Ron Summers

NIH Clinical Center

Dr. Summers is currently Chief of the Imaging Biomarkers and Computer-Aided Diagnosis (CAD) Laboratory at the NIH Clinical Center. His research interests include CT colonography, spine imaging, deep learning, CAD and development of large radiologic image databases. Dr. Summers clinical areas of specialty are gastrointestinal and genitourinary radiology.

His group were one of the pioneers in deep learning in radiology image analysis, publishing important and highly cited early works in MICCAI, CVPR, and IEEE TMI. Their research has shown that deep learning can provide dramatic improvements in sensitivity for automated detection of various diseases including pre-cancerous polyps in the colon, osseous metastases in the spine and enlarged lymph nodes in the mediastinum and retroperitoneum.

Dr. Summers group is an international leader in automated detection of spine disease on CT scans. His group has developed fully automated software that partitions and labels the spine, measures bone mineral density, and detects traumatic and compression fractures, degenerative disease and epidural masses. For example, the automated bone mineral densitometry permits opportunistic screening for osteoporosis on routine body CT scans obtained for other purposes.

The long-term vision of his group is to develop software that can accurately and fully automatically detect all diseases on abdominal CT scans. To achieve this bold vision, Dr. Summers has enumerated all of the radiographically detectable abnormalities on CT scans and have systematically developed automated software to detect them one by one. Examples include automated organ volumetrics, bowel analysis, lymphadenopathy detection and visceral fat measurement. Recently, they have developed more holistic assessments that enable detection of numerous different abnormalities using a single training set. An example is the Universal Lesion Detector that can identify numerous different lesion types on body CT scans.

Dr. Summers lab has also contributed multiple radiology datasets to the scientific community, including the large “Chest X-ray8” and “Deep Lesion” datasets.