Yang’s article published in Nature Machine Intelligence

Lin Yang, Ph.D.

Dr. Lin Yang, associate professor in the J. Crayton Pruitt Family Department of Biomedical Engineering, and researchers’ article, “Pathologist-level Interpretable Whole-slide Cancer Diagnosis with Deep Learning,” was recently published in Nature Machine Intelligence.

Most carcinoma identification requires microscopy-level image assessment for early tumor discovery and for developing therapies based on diagnostic pathology. Diagnosing pathology slides is a complicated task that requires years of pathologist training. Unlike other medical image types (for example, radiology images), digital pathology slides are obtained at very high resolution (more than ten gigapixels).

However, region- and cellular-level features, which have significant implications for diagnosis and treatment, can be subtle and confusing. Facing the enormous amounts of information that exists in huge slides and the heavy workload in clinics, even experienced pathologists are prone to make errors.

Qualified cancer diagnosis requires peer review and consensus, a standard that can be expensive to satisfy in hospitals and small cancer centers where experienced pathologists are scarce.

The team of researchers shows that artificial intelligence (AI) has the potential to assure trust in computer-aided diagnosis (CAD) in diagnostic pathology, to offer reliable diagnosis, objective second opinions, strong generalizability and cost-effective deployment, all of which have the potential to greatly improve the routine pathology experience.

The research team seeks to alleviate the above-demonstrated problems in diagnosis and will inspire the emergence of new AI-based CAD systems for various types of cancer. The first author, Dr. Zizhao Zhang, was a Ph.D student in Dr. Lin Yang’s lab and now continues his research at Google, LLC.

The paper provides an innovative and reliable means for making diagnostic suggestions and can be deployed at low cost as next-generation, artificial intelligence-enhanced CAD technology for use in diagnostic pathology.