a communication interface 713, and a memory 714. The database 720 can store pathology reports from which pathology features 13 and demographic data 14 are to be extracted. The database 720 can store pathology reports to be analyzed or summarized. The input/output interface 712 is coupled to the processor 711. The input/output interface 712 allows a user to operate the computing device 710 to perform the operations or methods disclosed herein (e.g., the method disclosed in FIG. 3 ). The communication interface 713 is coupled to the processor 711. The communication interface 713 allows the computing device 710 to communicate with the database 720. The communication interface 713 may support one or more of the following protocols: Universal Serial Bus (USB), Ethernet, Bluetooth, IEEE 802.11, 3GPP Long Term Evolution (LTE) (4G), and 3GPP New Radio (5G). The memory 714 may be a non-transitory computer-readable storage medium. The memory 714 is coupled to the processor 711. The memory 714 stores program instructions that can be executed by one or more processors (e.g., processor 711). When the program instructions stored in the memory 714 are executed, the program instructions cause one or more operations of the method disclosed in this disclosure to be performed.

For example, the program instructions may cause the computing device 710 to perform a set of actions including at least: obtaining a first data set from a pathology report; obtaining a second data set from a clinical trial; determining whether the first data set and the second data set match with respect to a first set of fields; when the first data set and the second data set match with respect to the first set of fields, determining a correlation value between the first data set and the second data set with respect to the second set of fields; and when the correlation value exceeds a threshold value, determining that the clinical trial is recommended.

The scope of this disclosure is not intended to be limited to the specific embodiments of the processes, machines, articles, compositions of matter, means, methods, steps, and operations described in this specification. Those skilled in the art will readily understand from the disclosure of this disclosure that currently existing or later developed processes, machines, articles, compositions of matter, means, methods, steps, or operations that perform substantially the same functions or achieve substantially the same results as the corresponding embodiments described herein can be utilized in accordance with this disclosure. Therefore, the appended claims are intended to include within their scope such processes, machines, articles, compositions of matter, means, methods, steps, or operations. Furthermore, each claim constitutes a separate embodiment, and the combination of each claim and embodiment is within the scope of this disclosure.

The methods, procedures, or operations according to the embodiments of the present disclosure may also be implemented on a programmed processor. However, the controller, flowchart, and module may also be implemented on a general-purpose or special-purpose computer, a programmed microprocessor or microcontroller and peripheral integrated circuit components, an integrated circuit, hardware electronics such as discrete component circuits or logic circuits, a programmable logic device, or the like. In general, any device having a finite state machine capable of implementing the flowcharts shown in the figures may be used to implement the processor functions of the present disclosure.

Alternative embodiments preferably implement the methods, procedures, or operations according to embodiments of the present disclosure in the form of a non-transitory computer-readable storage medium storing computer-programmable instructions. These instructions are preferably executed by a computer-executable component that is preferably integrated with the network security system. The non-transitory computer-readable storage medium can be stored on any suitable computer-readable medium, such as RAM, ROM, flash memory, EEPROM, optical storage (CD or DVD), hard drive, floppy drive, or any other suitable device. The computer-executable component is preferably a processor, but the instructions may alternatively or additionally be executed by any suitable dedicated hardware device. For example, one embodiment of the present disclosure provides a non-transitory computer-readable storage medium having computer-programmable instructions stored therein.

Although the present disclosure has been described using specific embodiments thereof, it is apparent that many alternatives, modifications, and variations may be apparent to those skilled in the art. For example, in other embodiments, various components of the embodiments may be interchanged, added, or substituted. In addition, not all elements shown in the figures are required for the operation of the disclosed embodiments. For example, it will enable those skilled in the art to carry out and utilize the teachings of the present disclosure by employing only the elements of the individual claims. Therefore, the embodiments of the present disclosure as described herein are intended to be illustrative and not restrictive. Various modifications may be made without departing from the spirit and scope of the present disclosure.

Although the numerous features and advantages of the present disclosure have been described in the foregoing description, along with details of its structure and function, this disclosure is illustrative only. The details, particularly in the shape, size, and arrangement of parts, may be varied substantially within the principles of the invention as indicated by the broad general meaning of the terms used in the appended claims.

11: Pathology Report 12: Pretrained Model 13: Pathology Features 14: Demographic Data 15: Clinical Trial Matching System 151: Operation 152: Operation 153: Operation 154: Operation

Claims (17)

A method for matching clinical trials comprises: obtaining a first dataset comprising pathological feature data and demographic data from a pathology report, wherein at least one description in the first dataset is obtained by performing a sequence tagging task on the pathology report using a pre-trained model, and the at least one description in the first dataset comprises a description of at least one of the following fields: surgery, histology, tumor size, stage, or PDL1; obtaining a clinical trial A method for determining whether the first data set and the second data set match with respect to a first condition set is provided. The method comprises: determining a correlation value between the first data set and the second data set with respect to a second condition set when the first data set and the second data set match with respect to the first condition set; and determining that the clinical trial is recommended when the correlation value exceeds a threshold value. The method of claim 1, wherein the relevance value is a sum of an individual relevance value for each of the second condition sets. The method of claim 2, wherein the individual relevance value is associated with an individual assigned weight (Wq). The method of claim 2, wherein the respective relevance value is associated with a respective inverse document frequency (IDF) of a respective keyword. The method of claim 1, wherein the first set of conditions includes one or more of: estimated glomerular filtration rate (EFGR), surgery, histology, pathological stage, age, sex, or smoking. The method of claim 1, wherein the second condition set comprises one or more of the following: ALK, ROS1, KRAS, BRAF, RET, NTRK, MET, P53, Her2, tumor size, tumor maximum diameter, programmed death-ligand 1 (PD-L1), lymph node metastasis, distant metastasis, CNS metastasis, bone metastases, wild type, anti-angiogenesis, platinum, EGFR TKIs, ALK inhibitors, PD-1/PD-L1 inhibitors, CTLA-4 inhibitors, radiation therapy, cisplatin/carboplatin, chemotherapy, systemic therapy, disease status, or Eastern Cooperative Oncology Group Performance Status (ECOG PS). The method of claim 1, wherein obtaining the first data set comprises: performing a classification task on the pathology report using the pre-trained model to obtain at least one status value in the first data set. The method of claim 7, wherein the classification task is performed to obtain a status value of one of the following fields: EGFR, ALK, ROS1, KRAS, BRAF, RET, NTRK, MET, P53, or Her2. The method of claim 1, wherein the pre-trained model is trained by a masked language model and/or next sentence prediction. A device for matching clinical trials, comprising: a processor; and a memory coupled to the processor, wherein the processor executes computer-readable instructions stored in the memory to perform operations, and the operations include: obtaining a first data set including pathological feature data and demographic data from a pathology report, wherein at least one description in the first data set is sequence-labeled on the pathology report by a pre-trained model The invention relates to a method for obtaining a clinical trial based on a first dataset, wherein the at least one description in the first dataset includes a description of at least one of the following fields: surgery, histology, tumor size, stage, or PDL1; obtaining a second dataset including pathological feature data and demographic data of a clinical trial; determining a correlation value between the first dataset and the second dataset with respect to a third condition set; and determining that the clinical trial is recommended when the correlation value exceeds a threshold value. The apparatus of claim 10, further comprising: determining whether the first data set and the second data set match a fourth condition set, wherein the correlation value is determined when the first data set and the second data set match the fourth condition set. The apparatus of claim 10, wherein the relevance value is a sum of an individual relevance value for each of the third condition sets. The apparatus of claim 12, wherein the respective relevance value is associated with a respective inverse document frequency (IDF) for a respective keyword. The device of claim 11, wherein the fourth set of conditions includes one or more of: estimated glomerular filtration rate (EFGR), surgery, histology, pathological stage, age, sex, or smoking. The device of claim 10, wherein the third condition group includes one or more of the following: ALK, ROS1, KRAS, BRAF, RET, NTRK, MET, P53, Her2, tumor size, tumor maximum diameter, programmed death ligand 1 (PD-L1), lymph node metastasis, distant metastasis, CNS metastasis, bone metastasis, wild type, anti-angiogenesis, platinum, EGFR TKI, ALK inhibitor, PD-1/PD-L1 inhibitor, CTLA-4 inhibitor, radiation therapy, cis-platinum/carboplatin, chemotherapy, systemic therapy, disease status, or Eastern Cooperative Oncology Group performance status (ECOG PS). The device of claim 10, wherein obtaining the first data set comprises performing a classification task on the pathology report using the pre-trained model to obtain at least one status value in the first data set. A non-transitory computer storage medium having program instructions stored thereon, wherein the program instructions, when executed by a processor, cause the processor to perform operations, the operations comprising: obtaining a first data set comprising pathological feature data and demographic data from a pathology report, wherein at least one description in the first data set is obtained by performing a sequence labeling task on the pathology report using a pre-trained model, and the at least one description in the first data set comprises a description of at least one of the following fields: The method comprises the steps of: determining surgery, histology, tumor size, stage, or PDL1; obtaining a second dataset of a clinical trial including pathological feature data and demographic data; determining whether the first dataset and the second dataset match with respect to a fifth condition set; determining a correlation value between the first dataset and the second dataset with respect to a sixth condition set when the first dataset and the second dataset match with respect to the fifth condition set; and determining that the clinical trial is recommended when the correlation value exceeds a threshold value.