Skip to Main Content

An official website of the United States government

Principal Investigator
Name
Biyue Zhu
Degrees
Ph.D.
Institution
Children's Hospital of Chongqing Medical University, China
Position Title
Associate Professor
Email
About this CDAS Project
Study
PLCO (Learn more about this study)
Project ID
PLCOI-1341
Initial CDAS Request Approval
Sep 28, 2023
Title
Precision Diagnosis of Cancers Based on Artificial Intelligence Pathology Analysis
Summary
Cancer is one of the main causes of death. Due to the high risk and high incidence of the disease, there is an urgent need for rapid, accurate, and cheap diagnosis of cancer. Pathological diagnosis is one of the important methods of clinical diagnosis and provides crucial information for prognosis and treatment. Traditionally, pathological diagnosis requires a pathologist to observe glass slides and give suggestions. Two decades ago, whole slide imaging (WSI) technology was developed. WSI scans the slide under a slide scanner microscope and generates a digital image of the whole field of view of a pathology slide, which can permanently store all the tissue information on the tissue slide. Compared with traditional glass slides, WSI has many advantages in slide preservation and management, such as teaching, remote diagnosis, image repeatability, etc., and effectively solves the problems of traditional glass slides that could be easily damaged, faded, lost, and difficult to retrieve. Also, WSI technology facilitates the easy collaboration of a multidisciplinary team at any time, place, and professionals to give quantitative analysis and annotation of the slides, and finally give an accurate and timely diagnosis for patients.

Artificial intelligence (AI) is expected to play a vital role in the new era of precision pathology diagnosis. Combining cutting-edge AI and WSI technology can integrate patient clinical information, multi-omics data, and tissue pathology images to facilitate precision diagnosis. Several studies have demonstrated the power of AI by developing an automated histopathological evaluation system for tumors. This tool can improve the accuracy of adult tumor pathology and clinical diagnosis, save time and cost, and also open up a new way to predict patient prognosis and molecular changes. Despite the rapid development of AI and WSI, powerful AI models are still in high demand.

Therefore, to improve the effectiveness and accuracy of clinical diagnosis, facilitate disease prediction and timely treatment of tumors, and save time and costs for the family, it is imperative to develop powerful AI tools for various cancer pathological diagnoses. To develop a smart AI model, large numbers of pathology images are highly required for model training.

This project will first collect the WSI images and clinical information of cancer patients, and use these data to train and develop an automated AI system. We envision that this new technology can be a versatile tool to improve the accuracy of clinical diagnosis, facilitate disease prediction and timely treatment of tumors, and save time and costs for the family. This tool can also be used to assist in the training of cancer pathologists. In addition, our results may facilitate the cancer etiology elucidation and drug discovery, and potentially improve the quality of life and survival rate of cancer patients.
Aims

This project will develop a new artificial intelligence (AI) system for the automated diagnosis of tumors through pathological images recorded by whole slide imaging (WSI) technology. We envision that this new technology can improve the accuracy of clinical diagnosis, facilitate disease prediction and timely treatment of cancers, and save time and costs for the family. Also, we hope this technology can help improve the quality of life and survival rate of cancer patients.
To achieve our goals, the following steps will be conducted:
①Image acquisition: collect whole slide images and clinical information on cancer patients.
②Preprocessing: converting the digital images into a format suitable for analysis, including normalization, filtering, and segmentation.
③Feature extraction: identifying and extracting relevant features from the digital images, such as morphological, textural, and intensity-based features.
④Phenotypic characterization: analyzing the extracted features to characterize the phenotypic characteristics of the tissue samples, including cellularity, nuclear morphology, and tissue architecture.
⑤Association analysis: investigating the association between the phenotypic characteristics and genetic variants, if applicable. This may involve techniques such as genome-wide association studies (GWAS) or machine learning approaches.
⑥Validation: validating the findings through independent cohorts or experimental evidence.

Collaborators

Dr. Xiyue Wang, a postdoc fellow recenly joined in the Department of Radiation Oncology, Stanford University School of Medicine, Stanford, CA 94305 USA (e-mail: xiyue.wang.scu@gmail.com)