Dr. Yao Lu gave the speech in the “Artificial Intelligence and Internet of Things” sub-forum.
Dr. Yao Lu introduced that medical imaging has become an important basis for clinical diagnosis, even in traditional Chinese medicine. With the separation of medical services and drugs, the hospital's revenue from drugs will drop sharply. However, the number of medical images is increasing by 40% per year while the number of doctors is too few to meet the clinical needs, so that the misdiagnosis and missed diagnosis are inevitable.
The artificial intelligence technology can improve the efficiency of doctors and effectively resolve the contradiction between the explosive imaging growth and the serious shortage of doctors. Medical AI is developed from computer-aided diagnosis (CAD). The goal of medical AI is to let the machine learn the doctor's experience and make logical judgment. The existing deep learning algorithms are usually dependent on data but ignoring the doctor's experience. According to this situation, we proposed a world-leading knowledge-based deep learning algorithm on small sample by combining knowledge-based machine learning algorithms and data-based deep learning algorithms. By now, our algorithm can be well converged and generalized on the data sets of 1000-2000 high-quality samples, which is difficult to achieve by the traditional deep learning algorithm.
Dr. Yao Lu took the breast cancer as an example to show how artificial intelligence assist doctors. The number of breast cancer patients in cities is growing at a rate of 14% per year in China, and the patients is younger. If the breast cancer is detected at an early stage, the 5-year survival rate of the patient is very high. But unfortunately, the detection rate of early breast cancer is still low in China, so the medical imaging-assisted early screening is very important.
At present, the data of mammography is about 300-400 layers. The contrast of the lesion is low, and there are large differences between individuals. It is very difficult for doctors to make accurate judgments in a short time. Our product is aimed to provide real-time detection of lesions and quantitative analysis based on artificial intelligence algorithms, which can meet the clinical needs well. In addition to assisting doctors in clinical diagnosis, it is important for early screening to provide breast cancer risk assessment by using a quantitative analysis model. We use the technology of intelligent quantitative analysis to establish a personalized breast cancer risk prediction model. The model evaluation results are highly consistent with the experience of senior doctors. Knowledge models and intelligent quantitative analysis are both the key technology of our advanced machine learning algorithms.
The key point of medical artificial intelligence is medical care. It is not so easy to meet the clinical requirements, for example it usually takes dozens of communication on a single feature with doctors. But we will polish our product based on clinical scenarios to make sure that it can assist doctors in making more accurate diagnosis and treatment decisions. We hope that all doctors will be more confident in clinical diagnosis and all patients will get better treatment with more dignity.
The future of medical artificial intelligence will be in the area of multi-omics and precision medicine. PVmed has always been a vision of “knowing the imaging, and knowing the doctor”. With combination of artificial intelligence and medical knowledge, we hope that a lot of simple repetitive work could be reduced, and our products could assist doctors to improve diagnostic efficiency and resolve more clinical pain points.
