Breast cancer cure is not a dream new application of artificial intelligence technology

"Artificial intelligence + medical" is a blue ocean field to be developed. A few days ago, "Tencent Miying" officially released an AI system for breast tumor screening, announcing the first use of artificial intelligence in the country to distinguish benign and malignant breast tumors, and automatically generate breast imaging reports And data system (BI-RADS) classification report. It is reported that "Miying" has entered clinical trials in more than 10 tertiary hospitals across the country, and it is expected to play a greater role in early screening, diagnosis and treatment of breast tumors in the future.

Early screening of breast cancer at younger age is the key

"Due to the preference for high-fat and oily eating habits, the increased pressure of life and work, and the increase in active and passive smoking, the incidence of breast cancer in China has been among the highest in the world in recent years, especially in cities such as Beijing, Shanghai, and Guangzhou. The high level of incidence in developed countries in Europe and the United States." Professor Lu Weiming, director of the Thyroid and Breast Surgery Department of the First Affiliated Hospital of Sun Yat-sen University, pointed out that breast cancer patients in China have the characteristics of early onset of onset and late treatment. The average age of onset of breast cancer in women is 10 younger than that of European and American countries. The peak age of onset is between 45 and 55 years old. Due to insufficient attention and lack of effective screening, the domestic breast cancer stage is late at the first diagnosis.

Lu Weiming said that mammography of mammography is a common screening method for breast tumors at home and abroad. Doctors need to spend a lot of energy to compare and diagnose multiple images. When the breast glands are dense and abundant, they can accurately distinguish between glands and Lesion is no small challenge. The emergence of artificial intelligence medical image analysis technology has brought breakthrough ideas for mammography.

It is reported that Tencent and the First Affiliated Hospital of Sun Yat-sen University, Sun Yat-sen Memorial Hospital of Sun Yat-sen University, Guangdong Provincial Hospital of Traditional Chinese Medicine, Guangdong Maternity and Child Health Hospital, and Guangdong Second People’s Hospital jointly developed the "Miying". The Second People's Hospital and more than 10 other tertiary hospitals across the country have entered clinical pre-trial trials. Experts say that it will help clinicians find breast tumors more accurately and efficiently, improve the diagnosis of early breast cancer in primary hospitals, and help patients obtain medical services for early screening, diagnosis and treatment.

Difficult cases + expert argumentation to improve the efficiency of "watching movies"

It is reported that the "Miying" breast tumor screening AI system uses a large number of multi-center data learning and expert argumentation and annotation to enhance the AI ​​model's ability to distinguish difficult cases.

Compared with traditional naked eye screening, this system has achieved two major breakthroughs: First, the traditional mammography examination requires the doctor to observe the patient’s breast image, and then diagnose the condition based on his own experience. “Miying” can automatically identify and locate suspicious lesions. , Mark the location of the tumor and calcification, to help the doctor quickly find the doubt. Secondly, it can further determine the degree of benign and malignant risk of tumors, and automatically generate breast imaging reports and data system (BI-RADS) classification reports. AI technology and doctors cooperate with each other to greatly improve the efficiency of doctors' "viewing films" and reduce the occurrence of "missed diagnosis".

Based on large-sample, non-homogeneous, multi-center test statistics, the "Miying" breast tumor screening AI system has a sensitivity of 99% and 90.2% for detecting breast calcification and malignant masses, respectively, and is sensitive to the diagnosis of benign and malignant breast tumors. The degree and specificity were 87% and 96%, respectively.