Premium
WE‐E‐103‐02: Past, Present and Future Roles of ROC Analysis in Medical Imaging and Quantitative Image Analysis
Author(s) -
Giger M,
Toledano A,
Myers K,
Jiang Y
Publication year - 2013
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4815603
Subject(s) - medical physics , receiver operating characteristic , medical imaging , modalities , clinical trial , computer science , precision medicine , data science , artificial intelligence , medicine , machine learning , pathology , social science , sociology
Receiver Operating Characteristic (ROC) analysis has been a mainstay of many research developments as well as various clinical studies/trials. It has provided medical physicists with a way to objectively measure how data are presented in an image, how people perceive those images, and how one can compare different observers or different imaging modalities with each other. ROC analysis plays an important role in both technology assessment and clinical decision‐making, especially as various aspects of imaging biomarkers and personalized medicine are evaluated. Over the past five years, on average, almost 40 papers/year that were published in MEDICAL PHYSICS utilized ROC analysis. The challenges and opportunities in ROC analysis research and in its application in various tasks are active areas, including expanding the mathematical formulation for multiple lesions per image, location‐based sensitivity, and evaluation without ground truth, as well as expanding its role in imaging biomarker validation, assessing response to therapy, theranostics, and image‐based phenotyping with genomics (image‐omics). Learning Objectives: 1. Review the mathematical foundations and implementation of ROC analysis in basic research and clinical trials; and understand the role and limitations of ROC analysis in large scale clinical studies/trials 2. Recognize advances in ROC analysis in order to incorporate multiple lesions per image, location‐based sensitivity, evaluation without ground truth, theranostics, and others 3. Appreciate the evolving role of ROC analysis in the evaluation of imaging biomarkers and image‐based phenotyping Research supported by NIH, DOE, and DOD. COI: Stockholder, Hologic, Inc Shareholder, Quantitative Insights, Inc Royalties, Hologic, Inc Royalties, General Electric Company Royalties, MEDIAN Technologies Royalties, Riverain Technologies, LLC Royalties, Mitsubishi Corporation Royalties, Toshiba Maryellen Giger Corporation Researcher, Koninklijke Philips Electronics NV Researcher, U‐Systems, Inc