China Has Made Their Own Artificial Intelligence To Cure Cancer

The world in which we all live in has been suffering from cancer for a long time. So how can we reduce the mortality rate of lung cancer? The key is early diagnosis and standard treatment." Li Weimin, president of West China Hospital of Sichuan University, told the Economic Daily reporter.

To help humans fight lung cancer, artificial intelligence is now showing its talents. Huaxi Hospital in China recently announced with Etutech to jointly develop the world's first lung cancer clinical research intelligent disease database and the world's first multidisciplinary intelligent diagnosis system for lung cancer. It is reported that the clinical research intelligent disease bank of lung cancer contains multi-dimensional data such as image, pathology, genetic test and medical record text of 28,000 lung cancer patients admitted to Huaxi Hospital since 2009, and has been cleaned, analyzed and reconstructed by artificial intelligence technology. The multi-disciplinary intelligent diagnosis system for lung cancer relies on the disease database and comprehensive multi-disciplinary clinical information for comprehensive diagnosis, which can realize the functions of nodule screening and diagnosis and coverage of all types of lung cancer. As an artificial intelligence "unicorn" company, Etu Technology recently announced that it has just completed a $200 million C+ round of financing. Its computer vision artificial intelligence technology is currently used in three major areas of security, finance, and medical care.

So, what can artificial intelligence do to fight against lung cancer? What are the difficulties to solve?

Artificial intelligence can help doctors see more accurately. Li Weimin told reporters: "In the early screening of lung cancer, the first is to find nodules. Artificial intelligence can significantly improve the accuracy of chest CT examination. For example, for nodules below 5 mm, the detection rate can reach 98% or more." Ni Hao, president of Figure Medical, said that this part of the technology is quite mature. "Compared with the 4.3% missed detection rate of the human doctor team, the false detection rate of artificial intelligence is only 0.7%." However, a nodule is not necessarily cancer, but it must be able to recognize its nature. In judging the nature of nodules, the focus is on tracking changes in the lesion. "How big it was last year, how big it was three months ago, but this judgment is not easy. The current practice is to first establish a 3D model of the lung, and then use deep learning techniques to segment the location of the lesion, and finally calculate its volume, And judge the signs and their changes." Ni Hao said.

In contrast, artificial intelligence is also creating new wisdom. In the traditional clinical research process, it is usually the doctors who first suspects that certain factors may be related to a certain problem, and then obtains data support through clinical experiments or patient medical records. By combining and integrating massive patient information, including multi-dimensional data such as electronic medical records, genes, and pathological tests, artificial intelligence can provide reference diagnosis and recommended treatment plans to doctors on each hand. On the other hand, artificial intelligence can also be used for auxiliary modeling which empowers clinical research to find the latest factors relevant to treatment.

"From these aspects, artificial intelligence can be replicated for the diagnosis. We also plan to expand the auxiliary diagnosis of artificial intelligence to 10 diseases such as liver cancer, breast cancer, stomach cancer and intestinal cancer. Ni Hao said.

However, artificial intelligence also has to "turn over the mountains and mountains." The first is the issue of data structuring and standardization. Just as patients complain that they don't understand the doctor's handwritten medical records, artificial intelligence also faces this problem. "The biggest difficulty is that many medical data structures are arbitrary and widely different. The doctors write different medical records in different formats and different ways of description. This requires the semantic understanding of artificial intelligence and also refers to the clinical guidelines for lung cancer according to the domestic and international guidelines for lung cancer. Studying the standardized fields and expanding the model of clinical information extraction is more complicated in terms of image data. We extract the model through the image structure of artificial intelligence deep neural network, including the type and size of the lesion," Li Weimin said.

The second is to solve the opacity problem of artificial intelligence. Influenced by the algorithm of deep learning, although artificial intelligence can make conclusions, it is difficult for human beings to understand its decision-making reasons. So how does it get the trust of doctors? “We have included some interpretable technical analysis indicators at the product level to provide evidence to doctors and tell them why they are so judged. For example, artificial intelligence tells doctors to observe signs such as lobulation and vacuoles. Therefore, it is judged to be malignant." Ni Hao told reporters.

In addition, artificial intelligence in the application scene "landing" always faces commercialization problems. Ni Hao said: "The 'itch point' is not commercialized, and the 'pain point' has commercial prospects. With the popularity of lung cancer screening in China, the number of screenings will definitely increase tremendously, and the hospital manpower is unbearable. 'Pain points' will inevitably require the help of artificial intelligence."

"The involvement of lung cancer screening and diagnosis means that artificial intelligence is moving from research to clinical practice." According to Jin Xiaotao, president of the China Health Information and Health Medical Big Data Association, the participation in artificial intelligence also helps to break the distribution of medical resources. Inequality is a “bottleneck”: “Big hospitals are overcrowded and basic medical institutions are difficult to improve their treatment capabilities. Artificial intelligence technology is used as a tool to break the information barrier and empower the grassroots of hospitals doctors more accurately. The diagnosis and treatment are an important means to solve the shortage of medical resources in our country."

If you really love my effort, PLEASE DO:

Yes, your every action motivate me to write more on Steemit. Without you, it would never be possible for me to reach here.