Rectum cancer neoadjuvant chemoradiotherapy curative effect prediction method and system based on magnetic resonance image analysis

Abstract

The invention provides a rectal cancer neoadjuvant chemoradiotherapy curative effect prediction method and system based on magnetic resonance image analysis, and aims to improve accurate prediction of curative effects of patients with local advanced rectal cancer (LARC). According to the method, multi-modal magnetic resonance images (T1WI, T2WI, DWI and DCE-MRI) are obtained, qualitative and quantitative analysis is combined, tumor morphology, diffusion characteristics and microcirculation parameters are extracted, clinical and pathological data of a patient are integrated, and a multi-dimensional feature set is constructed. Curative effect prediction is performed through machine learning and a deep learning model, and the model can predict treatment results of complete remission, partial remission, ineffectiveness and the like in a classified manner. The system accurately predicts the reaction of a patient to neoadjuvant radiotherapy and chemotherapy through a non-invasive method, supports formulation of a personalized treatment scheme, reduces excessive treatment, optimizes the treatment effect, and has good clinical application value.

Description

Technical Field

[0001] The present invention relates to the technical field of rectal cancer treatment auxiliary systems, and in particular to a method and system for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis. Background Art

[0002] Rectal cancer is one of the most common malignant tumors in the world, especially among digestive tract cancers. Patients with locally advanced rectal cancer (LARC) usually need to receive neoadjuvant chemoradiotherapy (NCRT) to reduce tumor volume and reduce tumor stage, thereby creating more favorable conditions for subsequent surgical treatment. However, clinically, it has been found that there are significant differences in the response of different patients to neoadjuvant chemoradiotherapy. Some patients can achieve complete remission (CR), while others have poor treatment effects or even no effect. Therefore, how to accurately predict the patient's response to neoadjuvant chemoradiotherapy before the start of treatment is of great significance for formulating personalized treatment plans, avoiding unnecessary treatment side effects, and improving treatment efficiency.

[0003] At present, the evaluation of the efficacy of chemoradiotherapy is mainly based on postoperative pathological examination and post-treatment imaging evaluation, such as traditional CT, MRI and other imaging methods. However, these methods can usually only provide efficacy information after the end of treatment, and cannot be predicted before or during chemoradiotherapy, lacking foresight. In addition, these traditional evaluation methods mainly rely on the subjective judgment of doctors, which may lead to a certain degree of misjudgment. With the development of medical imaging technology, magnetic resonance imaging (MRI) has gradually become an important tool for rectal cancer diagnosis and efficacy evaluation due to its advantages such as high soft tissue resolution and no radioactivity.

[0004] Especially in recent years, multi-parameter magnetic resonance imaging (mpMRI) technology has shown great potential in the qualitative and quantitative evaluation of tumors. mpMRI can combine multiple modalities such as T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI) and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) to provide rich histological and functional information. For example, diffusion-weighted imaging can reflect the cell density of the tumor and evaluate the freedom of water molecule movement in the tumor tissue by calculating the apparent diffusion coefficient (ADC); dynamic contrast-enhanced imaging can evaluate the blood perfusion state of the tumor and reflect changes in the tumor microcirculation. By comprehensively analyzing these imaging parameters, the morphological characteristics of the tumor and changes in its microenvironment can be comprehensively evaluated.

[0005] In addition, with the development of artificial intelligence technologies such as machine learning and deep learning, automated analysis methods based on medical imaging data have provided new ideas for predicting the effects of cancer treatment. Machine learning algorithms can accurately predict patients' responses to treatment by automatically extracting features from large-scale data sets and building classification or regression models. These algorithms can combine multi-dimensional information such as imaging data, clinical data, and pathological indicators to automatically screen out features that are highly correlated with efficacy and establish high-precision prediction models. Through these models, doctors can predict patients' sensitivity to radiotherapy and chemotherapy before treatment begins, thereby formulating more personalized treatment plans.

[0006] Although some studies have tried to use MRI images and machine learning algorithms to predict the efficacy of rectal cancer radiotherapy and chemotherapy, the following problems still exist:

[0007] Selection and fusion of image features: Traditional MRI image analysis methods mostly focus on a single modality, making it difficult to fully utilize the advantages of multimodal images (such as T1WI, T2WI, DWI, DCE-MRI), and lack effective fusion and comprehensive analysis of multimodal image features.

[0008] Generalization ability of machine learning models: During model training, how to avoid overfitting and improve the generalization ability of the model remains an important challenge. Especially on small sample medical datasets, the performance of the model is easily limited by the amount and quality of data.

[0009] Interpretability of clinical applications: Although deep learning models have achieved excellent performance in image recognition tasks, their "black box" characteristics lead to low interpretability of the models, which may cause certain trust barriers for clinicians when applying such models.

[0010] Therefore, there is an urgent need for a comprehensive method and system that combines multimodal magnetic resonance imaging, clinical and pathological data and uses machine learning algorithms to predict the efficacy of neoadjuvant chemoradiotherapy. This method can effectively predict the efficacy before treatment, help clinicians develop the best treatment plan based on the individual characteristics of patients, reduce treatment risks, and improve treatment effects.

[0011] The present invention aims to solve the above problems and provide a method and system for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis. It makes full use of multimodal MRI data, combines qualitative and quantitative analysis with clinical pathological indicators, and uses machine learning algorithms to construct a prediction model to provide LARC patients with personalized efficacy prediction plans, which has significant clinical application value. Summary of the invention

[0012] The purpose of the present invention is to address the deficiencies of the prior art and provide a method and system for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis, so as to solve the problems raised in the background technology.

[0013] To achieve the above object, the present invention provides the following technical solution: a method for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis, comprising the following steps:

[0014] Acquire multimodal magnetic resonance images of patients, including T1-weighted images (T1WI), T2-weighted images (T2WI), diffusion-weighted imaging (DWI), and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI);

[0015] The acquired multimodal magnetic resonance images were preprocessed, including noise removal, image normalization, registration, and tumor area segmentation, and the region of interest (ROI) was extracted;

[0016] The morphological characteristics of the tumor were obtained by qualitative analysis;

[0017] The diffusion characteristics and microcirculatory parameters of the tumor were obtained through quantitative analysis;

[0018] Combine clinical and pathological indicators to construct a feature set;

[0019] Use machine learning algorithms to establish efficacy prediction models, including feature extraction, selection, and model training;

[0020] Multimodal MRI images and clinical and pathological data of new patients were input to predict their sensitivity to neoadjuvant chemoradiotherapy.

[0021] As a preferred technical solution of the present invention, qualitative analysis includes evaluating the size, shape, boundary clarity and signal intensity of the tumor, and quantitative analysis includes calculating the apparent diffusion coefficient (ADC) value, blood flow (BF), blood volume (BV) and vascular permeability (PS) of the tumor.

[0022] As a preferred technical solution of the present invention, the apparent diffusion coefficient (ADC) is calculated by the following formula:

[0023]

[0024] Wherein, S(b) is the signal intensity of the diffusion weighted image when the diffusion sensitivity factor is b, S(0) is the signal intensity when the b value is 0, and b is the diffusion sensitivity factor.

[0025] As a preferred technical solution of the present invention, microcirculation parameters include blood flow (BF), blood volume (BV) and vascular permeability (PS), which are calculated by dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). The calculation formula of microcirculation parameters is as follows:

[0026]

[0027] Among them, K trans is the vascular permeability coefficient, C t (t) is the contrast agent concentration in the tumor tissue at time t, C p (t) is the contrast agent concentration in plasma at time t.

[0028] As a preferred technical solution of the present invention, clinical indicators include the patient's age, gender, tumor stage, and carcinoembryonic antigen (CEA) level; pathological indicators include Ki-67, p53, etc. All characteristics are normalized by the following formula:

[0029]

[0030] Where x is the original eigenvalue, x′ is the normalized eigenvalue, min(x) and max(x) are the minimum and maximum values ​​of feature x, respectively.

[0031] As a preferred technical solution of the present invention, the feature set is constructed by using a feature fusion method to combine imaging, clinical and pathological data to generate a feature vector F:

[0032] F=[F image , F clinical , F pathology ]

[0033] Among them, F image is the imaging feature, F clinical For clinical features, F pathology is the pathological feature. The feature vector is reduced in dimension by principal component analysis (PCA), and the formula is:

[0034] Z=XW

[0035] Among them, Z is the feature matrix after dimensionality reduction, X is the feature matrix, and W is the dimensionality reduction matrix.

[0036] As a preferred technical solution of the present invention, the efficacy prediction model adopts a machine learning algorithm, support vector machine (SVM), and its loss function is as follows:

[0037]

[0038] Among them, w is the weight vector, x i is the input feature, y i is the label, b is the bias, and C is the regularization parameter.

[0039] As a preferred technical solution of the present invention, the model training can adopt the Random Forest algorithm, and its integrated formula is:

[0040]

[0041] in, is the final prediction result, M is the number of trees, and fm(x) is the prediction result of the mth decision tree.

[0042] The present invention also provides a system for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer, comprising:

[0043] Image acquisition module, used to obtain T1WI, T2WI, DWI and DCE-MRI images of patients;

[0044] An image processing module, used for preprocessing the acquired magnetic resonance image and performing tumor area segmentation;

[0045] Feature extraction module, used to extract qualitative and quantitative image features;

[0046] A data entry module for entering clinical and pathological data of patients;

[0047] Model training module, used to train efficacy prediction models based on imaging, clinical and pathological features;

[0048] The prediction module is used to predict the efficacy of new patients.

[0049] As a preferred technical solution of the present invention, the model can integrate imaging, clinical and pathological data and use a neural network (such as a convolutional neural network, CNN) for deep learning, and its loss function is a cross entropy loss:

[0050]

[0051] Among them, y i is the actual label, p i is the probability value predicted by the model, and N is the number of samples. The model can automatically optimize and continuously improve the prediction performance.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] The present invention provides a method and system for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis. Through multimodal magnetic resonance imaging (such as T1WI, T2WI, DWI, DCE-MRI), rich tumor morphology and functional information are obtained, combined with clinical and pathological data, to comprehensively evaluate the patient's response to neoadjuvant chemoradiotherapy. This method improves the accuracy of efficacy prediction by integrating quantitative and qualitative features and using advanced machine learning and deep learning models, and can provide patients with personalized treatment plans, reducing the blindness of treatment and unnecessary side effects.

[0054] The present invention makes full use of machine learning technology to automatically screen imaging, clinical and pathological features, and continuously optimize and improve the prediction model. Through cross-validation and dynamic monitoring, the model can provide accurate efficacy predictions and support clinicians to make scientific decisions before treatment. Especially for patients who are insensitive or ineffective, the present invention can timely adjust the treatment strategy, optimize the treatment plan, and reduce treatment time and cost.

[0055] In summary, the present invention provides an efficient and accurate efficacy prediction tool through a non-invasive method, which has the functions of personalized treatment guidance and early intervention. Its wide application in neoadjuvant chemoradiotherapy for rectal cancer not only improves the prognosis of patients, but also optimizes the use of medical resources and promotes the precision and personalized development of rectal cancer treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A method flow chart of a method and system for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis;

[0057] Figure 2 This is a system block diagram of a method and system for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings and multiple embodiments.

[0060] In order to better explain the embodiments of the present invention, the following describes in detail the method and system for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis through several specific examples. Each example covers the complete process from image acquisition and processing to efficacy prediction.

[0061] Example 1: Prediction of therapeutic efficacy based on multimodal magnetic resonance images

[0062] participate Figure 1 and Figure 2 , patient data collection

[0063] Patients with LARC (locally advanced rectal cancer) were recruited and underwent standardized multimodal MRI scans, including T1-weighted images (T1WI), T2-weighted images (T2WI), diffusion-weighted imaging (DWI), and dynamic contrast-enhanced MRI (DCE-MRI). All images were obtained before treatment as baseline data.

[0064] Image preprocessing

[0065] The collected images are preprocessed, including the following steps:

[0066] Noise removal: Use median filtering or Gaussian filtering to denoise the image and reduce the impact of imaging noise on subsequent analysis.

[0067] Standardization: Standardize the grayscale values ​​of the image and normalize the grayscale values ​​of the image to the [0,1] interval to improve the consistency between images of different patients.

[0068] Image registration: Images of different modalities are registered using a rigid body registration algorithm to ensure that multimodal images of the same patient are aligned in the same space.

[0069] Tumor region segmentation

[0070] The tumor region is segmented on T2-weighted images and DWI images using a semi-automatic segmentation algorithm (such as region growing or convolutional neural network-based segmentation algorithm). The segmented region of interest (ROI) is used for subsequent feature extraction.

[0071] Qualitative and quantitative feature extraction

[0072] Qualitative features: By observing the tumor morphology of T2WI and DWI images, information such as tumor size, shape, boundary clarity, and signal uniformity can be extracted.

[0073] Quantitative characteristics: The apparent diffusion coefficient (ADC) value was calculated by the following formula:

[0074]

[0075] Wherein, S(b) is the signal intensity of the diffusion weighted image when the diffusion sensitivity factor is b, S(0) is the signal intensity when the b value is 0, and b is the diffusion sensitivity factor.

[0076] In addition, tumor microcirculation parameters such as blood flow (BF), blood volume (BV), and vascular permeability (PS) were extracted from DCE-MRI images, and the vascular permeability coefficient K was calculated. t rans

[0077]

[0078] Among them, K trans is the vascular permeability coefficient, C t (t) is the contrast agent concentration in the tumor tissue at time t, C p (t) is the contrast agent concentration in plasma at time t.

[0079] Feature Fusion

[0080] The qualitative and quantitative features extracted from imaging are fused with the clinical and pathological data of the patient. Clinical data include patient age, gender, tumor stage, CEA level, and pathological data include Ki-67, p53, etc. All data are normalized using the following formula:

[0081] Where x is the original eigenvalue, x′ is the normalized eigenvalue, min(x) and max(x) are the minimum and maximum values ​​of feature x, respectively.

[0082] Model training and efficacy prediction

[0083] The efficacy prediction model was trained by support vector machine (SVM). The model was optimized using the cross-validation method, using the following loss function:

[0084] Among them, w is the weight vector, x i is the input feature, y i is the label, b is the bias, and C is the regularization parameter.

[0085] After training, new patient characteristics are input into the model to predict their response to neoadjuvant chemoradiotherapy and output the patient's possible efficacy classification: complete remission (CR), partial remission (PR), stable disease (SD) or progressive disease (PD).

[0086] Example 2: Feature selection and prediction based on random forest

[0087] participate Figure 1 and Figure 2 , patient data collection and processing

[0088] Similar to Example 1, a multimodal magnetic resonance scan is performed on the patient and the image is preprocessed, including noise removal, standardization and registration.

[0089] Feature extraction

[0090] By segmenting the tumor area, imaging features are extracted from T2WI, DWI and DCE-MRI images, and clinical and pathological characteristics of the patient are extracted. Feature extraction includes the following:

[0091] Morphological characteristics: tumor volume, shape, etc.

[0092] Functional characteristics: apparent diffusion coefficient (ADC) values ​​and microcirculatory parameters (BF, BV, PS).

[0093] Feature Selection

[0094] The random forest algorithm is used for feature selection to screen out the features that contribute most to the prediction of therapeutic efficacy. The prediction formula of random forest is as follows:

[0095]

[0096] in, is the final prediction result, M is the number of trees, and fm(x) is the prediction result of the mth decision tree.

[0097] Model training and prediction

[0098] The selected features were used to train the efficacy prediction model, and the random forest algorithm was used to build a prediction model to predict the sensitivity of patients to neoadjuvant chemoradiotherapy based on their imaging, clinical and pathological characteristics. The model output results were complete remission (CR), partial remission (PR) or no effect (PD), and personalized treatment recommendations were provided based on these results.

[0099] Example 3: Multimodal image fusion based on deep learning

[0100] participate Figure 1 and Figure 2 , data preparation

[0101] T1WI, T2WI, DWI and DCE-MRI images were extracted from multimodal magnetic resonance images and preprocessed and regions of interest were segmented.

[0102] Multimodal image fusion

[0103] Convolutional Neural Network (CNN) is used for multimodal image fusion, taking image data of different modalities as input. Through multi-layer convolution operations, the features of each modality are extracted and fused in the fully connected layer.

[0104] Model Training

[0105] The cross entropy loss function L is used to train a deep neural network to predict the efficacy:

[0106]

[0107] Among them, y i is the actual label, p i is the probability value predicted by the model, and N is the number of samples. The model can automatically optimize and continuously improve the prediction performance. The weights are continuously updated through the back propagation algorithm to optimize the prediction performance of the model.

[0108] Prediction of therapeutic effect

[0109] New patient images are input into the trained deep learning model to predict their sensitivity to neoadjuvant chemoradiotherapy, and the output is complete remission (CR), partial remission (PR) or no effect (PD). The advantage of the deep learning model is that it can adaptively learn complex feature associations in imaging data.

[0110] Example 4: Integrated multi-model prediction

[0111] participate Figure 1 and Figure 2 , data collection and processing

[0112] Multimodal magnetic resonance images were acquired and conventional preprocessing (denoising, normalization, registration) and region of interest segmentation were performed.

[0113] Feature extraction

[0114] Morphological and functional features were extracted from T2WI, DWI and DCE-MRI, and multidimensional feature vectors were generated by combining clinical and pathological data.

[0115] Multi-model integration

[0116] Combining multiple prediction models, including SVM, random forest and CNN models, the prediction results of each model are integrated through a weighted voting mechanism, and the final prediction result is the weighted average of the prediction values ​​of each model: P final =αP SVM +βP RF +γP CNN

[0117] Among them, α, β, and γ are the weights of different models, and the weight ratio of each model is determined by model performance evaluation.

[0118] Prediction of therapeutic effect

[0119] The integrated model is used to predict the efficacy of patients, and the results are complete remission (CR), partial remission (PR) or no response (PD). This integrated strategy improves the accuracy and robustness of prediction by combining the advantages of different models.

[0120] Conclusion: The above specific embodiments describe in detail multiple application scenarios of the present invention, from traditional machine learning models (such as SVM, random forest) to deep learning models (such as CNN) and multi-model integration, showing how to use multimodal magnetic resonance images and clinical data to predict the efficacy of neoadjuvant chemoradiotherapy for rectal cancer. These embodiments provide clinicians with accurate efficacy prediction tools that can help formulate personalized treatment plans.

[0121] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in the field. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis. It is characterized in that The following steps are involved: Obtain multimodal magnetic resonance images of the patient, including T1-weighted images T1WI, T2-weighted images T2WI, diffusion-weighted imaging DWI, and dynamic contrast-enhanced magnetic resonance imaging DCE-MRI; Preprocess the acquired multimodal magnetic resonance images, including noise removal, image standardization, registration, and tumor area segmentation, and extract the region of interest (ROI); The morphological characteristics of the tumor were obtained by qualitative analysis; The diffusion characteristics and microcirculatory parameters of the tumor were obtained through quantitative analysis; Combine clinical and pathological indicators to construct a feature set; Use machine learning algorithms to establish efficacy prediction models, including feature extraction, selection, and model training; Multimodal MRI images and clinical and pathological data of new patients were input to predict their sensitivity to neoadjuvant chemoradiotherapy.

2. The method for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis according to claim 1, It is characterized in that The qualitative analysis includes the evaluation of the size, shape, boundary definition and signal intensity of the tumor, and the quantitative analysis includes the calculation of the apparent diffusion coefficient ADC value, blood flow BF, blood volume BV and vascular permeability PS of the tumor.

3. The method for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis according to claim 2, It is characterized in that The apparent diffusion coefficient ADC is calculated by the following formula: Wherein, S(b) is the signal intensity of the diffusion weighted image when the diffusion sensitivity factor is b, S(0) is the signal intensity when the b value is 0, and b is the diffusion sensitivity factor.

4. The method for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis according to claim 2, It is characterized in that The microcirculation parameters include blood flow BF, blood volume BV and vascular permeability PS, which are calculated by dynamic contrast-enhanced magnetic resonance imaging DCE-MRI. The calculation formula of microcirculation parameters is as follows: Among them, K trans is the vascular permeability coefficient, C t (t) is the contrast agent concentration in the tumor tissue at time t, C p (t) is the contrast agent concentration in plasma at time t.

5. The method for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis according to claim 1, It is characterized in that The clinical indicators include the patient's age, gender, tumor stage, and carcinoembryonic antigen (CEA) level; the pathological indicators include Ki-67 and p53. All characteristics are normalized by the following formula: Where x is the original eigenvalue, x′ is the normalized eigenvalue, min(x) and max(x) are the minimum and maximum values ​​of feature x, respectively.

6. The method for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis according to claim 1, It is characterized in that The feature set is constructed by using a feature fusion method to combine imaging, clinical and pathological data to generate a feature vector F: F=[F image ,F clinical ,F pathology ] Among them, F image is the imaging feature, F clinical For clinical features, F pathology is the pathological feature. The feature vector is reduced in dimension by principal component analysis (PCA), and the formula is: Z=XW Among them, Z is the feature matrix after dimensionality reduction, X is the feature matrix, and W is the dimensionality reduction matrix.

7. The method for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis according to claim 1, It is characterized in that The efficacy prediction model adopts a machine learning algorithm, support vector machine SVM, and its loss function is as follows: Among them, w is the weight vector, x i is the input feature, y i is the label, b is the bias, and C is the regularization parameter.

8. The method for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on magnetic resonance image analysis according to claim 1, It is characterized in that The model training adopts the random forest algorithm, and its integration formula is: in, is the final prediction result, M is the number of trees, f m (x) is the prediction result of the mth decision tree.

9. A system for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer based on the method according to any one of claims 1 to 8, include: Image acquisition module: used to obtain T1WI, T2WI, DWI and DCE-MRI images of patients; Image processing module: used for preprocessing the acquired magnetic resonance images and segmenting the tumor area; Feature extraction module: used to extract qualitative and quantitative image features; Data input module: used to input the patient's clinical and pathological data; Model training module: used to train efficacy prediction models based on imaging, clinical and pathological features; Prediction module: used to predict the efficacy of new patients.

10. The system for predicting the efficacy of neoadjuvant chemoradiotherapy for rectal cancer according to claim 9, in, The model can integrate imaging, clinical and pathological data and use neural networks for deep learning. Its loss function is cross entropy loss: Among them, y i is the actual label, p i is the probability value predicted by the model, and N is the number of samples.

Images