Researchers use AI to improve the discrimination between benign and malignant thyroid nodules
Science and Technology Daily, Beijing, May 14th (Reporter Liu Yuanyuan) A reporter learned from West Lake University on the 14th that the school ’s research team used artificial intelligence technology in the proteome big data of nearly a thousand patients with thyroid nodules and found that it helps to distinguish thyroid nodules Combination of benign and malignant protein molecular markers. This series of markers is expected to greatly improve the accuracy of the judgment of benign and malignant thyroid nodules.
The researchers said that thyroid nodules, or thyroid tumors, can be caused by a variety of factors and are more common in contemporary populations. Like most nodules, thyroid nodules also have a difference between benign and malignant. Benign thyroid nodules will not affect daily work and life, and malignant thyroid nodules need to be treated as soon as possible. However, in clinical practice, about 30% of thyroid nodules lack effective benign and malignant judgment methods.
In order to solve this problem, Guo Tiannan Laboratory of School of Life Science of West Lake University and Li Ziqing Laboratory of Engineering College have joined forces and cooperated with multiple clinical teams at home and abroad.
In this study, the experimental team analyzed tissue samples from 911 thyroid nodule carriers, performed data-independent proteomics analysis, and generated 2,421 proteomics data.
Due to the huge amount of proteome data involved in the experiment and the slight difference in molecular level of some thyroid nodules, the research team used artificial neural network technology for screening. They found 14 key protein combinations that distinguished between benign and malignant nodules, and these combinations constituted a model that can determine benign and malignant.
Subsequently, the research team used the model to predict unknown benign and malignant thyroid nodules, and then compared with the pathological results after clinical surgery. The results showed that the accuracy of this method reached 90% in judging the benign and malignant of 288 thyroid paraffin samples and 64 thyroid nodule puncture samples provided by four hospitals in China.
It is understood that the method is currently being tested in more clinical centers to further optimize the artificial intelligence model, and has applied for a patent.