A non-invasive early screening method and system for cancer based on cfDNA terminal sequence characteristics

Abstract

The present invention discloses a non-invasive early screening method and system for cancer based on the terminal sequence characteristics of cfDNA. The method uses a relatively low-depth whole-genome sequencing method to perform statistical analysis on the distribution frequency change trends of three consecutive groups of 3bp base combination categories upstream and downstream of the break position of cfDNA from tumors and cfDNA from healthy individuals, establish an early screening model for cancer, and achieve non-invasive early screening for cancer. The present invention draws an overall map based on the distribution of the types of combined DNA motif combinations in units of 3 bases for 9 consecutive positions at the 5' end, and uses the frequency as a training feature to establish an early diagnosis model for biliary and pancreatic tumors in the studied cohort, and evaluates the effectiveness of the model. This method uses a more accurate method of analyzing tumor DNA in the blood to find clues for early diagnosis of tumors, providing more solid and reliable data support for clinical precision applications.

Description

Technical Field

[0001] The present invention belongs to the field of medical detection technology, and specifically relates to a non-invasive early screening method and system for cancer based on cfDNA terminal sequence characteristics. Background Art

[0002] In recent years, liquid biopsy technology has been widely used in clinical practice, especially in assisting the diagnosis, treatment and postoperative monitoring of tumor patients. Compared with traditional intraoperative sampling, liquid biopsy obtains samples by drawing blood. There are free nucleic acid molecules (cfDNA) in plasma. In healthy people, cfDNA mainly comes from the natural apoptosis of lymphocytes in the blood. After a series of digestion treatments, the DNA molecules in the cell nucleus become fragmented nucleic acid molecules and are released into body fluids such as blood. When tumors occur in tissues, a large number of fragmented nucleic acid molecules of specific tissue tumor cells are released into the blood. At present, the conventional method for studying liquid biopsy and early cancer screening is to identify cfDNA released by tumors by detecting mutations in cancer-specific oncogenes or tumor suppressor genes. Whole genome sequencing (WGS) of blood cfDNA can identify chromosomal abnormalities in cancer patients, but because the number of abnormal chromosomal changes accounts for a small proportion of tumor-derived cfDNA, especially in the early stages of cancer, it may be challenging to detect such changes. A common limitation of using cfDNA mutation detection is the requirement to detect mutation differences at the genomic level that distinguish cfDNA, such as fetuses and mothers in non-invasive prenatal diagnosis, and tumors and normal people in tumor diagnosis. Diseases such as myocardial infarction, stroke, and autoimmune diseases are associated with elevated cfDNA levels, which may be the result of tissue damage, but due to the lack of differences in this DNA mutation change, it is not possible to monitor it specifically through cfDNA, which is very similar to the early stages of tumors. In addition, not all ctDNA derived from cancer cells carries mutation information, which urgently requires us to provide a new and more sensitive method to change the current conventional means of cfDNA detection.

[0003] The chromatin state within cells is not completely consistent. The open chromatin area shows loose nucleosome connections, which is conducive to the binding and function of transposases and other cell function regulators. Different cell populations have different functions to perform, so the open chromatin areas of different cell populations are also inconsistent. After tumor cells mutate, cell functions change, and the open chromatin areas also change compared to normal cells. Therefore, when blood cfDNA is produced by enzymatic hydrolysis, it will be preferentially cut more frequently at some open positions, resulting in position preference. Based on these theoretical foundations, some new major breakthroughs have been made in the field of cfDNA cancer liquid biopsy. The research group of Professor Dennis Lo of the Chinese University of Hong Kong, China, the founder of non-invasive prenatal testing (NIPT), confirmed that cfDNA end analysis technology can greatly improve the sensitivity of liquid biopsy based on somatic mutation detection, and is expected to become a cost-effective new cancer detection method. The study analyzed plasma samples from liver cancer, liver transplantation and hepatitis patient groups and found that the cancer signal of blood ctDNA can be found by tracking the preferred end coordinates of cfDNA, and the cancer signal can be tracked using specific cfDNA end coordinates (PNAS, 2018, doi / 10.1073 / pnas.1814616115). Through the analysis of these samples, the researchers found that the end coordinates of DNA fragments can be used as a marker for cancer, further promoting the in-depth research and clinical transformation of liquid biopsy in the field of early cancer screening. Using a group of plasma samples from liver transplant cases, they first identified nearly 10,000 liver-related preferred end coordinate sites, proving for the first time that at certain genomic coordinates, cfDNA molecules from the liver are more likely to have preferred ends than non-liver-derived molecules. Afterwards, the sequencing data of a group of liver cancer patients were analyzed and compared with samples from chronic hepatitis B patients, and the terminal coordinates of tumor-specific plasma DNA were found, among which 9.8 million and 13.8 million preferred terminal coordinate sites were identified in HCC and chronic HBV infection patients. The researchers said that the abundance of these circulating DNA molecules with cancer-specific terminal coordinates was also closely related to the patient's tumor DNA portion. At the same time, there are millions of tumor-related plasma DNA terminal coordinates throughout the genome, indicating that they are easier to detect than somatic mutations. This terminal coordinate-based analysis method may provide a more efficient and economical way to find signs of cancer in the blood. Therefore, combining somatic mutation detection with the terminal coordinates of tumor-associated DNA is expected to greatly improve the sensitivity of liquid biopsy. The analysis also found that the abundance of these cfDNA molecules with cancer-specific terminal coordinates was also closely related to the patient's tumor DNA composition, even in those samples that were only shallowly sequenced. This shows that somatic mutations and tumor-related preferred terminal abundance can be used as a recognizable feature of ctDNA.

[0004] At present, there are no reports on the use of cfDNA terminal base sequence to study the early screening of liquid biopsy for gallbladder and pancreatic tumors. Among them, the most relevant study to the present invention (PMID: 32111602): Based on the preference of the terminal position, the literature Jiang, P., et al., Plasma DNA end motif profiling as a fragmentomic marker incancer, pregnancy and transplantation. Cancer Discovery, 2020: p. CD-19-0622. The authors first proposed and verified the advantages of certain nucleotides at the terminal position of cfDNA. The study statistically analyzed the frequency of 256 combinations of 5' terminal four nucleotides in an independent cohort of hepatocellular carcinoma, HBV virus carriers and healthy individuals and found that the distribution of terminal motifs CCCA, CCAG, and CCTG in healthy individuals was significantly higher than that in hepatocellular carcinoma patients, while the distribution of TAAA, AAAA, and TTTT in hepatocellular carcinoma was significantly increased. However, this study only preliminarily investigated the preference of the tissue-specific cfDNA terminal base sequence motif in hepatocellular carcinoma, and the sample size was small, which has certain limitations.

[0005] As research progresses, we have found that compared with detection methods based on gene mutations, fragment end information detection may provide a cheaper and more sensitive method for finding cancer signals. Existing technologies generally connect specific markers to open regions of chromatin, capture DNA fragments in the region and sequence them to obtain effective diagnostic information for ctDNA; the experimental steps are many, the cost is high and it is very easy to increase human and machine errors. In addition, the existing technology for predicting fragment preference end coordinates has complex bioinformatics. Different individuals and different sequencing platforms will affect the identification results. In addition, the number of identified sites with different terminal base preference sequence coordinates is large, and it is difficult to distinguish the truly contributing features, which brings obstacles to the practical application of the technology. Summary of the invention

[0006] The purpose of the present invention is to provide a non-invasive early screening method and system for cancer based on the terminal sequence characteristics of cfDNA, mainly to solve the technical problems of low specificity, high false positive rate and high cost in the early screening of gallbladder and pancreatic malignancies in the prior art.

[0007] The technical solution adopted by the present invention to solve the above technical problems is as follows:

[0008] A non-invasive early screening method for cancer based on the terminal sequence characteristics of cfDNA, the method is: by using a relatively low-depth whole-genome sequencing method, the difference statistics of the distribution frequency change trends of three groups of 3bp base combination categories upstream and downstream of the break position between tumor-derived cfDNA and cfDNA from healthy individuals are performed, an early screening model for cancer is established, and non-invasive early screening for cancer is achieved. The relatively low-depth whole-genome sequencing method refers to a sequencing depth of 2X to 4X.

[0009] As a preferred embodiment, the statistical method for the distribution frequency of three consecutive groups of 3bp base combination categories upstream and downstream of the break site of the cfDNA is: count 9 bases from the 6th base at the 5' end of the cfDNA and group three consecutive bases into three groups of 3bp motifs; and count the frequency of occurrence of the three groups of 3bp motifs for each sample.

[0010] As a preferred implementation scheme, a random forest algorithm is used, and a 5-fold cross-validation method is repeated 30 times to obtain model coefficients and establish an early screening model for cancer.

[0011] As a preferred embodiment, the cancer is a pancreatic or bile duct malignancy, which includes pancreatic cancer, gallbladder cancer and bile duct cancer.

[0012] The present invention also provides a non-invasive early screening system for cancer based on the terminal sequence characteristics of cfDNA, the system comprising:

[0013] The terminal Motif feature extraction module is used to obtain the terminal sequence feature data of cfDNA in the sample;

[0014] The machine learning classification model building module is used to establish an early screening model for cancer based on statistical differences in the frequency distribution trends of three consecutive groups of 3bp base combination categories upstream and downstream of the break site between tumor-derived cfDNA and cfDNA from healthy individuals.

[0015] As a preferred embodiment, the terminal Motif feature extraction module includes:

[0016] The sequencing data alignment unit is used to align the sequencing data to the human reference genome hg19 after removing the sequencing adapters of the sequencing data;

[0017] The reads filtering unit is used to filter and screen the sequencing data. The filtering criteria are: only consider reads aligned to autosomes 1-22; the quality score is greater than 20; the insert length is between 150 and 600; both ends must be properly paired; the reference region of the read does not contain degenerate bases;

[0018] The combined motif calculation unit is used to count 9 bases starting from the 6th base at the 5' end and divide them into three groups as its combined motif; for the 3' end read, 9 bases starting from the 6th base at the 3' end are counted in the order of 3' to 5' and divided into three groups, and then reverse complemented as its combined motif;

[0019] The combined motif statistics unit is used to count the frequencies of three groups of 3bp motifs for each sample. Each group of 3bp motifs has 64 different combinations, with a total of 192 frequencies as its training variables.

[0020] As a preferred embodiment, the machine learning classification model building module includes:

[0021] The sample data classification unit is used to divide the samples into a training set and a test set in a ratio of 4:1, and to keep the distribution ratios of healthy controls and samples of various cancer types in the two sets consistent;

[0022] The model parameter acquisition unit is used to process the sample data in the training set; in the training set, the random forest algorithm is used, and the 30-fold repeated 5-fold cross validation method is used to obtain the model parameters;

[0023] The model performance evaluation unit is used to draw the receiver operating characteristic curve of the training set based on the model prediction value and pathological detection results of each sample in the training set.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] At present, imaging diagnosis and serum index diagnosis are widely used in clinical practice to identify malignant tumors of the gallbladder and pancreas. The specificity is poor, resulting in false positives, and the diagnosis of cfDNA-related mutation information is biased towards the late stage. The present invention has the following characteristics for the study of cfDNA fragmentation distribution as an early diagnosis system for cancer: a. The low-depth whole genome sequencing method is adopted. Compared with ultra-high depth or high-depth target sequencing, the sequencing cost is greatly reduced, and the cost is reduced; b. Abnormal changes in fragments earlier than cfDNA mutations can be detected at an early stage of cancer, and the method is more sensitive than detecting cfDNA mutation information; c. The patients used are all early-stage patients, and the changes in related characteristics are detected earlier than the patients used in the closest research, and after systematic and scientific analysis and verification, it has better diagnostic efficacy than the classification model of existing related research.

[0026] The present invention uses 60 cases of pancreatic and biliary tumors and 31 healthy controls plasma cfDNA detected in the clinic to perform low-depth (2X-4X) whole genome testing, taking into account the preference of the base arrangement at the 5' end of plasma cfDNA, and the base distribution frequency of 9bp upstream and downstream of the cfDNA break position has a trend of change. The overall map is drawn based on the distribution of the combined DNA motifs of 3 bases in 9 consecutive positions at the 5' end, and the frequency is used as a training feature to establish an early diagnosis model for pancreatic and biliary tumors in the studied cohort, and the model's effectiveness is evaluated. This method uses a more accurate method of analyzing tumor DNA in the blood to find clues for early diagnosis of tumors, providing more solid and reliable data support for clinical precision applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is the ROC curve diagram of the training set in Example 2 of the present invention.

[0028] Figure 2 This is the ROC curve diagram of the test set in Example 2 of the present invention. DETAILED DESCRIPTION

[0029] The technical solution of the present invention is described in detail below in conjunction with the embodiments. The reagents and biological materials used below are all commercial products unless otherwise specified.

[0030] Example 1

[0031] (1) Study cohort and clinical information

[0032] This study included 60 patients diagnosed with pancreatic and biliary tumors (pancreatic cancer, gallbladder cancer, and bile duct cancer) by tumor markers, imaging examinations (such as ultrasound, abdominal CT scan, etc.) and pathological examination results, and 31 healthy people. Blood samples were collected from patients and healthy individuals before surgery. Each enrolled patient was given an accurate diagnosis based on the results of pathological examination after surgery.

[0033] (2) Blood sample collection, separation and storage

[0034] Whole blood from preoperative cancer patients and healthy controls was collected in 10 ml free nucleic acid storage tubes (REF43803, BD, USA) and transported at room temperature. The whole blood samples received were separated into plasma by two-step centrifugation. First, the plasma and cellular components were separated by centrifugation at 4°C and 1,600 g for 10 minutes. The supernatant was carefully aspirated, taking care not to aspirate the white blood cell layer. The hemolysis grade of the plasma was recorded at the same time. Samples with hemolysis grade ≥5 were not included in subsequent studies. Secondly, the plasma was centrifuged again at 16,000 g for 15 minutes at 4°C to remove any remaining cells or cell debris. The supernatant was transferred to a centrifuge tube and divided into 1 ml tubes. The separated plasma samples were stored in a -80°C refrigerator.

[0035] (3) Extraction of blood cfDNA

[0036] The plasma sample was taken out of the -80℃ refrigerator and placed in a water bath, incubated statically at 37℃ for about 5 minutes, and the plasma was transferred to a low-temperature refrigerated centrifuge and centrifuged at 4℃ and 1,600g for 10 minutes, and the supernatant was carefully aspirated into a centrifuge tube. The plasma cfDNA was extracted using the QIAamp Circulating Nucleic Acid Kit (55114, Qiagen, Shanghai, China) kit from 1ml of plasma. The specific operation steps refer to the product manual, and 30μl EB was finally used to elute the cfDNA. The total amount of extracted cfDNA was quantified using the Qubit fluorescence quantifier and the corresponding reagents (Q32854, Thermo Fisher, USA). The Agilent 2100 bioanalyzer and the corresponding Agilent High Sensitivity DNA Kit & Reagents (5067-4626, Agilent, USA) were used to detect the distribution of cfDNA fragments.

[0037] (4) Blood cfDNA library construction and WGS sequencing

[0038] The samples that passed the cfDNA quality control were used for cfDNA library construction and WGS sequencing. The KAPADNA Hyper Prep kit (KK8504, KAPA, USA) was used for library preparation. The detailed operation process is referred to the product manual. The input amount of each cfDNA sample was 10ng, and then the base ends were filled and A tails were added. Then, the adapters were connected, the adapters were purified, and the PCR amplification was performed for 7 cycles to enrich the library. After purification, the DNA was eluted with 25μl elution buffer, and the plasma cfDNA library concentration was determined by Qubit, and the fragment distribution of the plasma cfDNA library was determined by 4150. The library that passed the quality inspection was selected for whole genome sequencing using the NovoSeq 6000 platform, with a sequencing strategy of 2x150bp and a sequencing volume of ~10G (sequencing depth of about 3×).

[0039] (5) 5' end motif feature extraction

[0040] Based on DNA sequencing detection technology, cfDNA in the patient's peripheral blood is obtained. The analysis process of sequencing data is as follows:

[0041] 1) Sequencing data alignment. After removing the sequencing adapters of the DNA sequencing data, the sequencing data were aligned to the human reference genome hg19 (genome download link: ftp: / / ftp-trace.ncbi.nih.gov / 1000genomes / ftp / technical / reference / human_g1k_v37.fasta.gz) using BWA software (version: 0.7.17-r1188);

[0042] 2) Read filtering. Only reads that map to autosomes 1-22 are considered; the quality score is greater than 20; the insertion size is between 150 and 600 bp; both ends must be properly paired; the reference region of the read does not contain degenerate bases;

[0043] 3) Combined motif calculation. For the 5' end read, count 9 bases from the 5' end and divide three consecutive bases into three groups, which is its combined motif. Since the sequencing data is added with UMI, there is a UMI sequence in front of each read. The actual calculation is to count 9 bases from the 6th base (inclusive) of the 5' end and divide them into three groups as its combined motif. Similarly, for the 3' end read, count 9 bases from the 6th base (inclusive) of the 3' end in the order of 3' to 5' and divide them into three groups, and then reverse complement them as its combined motif;

[0044] 4) Combined motif statistics. Each group of 3bp motifs has 64 different combinations. There are 4 base choices at one position, and the total number of 3bp base combinations is 4 to the power of 3, for a total of 64 combinations. Three consecutive 3bp combinations (9bp base positions) are 64+64+64, for a total of 192 combinations. For each sample, the frequencies of its three groups of 3bp motifs are counted, and a total of 192 frequencies are used as its training variables.

[0045] (6) Establishment of machine learning classification model

[0046] 1) Divide the samples into training set and test set. Divide all samples into training set and test set in a ratio of 4:1, and keep the distribution ratio of healthy controls and various cancer samples in the two sets consistent;

[0047] 2) Process the sample data in the training set. In the training set, use 30 repetitions of the 5-fold cross validation method to obtain the model parameters;

[0048] 3) Evaluate the effectiveness of the model. According to the model prediction value and pathological test results of each sample in the training set, draw the receiver operating characteristic curve (ROC curve) of the training set. Based on the prediction value, set a series of thresholds to divide the training set into healthy people and cancer patients, and then use the pathological test results as the true value to evaluate the model prediction performance. The model prediction performance evaluation method includes the area under the ROC curve (AUC, Area UnderCurve, range 0 to 1), positive predictive value (PPV, Positive Predictive Value, range 0 to 1), specificity (range 0 to 1), accuracy (range 0 to 1) and sensitivity (range 0 to 1). The higher the value, the better the effect.

[0049] Example 2

[0050] (1) Study cohort and clinical information

[0051] A total of 60 patients diagnosed with pancreatic and biliary tumors (29 pancreatic cancers, 15 gallbladder cancers, and 16 bile duct cancers) and 31 healthy people were included in this study. Blood samples were collected from patients before surgery. Each enrolled cancer patient was given an accurate diagnosis based on the results of pathological examination after surgery. All samples were divided into training and test sets in a ratio of 4:1, and the distribution ratios of healthy controls and various cancer samples in the two sets were kept consistent. Table 1 shows the grouping information of healthy and patient controls. The analysis results showed that there was no significant difference in the gender ratio and the distribution ratio of the number of healthy controls and cancer patients between the two groups of samples.

[0052] Table 1: Queue information

[0053]

[0054]

[0055] (2) Health and cancer classification scoring model

[0056] Using the training set and the results of pathological tests, the Random Forest algorithm is used to build a scoring model for cancer patients and healthy people. The model consists of three parts: variables, model formulas, and predicted values. The process is as follows:

[0057] ① Model variables and parameters. In the training set, the model used the frequencies of 192 3bp motifs as feature variables (Table 2), and used 30 repeated 5-fold cross validation methods to obtain model coefficients.

[0058] Table 2: Example of model input variables

[0059] Serial number Model variables 3bp motif frequency 1 AAA Freq1 2 AAT Freq2 …… …… …… 191 CCG Freq191 192 CCC Freq192

[0060] ② Scoring model. The scoring model formula is as follows:

[0061]

[0062] where x i is the input variable, P k (x i ) is the probability that the k-th decision tree predicts that sample i is cancer. Each decision tree is constructed based on the Gini impurity. G The calculation formula is:

[0063] I G =2p c (1-p c )

[0064] Where Pc is the proportion of cancer samples in a node after each branch of the decision tree.

[0065] Using the classification model and the combined motif frequency of each sample, the class prediction result of each sample can be obtained.

[0066] ③Model effectiveness evaluation.

[0067] In order to build an early screening model for cancer patients, the training set samples were divided into healthy people and cancer patients using the predicted values. The pathological test results were taken as the true values, and the ROC curve of the training set was drawn according to the predicted values ​​and pathological results in the cohort. Its AUC value reached 1, see Figure 1 , which is the ROC curve of the training set. The PPV (accuracy), specificity and sensitivity predicted by the training model are 100%, 100% and 100% respectively. The model performance evaluation results are shown in Table 3. The results show that in the training set, this risk prediction model has high sensitivity and NPV, and the model has better predictive efficacy for early diagnosis of cancer.

[0068] ④Verification of the predictive effectiveness of the discriminant model.

[0069] In order to verify the effectiveness of the discriminant model, the test set patients were divided into healthy and cancer patient groups (the same as the training set) based on the threshold set by the prediction value. The pathological test results were used as the true value, and the effectiveness of the model was verified according to the classification model and prediction value determined in the training set. The ROC curve of the test cohort was drawn, and the AUC value of the test set reached 1, see Figure 2, which is the ROC curve of the test set. The model prediction performance was evaluated, including accuracy, specificity and sensitivity, which were 85.7%, 71.4% and 100% respectively (Table 3). The results showed that in the test set, this risk prediction model also had high specificity, sensitivity and accuracy, that is, the model prediction performance was better.

[0070] Table 3: 192 variables model performance evaluation

[0071]

[0072] The above are only some preferred embodiments of the present invention, and the present invention is not limited to the contents of the embodiments. For those skilled in the art, various changes and modifications can be made within the scope of the technical solution of the present invention, and any changes and modifications made are within the protection scope of the present invention.

Claims

1. A non-invasive early screening method for cancer based on the terminal sequence characteristics of cfDNA, the method comprising: performing statistical analysis of the distribution frequency trends of three consecutive groups of 3 bp base combination categories upstream and downstream of the break position between tumor-derived cfDNA and cfDNA from healthy individuals by whole-genome sequencing at a relatively low depth, establishing an early screening model for cancer, and realizing non-invasive early screening for cancer; the statistical method for the distribution frequency of three consecutive groups of 3 bp base combination categories upstream and downstream of the break position of cfDNA is as follows: counting 9 bases from the 5' end of cfDNA and grouping three consecutive bases into three groups of 3 bp motifs; and counting the occurrence frequencies of the three groups of 3 bp motifs for each sample; the cancer is pancreatic cancer, gallbladder cancer and / or bile duct cancer.

2. The non-invasive early screening method for cancer based on cfDNA terminal sequence characteristics according to claim 1, characterized in that: The random forest algorithm was used with 30 repeated 5-fold cross validation method to obtain the model coefficients and establish an early screening model for cancer.

3. A non-invasive early screening system for cancer based on cfDNA terminal sequence characteristics, characterized in that: The system comprises: The terminal Motif feature extraction module is used to obtain the terminal sequence feature data of cfDNA in the sample; A machine learning classification model building module is used to establish an early screening model for cancer based on statistical differences in the distribution frequency trends of three groups of 3 bp base combination categories upstream and downstream of the break position between tumor-derived cfDNA and cfDNA from healthy individuals; the cancer is pancreatic cancer, gallbladder cancer and / or bile duct cancer; The terminal Motif feature extraction module includes: The sequencing data alignment unit is used to align the sequencing data to the human reference genome hg19 after removing the sequencing adapters of the sequencing data; The reads filtering unit is used to filter and screen the sequencing data. The filtering criteria are: only consider reads aligned to autosomes 1-22; the quality score is greater than 20; the insert length is between 150 and 600; both ends must be properly paired; the reference region of the read does not contain degenerate bases; The combined motif calculation unit is used to count 9 bases from the 5' end of cfDNA and divide them into three groups as its combined motif; for the 3' end read, 9 bases are counted from the 3' end of cfDNA in the order of 3' to 5' and divided into three groups, and then reverse complemented as its combined motif; The combined motif statistics unit is used to count the frequencies of three groups of 3 bp motifs for each sample. Each group of 3 bp motifs has 64 different combinations, with a total of 192 frequencies as its training variables.

4. The non-invasive early screening system for cancer based on cfDNA terminal sequence characteristics according to claim 3, characterized in that: The machine learning classification model building module includes: The sample data classification unit is used to divide the samples into a training set and a test set in a ratio of 4:1, and to keep the distribution ratios of healthy controls and samples of various cancer types in the two sets consistent; The model parameter acquisition unit is used to process the sample data in the training set; in the training set, the random forest algorithm is used, and the 30-fold repeated 5-fold cross validation method is used to obtain the model parameters; The model performance evaluation unit is used to draw the receiver operating characteristic curve of the training set based on the model prediction value and pathological detection results of each sample in the training set.

Images