Linking AI Team Remote Cancer Radiotherapy Service Platform

Based on cloud computing and artificial intelligence technology, Linking builds an AI Team Remote Radiotherapy System, providing a broad range of patients with tumor target area mapping from well-known experts, dosage recommendation of radiotherapy, video group consultation, etc. This project has received support from a number of esteemed authorities such as the national key research and development plan, Beijing Collaborative Innovation, etc. Linking has developed a cloud SAAS-based product that utilizes advanced cloud platform processing and analysis technologies to deliver intelligent diagnosis and treatment tools to medical professionals. These technologies have enabled Linking to provide Internet + telemedicine services for tumor remote radiotherapy, staged diagnosis and treatment, as well as MDT comprehensive group consultation. Innovation Point 1: This project is the "Automatic Mapping Methods, Devices, and Storage Media for Endangered Organs based on Neural Networks" (Invention Patent No. 2018102392636). It has solved the efficiency and stability issues of endangered organ mapping for the first time, realizing the clinical application of artificial intelligence in automatic mapping. It adopts semantic segmentation and integrates various structures such as U-Net, Res-Net, Dilated Convolution, etc. to achieve the division of organs with high precision. It has developed an automatic part classification algorithm and a depth learning network-based segmentation algorithm to ensure stable automatic mapping of over 120 endangered organs in the head, neck, chest, abdomen, and pelvic cavity. This boosts efficiency by 80-90% and drastically reduces processing time from hours of manual work to just a few minutes. Innovation Point 2: This project is a "Method, Device, and Storage Media based on Deep Learning for Target Area Automatic Mapping" (Invention Patent No. 2018101344071). It has put into practice for the first time the automatic mapping of tumor target area based on artificial intelligence, making it possible to map the radiotherapy target area. The automatic mapping of the tumor target areas has always been a difficult in the industry. In the context of the automatic planning for radiotherapy applications, Linking has developed a multi-layer neural network structure with feedback and excitation mechanisms based on conventional depth. Leveraging the principles of radiotherapy mapping, as well as the constant rules and change characteristics inherent in treatment planning, the network incorporates a multiscale attention mechanism to regulate model parameters in the neural architecture. This approach achieves emphasis on reinforcement learning and enables automatic mapping of tumor target areas, even when working with small sample sizes. As a result, a successful convergence effect is achieved, leading to the accurate and efficient automatic mapping of tumor target areas. Innovation Point 3: This project is the "Methods and Systems for Artificial Intelligence-led Radiotherapy Program" (Invention Patent No. 2019108206503). This Program introduces the concept of "Automatic Angle" for the first time, solving the problem of excessive case samples of traditional predictive models, and forming the automation solution of the radiotherapy plan based on artificial intelligence. On the basis of automatically mapping the results of the endangered organs and the tumor target areas, the prescription is determined according to the disease information corresponding to the medical image, the geometric anatomical structure and a preset desease-prescription template library, the irradiation angle of radiotherapy is determined based on an "automatic angle" algorithm, and the irradiation angle is input into a dosage prediction model based on deep learning to obtain a radiation dosage distribution result. It takes the result of radiotherapy distribution as a reference, and adopts the reverse optimization algorithm based on the dosage distribution or DVH guidance to optimize and generate the executable radiotherapy plan. The innovation has been patented. The executable radiotherapy plan includes forward radiotherapy plan, stereoscopic radiotherapy plan, and intensity modulated radiotherapy plan, wherein the intensity modulated radiotherapy plan includes dynamic intensity modulated radiotherapy plan, static intensity modulated radiotherapy plan, volume intensity modulated radiotherapy plan, and rotation intensity modulated radiotherapy plan. This Algorithm has been integrated into Linking's Automatic Radiotherapy Program system, which has passed the testing of the Beijing Medical Device Inspection Institute, met the rules and regulations relating to radiotherapy such as YY/0775-2210, YY/T0889-2013, YY/0637-2009, and realized the outstanding performance recognized by the international community.