Gastric cancer TNM staging prediction system based on multi-modal deep learning

Abstract

The invention discloses a gastric cancer TNM staging prediction system based on multi-modal deep learning, which comprises a data acquisition module, an image preprocessing module, a text preprocessing module, a prediction model training module and a prediction module, and the application of the modules comprises the following steps: acquisition and data cleaning of CT image and text data; preprocessing the CT image, extracting mark data, establishing a mask image, and multiplying the mask image with an original image to obtain an effective identification area; performing pre-training on the text data by utilizing XGBoost and LightGBM algorithms in machine learning to obtain a feature importance sequence so as to obtain an effective data sample; and deep learning is performed on the deep and shallow level features of the image and text data, and training is performed by using a convolutional neural network to obtain a classification result of a correct TNM period. Compared with traditional artificial identification, the method has more modal reference bases, has very high identification accuracy, can reduce the business burden of professional doctors, and solves the current situation of unbalanced medical resources in China.

Description

technical field [0001] The invention belongs to the field of medical artificial intelligence, and in particular relates to a gastric cancer TNM staging prediction system based on multimodal deep learning. Background technique [0002] Gastric cancer (GC) has a high incidence in East Asia and remains the fifth most common malignancy and the third leading cause of cancer death worldwide. Accurate tumor local depth invasion (T), local lymph node invasion (N) and distant metastasis (M) are the key to the treatment strategy. In clinical practice, computed tomography (CT) has been widely used in preoperative staging. However, the accuracy of preprocessing staging is not satisfactory. Currently, CT images of gastric cancer are manually evaluated by radiologists. Accurate TNM staging of gastric cancer is an important and challenging task for radiologists. In a recent study, the accuracy of CT for T and N staging was 60%-78% and 56%-75.9%, respectively. The poor resolution of CT...

AI Technical Summary

Problems solved by technology

[0006] However, in reality, models based on a single modality (e.g., text, image) may not be sufficient for the task

Images