Since AlphaGo defeated the human player Lee Sedol in Go last year, artificial intelligence (AI) has spread throughout the streets. Since then, AI has achieved various results in various industries and continues to heat up. Medical care is one of the main application areas of arti

Since AlphaGo defeated the human player Lee Sedol in the field of Go last year, artificial intelligence (AI) has spread throughout the streets. Since then, as AI has achieved various results in various industries, it has continued to heat up. Medical treatment is artificial intelligence. One of the main application areas.

It is reported that Airdoc has achieved results similar to those of human doctors in terms of specificity and sensitivity in the fields of diabetic retinopathy, strabismus and orbital disease, and has begun to be used in clinical applications. Has the era of medical artificial intelligence arrived?

Application of AI in the medical field

Artificial intelligence is the use of machines to simulate the working methods of human brain neurons, so as to think and master the intelligence of learning like humans. Compared with humans, artificial intelligence has the characteristics of fast speed, a large number of fortunes, and no fatigue, and can provide help to humans in many fields.

Unlike many industries, doctors are an industry that requires a lot of experience. They need to master the medical knowledge accumulated by countless predecessors and verify theories and accumulate experience in clinical work. AI based on deep learning is the same as doctors from some perspectives. It requires a lot of learning to master, and continuous improvement of algorithm models in practice.

There are two main directions for artificial intelligence in the medical field. One is based on natural language processing, which diagnoses diseases based on medical records and symptoms; the other is based on computer vision , which diagnoses diseases by identifying medical images.

Take lung cancer recognition as an example. AI can diagnose lung cancer in two ways. One is based on natural language processing. The representative company is Watson of IBM. In 4 years, Watson studied 200 textbooks in the field of oncology and 290 After reading several medical journals and more than 15 million documents, Watson began to be used clinically. By inputting patient medical record information and symptoms into the system, it can identify lung cancer. Today, Watson can already provide doctors with diagnostic suggestions in the fields of lung cancer, breast cancer, rectal cancer, colon cancer, gastric cancer, and cervical cancer.

Another diagnostic method is based on computer vision. The representative company is the domestic Airdoc. Since lung cancer does not have any clinical symptoms in the early stage, if there are symptoms, it is in the middle and late stages. Currently, about 75% of lung cancer patients are in the late stage when diagnosed. lung Effective diagnosis and timely treatment of nodules are crucial. The algorithm developed by Airdoc based on deep learning can quickly detect pulmonary nodules through pulmonary nodule location, nodule size judgment, and calculation of lung nodule malignancy indicators.

Advantages of AI in medical imaging

Although Watson can accurately identify many cancers, the initial investment was equally huge. With the efforts of many experts, it still took 4 years and a huge amount of money to achieve the current results. But the field of computer vision is different. In 2012, Professor Hinton adopted a new algorithm called "Dropout" to avoid overfitting. This was a revolutionary breakthrough that made computer vision more convenient and accurate. Artificial intelligence computer vision based on deep learning has entered a stage of rapid development and is now able to contribute to various fields.

Today, AI+ medical imaging has left the laboratory, and the next step will be a wave of commercialization. For example, Enlitic collects X-ray , MRI, CT scan, 3D medical and other imaging data to share or view medical images, thereby increasing the speed of discovering tumors and other diseases.

Foreign research on AI has exceeded 60 years, but the domestic AI field is catching up. The total number of Chinese AI articles accounts for more than half of the world's, with 42.8% of citations, and there are many Chinese who influence the world in the field of deep learning.

As early as 2012, Airdoc members began to conduct research on artificial intelligence. Airdoc spent a lot of time collecting hundreds of thousands of fundus photos from many top domestic and foreign hospitals, and finally developed the Airdoc diabetic retinopathy auxiliary diagnosis model. , and achieved results comparable to those of human doctors in terms of main indicators such as sensitivity and specificity.

Future diagnosis and treatment model

The development of modern medicine depends on three changes in the last century. In 1935, sulfonamide was confirmed to have a bactericidal effect. In the 1940s, the artificial synthesis of sulfonamide drugs was realized, which promoted the rapid development of pharmaceutical and chemical technology. This is The first revolution; the second revolution is through the latest radiological diagnostic technology, which can detect early tumors and many other early lesions; the third revolution is the production of biological products, which has opened up a new concept of biological treatment of diseases.

With the development of medical informatization in recent years, "medical diagnostic innovation", "smart medical care" and "health big data" have become frequently appearing words. Internet medical care has benefited tens of thousands of patients. In order to promote the health industry, the country has launched the " Healthy China 2030 Planning Outline ". This is the best era for medical care, and this is the best era for AI medical care.

Although today's AI is still in its infancy and does not have further interaction and communication capabilities, it has achieved great results in the fields of preliminary diagnosis and disease warning and prediction. This year's national " Two Sessions " government work report It was mentioned that various forms of medical consortium construction pilots will be fully launched to enhance grassroots service capabilities and facilitate people to seek medical treatment nearby. The diagnosis and treatment model of AI+ doctors can assist grassroots medical institutions to meet this challenge.