The reporter learned on January 7 that ByteDance's medical brand "Xiaohe Health" has developed a colonoscopic AI auxiliary diagnosis software. It can use artificial intelligence technology to assist clinicians in finding and identifying colorectal cancer lesions in real time. At present, the sensitivity and specificity of the software to the standard diagnosis of colorectal cancer (two key indicators reflecting the level of medical screening) have been increased to more than 95%.
It is understood that colonoscopic AI-assisted diagnosis software is the first medical research project of Xiaohe Health. During the research and development process, a number of domestic top three hospitals provided support to Xiaohe Health, authorizing it to analyze more than 500,000 medical cases and more than 14 million medical pictures.
The project researchers said that based on ByteDance's technical advantages in the field of machine learning and image recognition, the colonoscopic AI software developed this time has strong deep learning capabilities, and the processing capacity and processing speed of colorectal cancer medical images and data have been greatly improved.
According to the data, ByteDance has applied for 15 invention patents for related AI technologies, mainly focusing on polyp recognition, polyp typing, endoscopic image analysis and processing and other fields. After the combination of these technologies, this AI colonoscopy software can achieve self-supervised training without adding additional data, and continuously improve its learning ability and clinical accuracy. When processing colonoscopy videos in real time, it can analyze 30 to 50 medical images per second and give real-time feedback and diagnostic prompts to doctors during the examination.
Compared with other cancers, colorectal cancer is relatively preventable and controllable, and is listed as the "most likely cancer to be destroyed" alongside cervical cancer. The key to preventability and controllability lies in early prevention, early screening, early treatment, colonoscopy is one of the most important means of early screening.
The Health Times has reported that China needs 350,000 gastrointestinal endoscopists, but the actual number is less than 40,000, and even fewer skilled operators. A gastroenterologist from novice to skilled, more than 10 years of training and experience. Therefore, the importance of AI medical technology in assisting clinicians and enhancing clinical capabilities has become more prominent.
The relevant person in charge of Xiaohe Health revealed that the reason for the development of colonoscopic AI instruments and supporting software is to help solve the shortage of gastrointestinal endoscopists in China and the high rate of missed diagnosis of traditional colonoscopy. Xiaohe Health conducted internal research in the early days, focusing on byteDance's technology patents. "After research and judgment, we finally chose colonoscopic AI technology as the first scientific research project. Through this technology, we can apply our own technical advantages to medical scenarios. The person in charge said.
In order to understand the actual needs of front-line doctors, before product development, all R&D engineers interned in cooperative hospitals, observing the operation and examination process every day, and watching a large number of intestinal videos. According to the practice, the engineers decided the research and development direction of colonoscopic AI technology: in addition to assisting doctors in finding and identifying colorectal cancer lesions, it also has functions such as intestinal cleanliness assessment and automatic switching of gastroscopic mode. In addition, colonoscopy will have a special reminder when entering and exiting the patient's intestinal lumen to help the doctor determine whether the time is in line with the standard requirements; in order to better assist the novice doctor, the software has also developed the auxiliary function of intestinal examination.
Pictured: Xiaohe Health Colonoscopy AI software analyzes and identifies colorectal polyposis lesions
"Assistive functions like intestinal examinations can assist doctors in predicting the direction of the colonoscope when the intestinal lumen is not visible. This helps a lot for some novice doctors. The above-mentioned person in charge told reporters that many doctors have mentioned during the investigation that improper colonoscopic operations may cause pain, and even damage the intestinal mucosa and cause perforation. This has also become one of the problems that this AI colonoscopy software focuses on solving.
Wu Xiaojian, vice president of the Sixth Affiliated Hospital of Sun Yat-sen University, believes that the combination of artificial intelligence and colonoscopy is one of the current trends, and the colonoscopic AI technology of Xiaohe Health has outstanding performance in sensitivity and specificity, two key indicators that reflect the level of medical screening. If related technologies are put into clinical practice on a large scale, they will be able to assist front-line doctors in a large number of front-line doctors to improve the diagnosis efficiency of colorectal cancer, colorectal polyps and other lesions, and reduce the rate of missed diagnoses.
The relevant person in charge of Xiaohe Health said that Xiaohe Health will take malignant tumors as a key project for continuous scientific research. In the future, ByteDance will also invest more resources in technology research and development, combining medical needs with its own technical advantages to benefit more ordinary patients.
![Colonoscopy without anesthesia or pain? Provincial women and children carry out a new method of water injection colonoscopy[fig]](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACwAAAAAAQABAAACAkQBADs=)