The goal of this project is to develop a based computer aided diagnosis (CAD) algorithm for identification of regions at risk for developing esophageal adenocarcinoma (EAC) in optical coherence tomography (OCT) scans of the esophagus. EAC is one of the deadliest cancers with a 5-year survival rate of less than 20%; yet the standard of care for detecting precursors to EAC is widely recognized to be inadequate. Just recently, a study found that 25% of patients who underwent a standard endoscopic surveillance exam which was found to be clear then went on and progressed to EAC within one year. Clearly todays approach is not working and a significant percentage of disease is being missed. While comprehensive esophageal OCT imaging has shown great potential in addressing this unmet clinical need, one of the main limiters to wider adoption and impact of this technology is the challenge of interpreting the large volume of high-resolution images in real-time. A CAD algorithm would allow OCT to realize its promise in this field and significantly improve the standard of care. Here we propose the development of a deep learning CAD algorithm which will operate on a full patient level volumetric dataset with awareness of the anatomy, robust against image quality and motion artifacts, and trained and validated against a large dataset (>1000 patients). We will aim to go above the performance threshold set by the American Society for Gastroenterology (ASGE) for the performance of advanced imaging in the detection of high grade dysplasia in BE, and include low grade dysplasia while maintaining a Sensitivity and Specificity of 90/80%.
Public Health Relevance Statement: PROJECT NARRATIVE The goal of this project is to develop a computer aided diagnosis (CAD) algorithm for use with the NinePoint Medical OCT imaging system and provide early detection of pre-cancerous lesions. Earlier detection leads to earlier treatment, more positive health outcomes for patients and lower healthcare system costs.
Project Terms: Address; Adoption; Algorithms; American; Anatomy; Awareness; Barrett Esophagus; base; Biopsy; case-based; classification algorithm; Clinical; Clinical Management; Clinical Research; Clinical Trials; Computer software; Computer-Assisted Diagnosis; cost; Data; Data Set; deep learning; design; Detection; Development; diagnosis quality; Disease; Dysplasia; Early Diagnosis; Early treatment; effective therapy; Esophageal Adenocarcinoma; Esophageal Diseases; Esophageal Intraepithelial Neoplasia; Esophagus; Feedback; Gastroenterology; Goals; Grant; Health; Health Care Costs; Healthcare Systems; High grade dysplasia; high resolution imaging; Image; imaging platform; imaging system; improved; in vivo; interest; large datasets; Lead; Lesion; Machine Learning; Malignant Neoplasms; Medical; meetings; Morphologic artifacts; Motion; Optical Coherence Tomography; Patient-Focused Outcomes; Patients; Performance; Phase; Physicians; premalignant; Process; prospective; Reporting; Risk; Sample Size; Scanning; Sensitivity and Specificity; Small Business Innovation Research Grant; Societies; software development; Software Tools; Speed; standard of care; Survival Rate; Techniques; Technology; Time; Training; validation studies
