An artificial intelligence deep learning model for predicting the sensitivity of neoadjuvant chemotherapy in colorectal cancer

Abstract

This invention discloses an artificial intelligence deep learning model for predicting the sensitivity of neoadjuvant chemotherapy in colorectal cancer, belonging to the field of colorectal cancer treatment prediction technology. Its key technical points include the following steps: S1, data collection; S2, patient grouping; S3, screening predictive factors; S4, artificial neural network modeling; and S5, model validation. This invention successfully screens common clinical indicators for predicting the sensitivity of neoadjuvant chemotherapy in colorectal cancer and constructs an artificial neural network risk prediction model. This model can be applied to clinical practice, providing predictions for the sensitivity of neoadjuvant chemotherapy in colorectal cancer, thereby assisting clinicians in achieving more accurate predictions of chemotherapy efficacy. This invention helps optimize clinical treatment strategies, promotes the development of individualized neoadjuvant therapy regimens for colorectal cancer, and provides new ideas and methods for precision medicine.

Description

Technical Field

[0001] This invention relates to the field of colorectal cancer treatment prediction technology, specifically to the construction method and application of an artificial intelligence deep learning model for the sensitivity of colorectal cancer patients to neoadjuvant chemotherapy. Background Technology

[0002] Colorectal cancer is one of the most common malignant tumors in the world, with the highest incidence and mortality rates. In my country, it is also the most common gastrointestinal malignancy, with colon cancer accounting for approximately 40%. Clinically, colorectal surgeons most frequently encounter resectable colon cancer. Over the past few decades, the combined treatment approach of radical surgery and postoperative adjuvant chemotherapy has improved the 5-year overall survival rate of patients with radically resectable colon cancer by about 20%. However, postoperative local recurrence and distant metastasis remain the leading causes of death. Studies have shown that although postoperative adjuvant chemotherapy can improve disease-free survival in patients with stage II-III colon cancer, the 5-year recurrence rate may exceed 25%. Therefore, the efficacy of existing adjuvant chemotherapy strategies has not reached a completely satisfactory level and further optimization is needed.

[0003] Neoadjuvant chemotherapy, first proposed in 1982, refers to a comprehensive treatment regimen for patients with potentially resectable tumors. To reduce tumor stage, simplify surgery, or improve survival, systemic chemotherapy is administered before tumor resection or radiotherapy, followed by the completion of full-course chemotherapy after surgery or radiotherapy. Neoadjuvant chemotherapy was initially applied to malignant tumors such as breast cancer, gastric cancer, and esophageal cancer, achieving good results. Subsequent studies have also found that neoadjuvant chemoradiotherapy for advanced rectal cancer often results in satisfactory tumor downstaging, significantly reduces local recurrence rates, increases sphincter preservation rates, and improves postoperative organ function; some patients even achieve complete pathological remission. These findings have greatly inspired improvements in colon cancer treatment, leading researchers to explore the effects of neoadjuvant chemotherapy in colon cancer at different stages.

[0004] Compared to traditional surgery and postoperative chemotherapy, neoadjuvant chemotherapy can effectively shrink the primary tumor preoperatively, increase the resection rate of radical surgery, reduce intraoperative tumor cell seeding and dissemination, prevent metastasis and spread, effectively eliminate micrometastases and subclinical lesions, and reduce the risk of postoperative metastasis and recurrence. However, individual responses to neoadjuvant chemotherapy vary greatly, with approximately 20% of patients not benefiting from it. Therefore, exploring and assessing the sensitivity of colorectal cancer patients to neoadjuvant chemotherapy is of great significance for developing preoperative treatment plans. Some guidelines recommend determining the choice of neoadjuvant therapy based on tumor risk stratification. For advanced colorectal cancer patients with a low risk of local recurrence, neoadjuvant chemotherapy is not very meaningful, and radical resection surgery can be performed directly. However, there is currently no consensus on the indicators for predicting the sensitivity of neoadjuvant chemotherapy.

[0005] With the increasing availability of electronic health data, the application and exploration of more robust and advanced computational methods in disease prediction have become more practical. Machine learning algorithms detect useful information from large, unstructured, and complex datasets, and the predictive models they construct are constantly improving in their applicability and effectiveness in predicting diseases, leading to their widespread development and application in the medical field. Artificial neural networks are multi-layered complex models formed by neurons (perceptrons) connected by synapses (weights). They are mathematical models that simulate the learning process of the human brain to perform machine learning, pattern recognition, and prediction. They can efficiently mine deep information from large amounts of electronic health records using computers to construct accurate prediction, diagnosis, and prognosis models for diseases in various systems, thereby guiding clinical medical decision-making. However, currently, there are no applications of artificial intelligence-based deep learning predictive models in the area of ​​neoadjuvant chemotherapy sensitivity in colorectal cancer.

[0006] Therefore, this invention proposes to construct an artificial intelligence deep learning model for predicting the sensitivity of colorectal cancer to neoadjuvant chemotherapy. This model will help to more accurately assess the sensitivity of colorectal cancer patients to neoadjuvant chemotherapy before treatment, thereby providing stronger support for clinical decision-making, optimizing patient treatment plans, and improving the overall survival rate and quality of life of colorectal cancer patients. This invention aims to contribute new strategies and tools to personalized treatment and precision medicine for colorectal cancer. Summary of the Invention

[0007] The purpose of this invention is to address the aforementioned problems by providing an artificial intelligence deep learning model for predicting the sensitivity of neoadjuvant chemotherapy in colorectal cancer. This invention successfully screens common clinical indicators for predicting the sensitivity of neoadjuvant chemotherapy in colorectal cancer and constructs an artificial neural network risk prediction model. This model can be applied in clinical practice to predict the sensitivity of neoadjuvant chemotherapy in colorectal cancer, thereby assisting clinicians in achieving more accurate predictions of chemotherapy efficacy. This invention helps optimize clinical treatment strategies, promotes the development of individualized neoadjuvant therapy regimens for colorectal cancer, and provides new ideas and methods for precision medicine.

[0008] To achieve the above objectives, the technical solution of the present invention is as follows: This invention provides a method for constructing an artificial intelligence deep learning model to predict the sensitivity of colorectal cancer to neoadjuvant chemotherapy, characterized by the following steps: S1. Data collection: Collect data on radical resection colorectal cancer patients who received neoadjuvant chemotherapy as a training set for building a predictive model; S2. Patient grouping: Patients were divided into sensitive and non-sensitive groups based on the tumor regression grading (TRG) of the Chinese guidelines for the diagnosis and treatment of colorectal cancer to assess the sensitivity to neoadjuvant chemotherapy. S3. Screening predictive factors: By comparing the general information, laboratory test indicators and tumor characteristics of sensitive and non-sensitive patients, indicators that can be used to predict the sensitivity of neoadjuvant chemotherapy for colon cancer are screened out. S4. Artificial Neural Network Modeling: Use the "neuralnet" package in R software to draw an artificial neural network prediction model; S5. Model Validation: Collect data from radical resection colorectal cancer patients who also received neoadjuvant chemotherapy as a validation set to validate the artificial neural network prediction model.

[0009] Furthermore, in step S1, a total of 16 patient data items were included, including 2 general data items, 7 laboratory test indicators, and 7 tumor characteristics.

[0010] Furthermore, in step S3, t-test and chi-square test were used to screen out 6 indicators that can be used to predict the sensitivity of neoadjuvant chemotherapy for colorectal cancer, including the patient's preoperative lymphocyte count, preoperative T stage, pathological type, neutrophil-to-lymphocyte ratio (NLR), mismatch repair system (MMR) protein, and cell proliferation marker (Ki-67).

[0011] Further, in step S4, the neural network model can be divided into an input layer, an output layer, and a hidden layer. Each node in the input layer corresponds to a predictor variable. The six selected predictor factors are used as the input layer, and the patient's sensitivity to neoadjuvant chemotherapy is used as the output layer. The hidden layer is defined as 10. The artificial neural network prediction model is drawn using the "neuralnet" package in R software, and the importance ranking of the predictor factors is obtained.

[0012] Furthermore, in step S5, data on radical resection colorectal cancer patients who received neoadjuvant chemotherapy are collected as a validation set, input into the constructed model for validation, and the validated artificial neural network prediction model is output.

[0013] This invention also provides an application of the artificial intelligence deep learning model for predicting neoadjuvant chemotherapy sensitivity in colorectal cancer as described above, using the artificial neural network prediction model for neoadjuvant chemotherapy sensitivity in colorectal cancer constructed by the method for assessing the sensitivity of colorectal cancer patients to neoadjuvant chemotherapy.

[0014] The difficulty and significance of the technical problem solved by this invention lie in: Compared to traditional surgery and postoperative chemotherapy, neoadjuvant chemotherapy can effectively shrink the primary tumor before radical resection of colorectal cancer, improve the resection rate, and reduce the risk of postoperative metastasis and recurrence. However, individual responses to neoadjuvant chemotherapy vary greatly, and some patients who are not sensitive to it may delay surgery. Therefore, exploring and assessing the sensitivity of colorectal cancer patients to neoadjuvant chemotherapy is of great significance for developing preoperative treatment plans, but there is currently no consensus on the indicators for predicting the sensitivity of neoadjuvant chemotherapy.

[0015] With the rapid development of healthcare information technology, a large amount of electronic medical information remains to be explored. In the treatment of colorectal cancer, chemotherapy sensitivity may be affected by a variety of factors, and these factors may have nonlinear interactions. How to better handle these complex nonlinear relationships, simulate the complex patterns in biomedical data, and achieve more accurate, rapid, and comprehensive prediction of neoadjuvant chemotherapy sensitivity in colorectal cancer is also one of the problems to be solved.

[0016] Therefore, this invention aims to develop a deep learning-based model for predicting the sensitivity of neoadjuvant chemotherapy in colorectal cancer. By comprehensively considering the patient's general information, laboratory test data, and tumor characteristics, this invention hopes to fully explore the application potential of common clinical indicators in predicting the sensitivity of neoadjuvant chemotherapy in colorectal cancer. This invention aims to provide clinicians with a precise decision-making tool, enabling them to develop personalized treatment plans based on the patient's benefit from neoadjuvant chemotherapy. This will not only help optimize clinical treatment strategies and improve postoperative survival rates for colorectal cancer patients, but will also significantly improve patients' quality of life, paving new avenues for personalized treatment and precision medicine in colorectal cancer.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention fully explores the application potential of common clinical indicators in predicting the sensitivity of neoadjuvant chemotherapy in colorectal cancer by combining general patient information, laboratory test data, and tumor characteristics. Compared with traditional models that rely on only a single indicator, this invention is based on artificial intelligence deep learning, which can handle complex nonlinear relationships and quickly find optimal solutions to the problem, thus helping to improve the accuracy of predicting the sensitivity of colorectal cancer to neoadjuvant chemotherapy. Traditional treatment methods lack consideration for individual patient differences. This invention can help clinicians develop personalized treatment plans based on the patient's benefit from neoadjuvant chemotherapy, reduce ineffective treatments, improve treatment efficiency, and is expected to promote the development of colorectal cancer treatment towards personalization and precision. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the construction of an artificial intelligence deep learning model for the sensitivity of neoadjuvant chemotherapy to colorectal cancer in an embodiment of the present invention.

[0019] Figure 2This is a network structure diagram of the artificial neural network model in an embodiment of the present invention.

[0020] Figure 3 This is the importance ranking result of the artificial neural network model for predicting factors in this embodiment of the invention.

[0021] Figure 4 This is a receiver operating characteristic curve used in this embodiment of the invention to evaluate the predictive effect of the invention by using data from 118 patients with radical resection of colon cancer who received neoadjuvant chemotherapy as a training set.

[0022] Figure 5 This is a receiver operating characteristic curve (ROC) of 50 patients with radical resection of colon cancer who underwent neoadjuvant chemotherapy and received neoadjuvant chemotherapy, used as a validation set in an embodiment of the present invention to evaluate the predictive effect of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solution of the present invention will be described in further detail below with reference to the embodiments and accompanying drawings. Obviously, the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort should fall within the scope of protection of the present invention.

[0024] Example:

[0025] The solution provided by this invention is: an artificial intelligence deep learning model for predicting the sensitivity of colorectal cancer to neoadjuvant chemotherapy, and the method for constructing this prediction model is as follows: Figure 1 As shown, it includes the following steps.

[0026] Step 1: Data Collection In this embodiment, data from 118 patients with radical resection of colon cancer who underwent neoadjuvant chemotherapy and were admitted to the Second People's Hospital of Nanning City were collected as a training set to build a prediction model. A total of 16 preoperative characteristics of patients were included, including 2 general information items: gender (male, female) and age; 7 laboratory test indicators: neutrophil count, lymphocyte count, neutrophil-to-lymphocyte ratio (NLR), fibrinogen level, albumin level, carcinoembryonic antigen (CEA) level (≤5ng / ml, >5ng / ml), and carbohydrate antigen 199 (CA199) level (≤37U / ml, >37U / ml); and 7 tumor characteristics: maximum tumor diameter (≤5cm, >5cm), tumor location (left colon, right colon), tumor pathological type (mucinous adenocarcinoma, adenocarcinoma), preoperative T stage (T1-T2, T3-T4), preoperative N stage (N0, N+), MMR protein (dMMR, pMMR), and Ki-67 (high expression, low expression).

[0027] Step 2: Patient grouping: In this embodiment, the sensitivity to neoadjuvant chemotherapy was assessed according to the Tumor Regression Grading (TRG) system in the Chinese guidelines for the diagnosis and treatment of colorectal cancer, classifying patients into sensitive and insensitive categories. Complete tumor regression (TRG grade 0) was defined as no residual cancer cells under light microscopy; moderate tumor regression (TRG grade 1) was defined as a single or small focal residual cancer cell; mild tumor regression (TRG grade 2) was defined as residual tumor with abundant fibrotic stroma; and no tumor regression (TRG grade 3) was defined as extensive residual tumor with few or no necrosis of cancer cells. TRG grades 0-2 were classified as sensitive (labeled 0), and TRG grade 3 was classified as insensitive (labeled 1).

[0028] Step 3: Screening predictor factors: In this embodiment, by comparing the general information, laboratory test indicators and tumor characteristics of sensitive and non-sensitive patients, six indicators that can be used to predict the sensitivity of neoadjuvant chemotherapy for colorectal cancer were screened out using t-test and chi-square test. These indicators include the patient's preoperative lymphocyte count, preoperative T stage, pathological type, NLR, MMR protein, and Ki-67.

[0029] Step 4: Artificial Neural Network Modeling In this embodiment, the neural network model can be divided into an input layer, an output layer, and a hidden layer. Each node in the input layer corresponds to a predictive variable. The six selected predictive factors (preoperative lymphocyte count, preoperative T stage, pathological type, neutrophil-to-lymphocyte ratio (NLR), mismatch repair system (MMR) protein, and cell proliferation marker (Ki-67)) are used as the input layer, and the patient's neoadjuvant chemotherapy sensitivity (sensitive and insensitive) is used as the output layer. The hidden layer is defined as 10. The artificial neural network prediction model is drawn using the "neuralnet" package in R software, and the importance ranking of the predictive factors is obtained.

[0030] Visualization of the network structure diagram of the artificial neural network model ( Figure 2 In the connection weights between neurons in the network, positive weights are connected by blue lines, and negative weights are connected by gray lines. The thickness of the lines reflects the magnitude of the weights. The importance of each independent variable to the model's prediction results is also calculated and visualized. From the results, it can be found that... Figure 3 The most important variables for predicting the sensitivity of neoadjuvant chemotherapy in colorectal cancer patients are, in order of importance: pathological type, Ki-67, NLR, preoperative T stage, MMR protein, and lymphocyte count.

[0031] Using the predicted probability as the test variable and the sensitivity of colorectal cancer patients to neoadjuvant chemotherapy as the state variable, and assigning the state variable a value of "1", the receiver operating characteristic curve of the model training set is obtained (e.g., Figure 4 As shown in the figure, the AUC of the training set was 0.965 (95% CI: 0.933-0.998), indicating good model discrimination. The optimal cutoff value of the model was calculated to be 0.741 using the receiver operating characteristic curve. The sensitivity corresponding to the optimal cutoff value was 85.3%, and the specificity was 100%.

[0032] Step 5: Model Validation In this embodiment, in order to verify the predictive effect of the above-mentioned artificial neural network model on the sensitivity of neoadjuvant chemotherapy in colorectal cancer patients, data from 50 radical resection colorectal cancer patients who also received neoadjuvant chemotherapy at the Second People's Hospital of Nanning City were collected as a validation set.

[0033] The validation set receiver operating characteristic curves are shown (e.g.) Figure 5 As shown in the figure, the AUC value was 0.876 (95% CI: 0.776–0.976), the optimal cutoff value was 0.855, the sensitivity corresponding to the optimal cutoff value was 75.0%, and the specificity was 95.5%. This artificial neural network model performed well on the validation dataset and has certain practical value.

[0034] It is worth noting that, through the above embodiments of the present invention, common clinical indicators for predicting the sensitivity of neoadjuvant chemotherapy in colorectal cancer are screened, and an artificial neural network risk prediction model is constructed. This model can be applied to clinical practice, providing predictions for the sensitivity of neoadjuvant chemotherapy in colorectal cancer, thereby assisting clinicians in achieving more accurate predictions of chemotherapy efficacy. The present invention helps optimize clinical treatment strategies, promotes the development of individualized neoadjuvant therapy regimens for colorectal cancer, and provides new ideas and methods for precision medicine.

[0035] The above specific embodiments are merely explanations of the present invention and are not intended to limit the present invention. After reading this specification, those skilled in the art can make modifications to these embodiments without contributing any inventive step, but as long as they are within the scope of the claims of the present invention, they are protected by patent law.

Claims

1. An artificial intelligence deep learning model for predicting the sensitivity of neoadjuvant chemotherapy in colorectal cancer, characterized in that: Includes the following steps: S1. Data collection: Collect data on radical resection colorectal cancer patients who received neoadjuvant chemotherapy as a training set for building a predictive model; S2. Patient grouping: Patients were divided into sensitive and non-sensitive groups based on the tumor regression grading (TRG) of the Chinese guidelines for the diagnosis and treatment of colorectal cancer to assess the sensitivity to neoadjuvant chemotherapy. S3. Screening predictive factors: By comparing the general information, laboratory test indicators and tumor characteristics of sensitive and non-sensitive patients, indicators that can be used to predict the sensitivity of neoadjuvant chemotherapy for colon cancer are screened out. S4. Artificial Neural Network Modeling: Use the "neuralnet" package in R software to draw an artificial neural network prediction model; S5. Model Validation: Collect data from radical resection colorectal cancer patients who also received neoadjuvant chemotherapy as a validation set to validate the artificial neural network prediction model.

2. The method for constructing an artificial intelligence deep learning model for predicting neoadjuvant chemotherapy sensitivity in colorectal cancer according to claim 1, characterized in that: In step S1, a total of 16 patient data items were included, including 2 general data items, 7 laboratory test indicators, and 7 tumor characteristics.

3. The method for constructing an artificial intelligence deep learning model for predicting neoadjuvant chemotherapy sensitivity in colorectal cancer according to claim 1, characterized in that: In step S3, t-test and chi-square test were used to screen out 6 indicators that can be used to predict the sensitivity of neoadjuvant chemotherapy for colorectal cancer, including the patient's preoperative lymphocyte count, preoperative T stage, pathological type, neutrophil-to-lymphocyte ratio (NLR), mismatch repair system (MMR) protein, and cell proliferation marker (Ki-67).

4. The method for constructing an artificial intelligence deep learning model for predicting neoadjuvant chemotherapy sensitivity in colorectal cancer according to claim 1, characterized in that: In step S4, the neural network model can be divided into an input layer, an output layer, and a hidden layer. Each node in the input layer corresponds to a predictor variable. The six selected predictor factors are used as the input layer, and the patient's sensitivity to neoadjuvant chemotherapy is used as the output layer. The hidden layer is defined as 10. The artificial neural network prediction model is drawn using the "neuralnet" package in R software, and the importance ranking of the predictor factors is obtained.

5. The method for constructing an artificial intelligence deep learning model for predicting neoadjuvant chemotherapy sensitivity in colorectal cancer according to claim 1, characterized in that: In step S5, data on radical resection colorectal cancer patients who received neoadjuvant chemotherapy are collected as a validation set, input into the constructed model for validation, and the validated artificial neural network prediction model is output.

6. The method for constructing an artificial intelligence deep learning model for predicting neoadjuvant chemotherapy sensitivity in colorectal cancer as described in any one of claims 1-5, characterized in that: The artificial intelligence deep learning model for predicting neoadjuvant chemotherapy sensitivity in colorectal cancer, constructed using the aforementioned method, was used to assess the sensitivity of colorectal cancer patients to neoadjuvant chemotherapy.

Images