The First In The World, Artificial Intelligence Platform For Early Diagnosis Of Colon Cancer And Peritoneal Metastasis!
Peritoneal metastasis is generally considered to be the terminal stage of bowel cancer, with a poor prognosis. At present, the diagnosis of peritoneal metastasis of colorectal cancer is mainly through imaging methods, and the sensitivity is lacking, especially for small peritoneal metastases below 5 mm.
Recently, the colorectal surgery research team of the Sixth Affiliated Hospital of Sun Yat-sen University and Shenzhen Tencent AIlab have cooperated and successfully developed the world's first AI platform for the diagnosis of colon cancer peritoneal metastasis, which can automatically identify the characteristics of the primary tumor. The imaging features of the tumor adjacent to the peritoneum were extracted, and an artificial intelligence-based SVM classifier was constructed. The AI model automatically recognizes and diagnoses all verified images in only 34 seconds. The accuracy is as high as 94%, the AUC is 0.922, and the sensitivity and specificity are as high as 94%.
This original research result was published in Annals of Surgery under the title of "Using Deep Learning to Build an Artificial Intelligence System to Diagnose Colon Cancer and Peritoneal Metastasis". Dr. Yuan Zixu of the hospital was the first author, Professor Wang Hui was the last corresponding author, and Dr. Cai Jian, the deputy chief physician, Dr. Cao Wuteng of the imaging department, and Dr. Zhao Yebiao made important contributions to the paper.
It is understood that Annals of Surgery, the top journal in the field of surgery, was published as early as 1885. It published many surgical "milestone" papers. It is a benchmark in the field of surgery and leads the development direction of international surgery. The current impact factor is 10.13. Minute.
The world's first AI platform for diagnosing peritoneal metastasis of bowel cancer! The future is expected to extend the survival time of patients with bowel cancer
Artificial Intelligence (AI) is a new type of intelligent technology science that simulates the learning of the human brain and extends human capabilities. In recent years, AI has been widely used in the medical field, especially in tumor diagnosis. AI is good at automatic medical images (imaging and pathology) For identification and diagnosis, the deep learning algorithm after the AI update has more advantages, which greatly improves the sensitivity and accuracy of AI diagnosis.
For a long time, peritoneal metastasis is considered to be the terminal stage of bowel cancer, and the prognosis is very poor. However, the current clinical diagnosis of colon cancer peritoneal metastasis is mainly through imaging methods, and there is a lack of sensitivity, especially for small peritoneal metastatic lesions below 5 mm. Therefore, the research group of Professor Wang Hui of the hospital has paid unanimous attention to how to diagnose colon cancer peritoneal metastasis early.
The incidence of colon cancer with simultaneous peritoneal metastasis (PC) is about 5-10%, and the incidence of recurrence with peritoneal metastasis is 25-44%. "If peritoneal metastasis can be diagnosed early, it can increase the chance of complete cytoreductive surgery, and it can significantly extend the survival time of patients with bowel cancer in the future." Professor Wang Hui said. Since 2018, the team has established a cooperative relationship with Shenzhen Tencent AI lab and developed a ResNet3D system based on Convolutional Neural Network (CNN). After investigation, this is the world's first AI platform for diagnosing colon cancer peritoneal metastasis. It can automatically identify the characteristics of the primary tumor, and at the same time extract the imaging features of the tumor adjacent to the peritoneum, and build an artificial intelligence-based SVM classifier. The training group included 19,814 CT images, and the validation group included 7,837 CT images.
The study found that ResNet3D's AI system automatically identified and diagnosed all verification images in only 34 seconds. The accuracy of the "ResNet3D+SVM classifier" in the diagnosis of colon cancer peritoneal metastasis is as high as 94%, the AUC is 0.922, and the sensitivity and specificity are as high as 94%, which is significantly better than the diagnostic ability of conventional enhanced CT.
What is the clinical value of this achievement? Yuan Zixu said, “The AI platform we developed is a new non-invasive diagnostic system based on the enhanced CT images that are routinely used in abdominal tumors. It can not only automatically identify the characteristics of the primary tumor, but also integrate the characteristics of the surrounding peritoneum. It is clinically practical. It is very strong. It provides references for clinicians to formulate surgical plans, and also provides a basis for colorectal cancer patients to choose appropriate treatment." According to reports, the AI platform can recognize imaging images from other hospitals or centers, so the next step is to use the AI system. Transplant to other hospitals, use a larger independent cohort, conduct external verification to prove its universal applicability, and strive to solve the worldwide problem of difficult diagnosis of colon cancer and peritoneal metastasis.
