Colon cancer, the second leading cause of cancer deaths for men and women in the United States, can be prevented by early detection and removal of its precursor lesions. Computed tomographic colonography (CTC), also known as virtual colonoscopy, could substantially increase the capacity, safety, and patient compliance of colorectal examinations. However, an FDA panel has recently identified two remaining concerns about CTC: patient adherence, and the detection of small polyps and flat lesions. Our clinical multi-center trial showed that laxative-free preparation by oral ingestion of a contrast agent (iodine) to indicate fecal materials for electronic cleansing (EC), followed by computer-aided detection (CADe), makes CTC easy to tolerate for patients while enabling the detection of ?10 mm lesions at sensitivity comparable to that of optical colonoscopy. However, small polyps and flat lesions were a significant source of false negatives, because EC produced image artifacts that imitated such lesions. Because laxative-free CTC addresses the concern of patient adherence, the only remaining concern about CTC is the detection of small polyps and flat lesions. The goal of this project is to develop a novel multi-material deep-learning scheme, hereafter denoted as Deep- ECAD, that integrates EC and CADe for the detection of small polyps and flat lesions in laxative-free spectral CTC (spCTC), where spectral imaging and deep learning will be used to overcome the above limitations of conventional CTC.
Our specific aims are to (1) establish a laxative-free ultra-low-dose spCTC image database, (2) develop a multi-material deep-learning method for EC, (3) develop deep radiomic detection of small polyps and flat lesions, and (4) evaluate the clinical benefit of Deep-ECAD with laxative-free cases. Successful development of the proposed Deep-ECAD scheme will substantially improve human readers? performance in the detection of small polyps and flat lesions while minimizing the inconveniences of bowel preparation and radiation risk to patients. Such a scheme will make laxative-free spCTC a highly accurate and acceptable screening option for large populations, in particular, Medicare population, leading to an increased screening rate, promoting early diagnosis of colon cancer, and ultimately reducing mortality due to colon cancer.

Public Health Relevance

Successful development of the proposed deep-learning EC-CADe scheme for detecting small polyps and flat lesions in ultra-low-dose laxative-free spCTC (Deep-ECAD) will substantially improve reader performance in the detection of small polyps and flat lesions while minimizing the inconveniences of bowel preparation and radiation risk to patients. Such a scheme will make laxative-free CTC a highly accurate and acceptable screening option for large populations, in particular, Medicare population, leading to an increased screening rate, promoting early diagnosis of colon cancer, and ultimately reducing mortality due to colon cancer.

Agency
National Institute of Health (NIH)
Institute
National Institute of Biomedical Imaging and Bioengineering (NIBIB)
Type
Exploratory/Developmental Grants (R21)
Project #
1R21EB024025-01
Application #
9297792
Study Section
Biomedical Computing and Health Informatics Study Section (BCHI)
Program Officer
Krosnick, Steven
Project Start
2017-07-05
Project End
2019-04-30
Budget Start
2017-07-05
Budget End
2018-04-30
Support Year
1
Fiscal Year
2017
Total Cost
Indirect Cost
Name
Massachusetts General Hospital
Department
Type
DUNS #
073130411
City
Boston
State
MA
Country
United States
Zip Code
02114
Tachibana, Rie; Näppi, Janne J; Ota, Junko et al. (2018) Deep Learning Electronic Cleansing for Single- and Dual-Energy CT Colonography. Radiographics 38:2034-2050