Westlake News WETALK

Lin Yang: Let AI Learn To Recognize Cancer Cells

20, 2022

Email: zhangchi@westlake.edu.cn
Phone: +86-(0)571-86886861
Office of Public Affairs

The Connecting Great Minds series of talks aims to share the latest achievements in basic scientific research and original technological innovation in the frontier fields of science with our Westlake community.

On April 16, five scholars from Westlake University shared their thoughts during the sixth session of this event. Let’s take a look at the research of Prof. Lin Yang of the Artificial Intelligence and Biomedical Imaging Laboratory with the School of Engineering at Westlake University.

The most beautiful words in the world are not "I love you", but "your tumor is benign".

—Woody Allen

A worrying situation is when doctors need to determine within minutes whether there is a problem with a pathological slide that shows millions of cells. Instantly discovering the presence or absence of cancer cells in millions of cells depends on the experience and ability of the individual doctor.

It is good for patients to hear "your tumor is benign", but there remains the risk that the doctor has made a mistake. This can have dire consequences.

It may not be easy for the human eye to tirelessly identify problems from millions of cells; but if these eyes can be escorted by artificial intelligence, it may not be a problem.

Yang is solving the problem of cancer screening through research on artificial intelligence and digital medicine.

Research in this field is divided into three areas—textual, pathological, and genetic and molecular—corresponding to three important directions for AI and digital medicine. The first direction is in speech and text, which has been widely used in major hospitals, and gradually builds standardized clinical pathology reports to form structured data. The second direction is pathological diagnosis, which will become the basis for precise treatment. The third direction is genetic testing, including proteomics and single-cell sequencing, such as the study of extracting genetic information from single cancer cells.

Yang hopes to use artificial intelligence to analyze this information to provide more assistance for doctors to make accurate diagnosis and treatment. In the past 20 years of his research career, he has focused on AI-assisted pathological diagnosis and treatment.

Pathology is called the foundation of medicine, and pathologists are called the doctors of doctors. Under the current medical conditions, many tumor patients need pathological tests to determine whether a tumor is malignant or benign, and decide the treatment plan. However, even if the lesion tissue is taken from the same person, each slice is different, which requires the pathologist to be cautious in the diagnosis and judgment to ensure accuracy. The reality is that there is a shortage of 90,000 licensed pathologists nationwide with nearly 5 million new cancer patients each year.

The application of artificial intelligence can fill the gap to a certain extent, and the learning ability of machines has already made it possible to give accurate pathological judgments. Take cervical cancer screening as an example. Every cervical slice needs to be observed by a doctor cell by cell, and if problems are found, re-examination is carried out. If the slices are digitized, artificial intelligence after deep learning can quickly detect each cell, spot problem cells and conduct in-depth analysis. Since his Ph.D. study in 2004, Yang and his research team have been committed to exploring and developing an interpretable general artificial intelligence model, adding wisdom to the whole process of pathological diagnosis. Recently, Yang’s team research on explainable artificial intelligence deep-learning networks was also published in the Nature Machine Learning.

Yang said, "In cancer screening, the real problem we face is that there are simply not enough qualified doctors, and the current cytology doctors are overloaded most of the time, and this situation cannot be changed in the short term. Only AI can significantly reduce medical insurance expenditures while also significantly improving the quality of cervical cancer screening.” Yang is very confident that this can be achieved, "I hope to use artificial intelligence combined with digital pathology to decode the secrets behind cells, and truly contribute to this area.”



Lin Yang: Let AI Learn To Recognize Cancer Cells