AAPM Task Group Report 273 - Recommendations on Best Practices for AI and Machine Learning for Computer-Aided Diagnosis in Medical Imaging

Aug 4, 2026

Upload by Jiven

Peer-reviewed recommendations for applying artificial intelligence and machine learning to computer-aided diagnosis in medical imaging. The report organizes guidance across data sourcing and sampling, including ethics, de-identification, diversity and inclusion, quality, augmentation, and harmonization. It further addresses reference standards and annotation practices, comparing objective versus subjective targets and multiple annotation granularities. It concludes with practical model development considerations, covering sampling strategies and learning paradigms, from supervised to self-supervised approaches.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages45

Reading Time1 h 53 min

AccessFree to download

Abstract

1 Introduction

2 Data

2.1 Data Collection

2.2 Data Augmentation

2.3 Data Harmonization

2.4 Take Home Message on Data

3 Reference Standards

3.1 Objective vs. Subjective Reference Standards

3.2 Annotation Granularity

3.3 Methods for Acquiring Annotations

3.5 Take Home Message on Reference Standards

4 Model Development

4.1 Data Sampling Strategies

4.2 Machine Learning Strategies

POLITECNICO DI TORINO Repository ISTITUZIONALE AAPM Task Group Report 273: Recommendations on Best Practices for AI and Machine Learning for Computer-Aided Diagnosis in Medical Imaging Original AAPM Task Group Report 273: Recommendations on Best Practices for AI and Machine Learning for Computer-Aided Diagnosis in Medical Imaging / Hadjiiski, Lubomir; Cha, Kenny; Chan, Heang-Ping; Drukker, Karen; Morra, Lia; Näppi, Janne J. ; Sahiner, Berkman; Yoshida, Hiroyuki; Chen, Quan; Deserno, Thomas M. ; Greenspan, Hayit; Huisman, Henkjan; Huo, Zhimin; Mazurchuk, Richard; Petrick, Nicholas; Regge, Daniele; Samala, Ravi; Summers, Ronald M. ; Suzuki, Kenji; Tourassi, Georgia; Vergara, Daniel; Armato III, Samuel G.. -In: MEDICAL PHYSICS. -ISSN 0094-2405. - :2ai2aiiy):, pp. 1-24. [10.1002/mp.16188] This version is available at: 11583/2973414 since: 2023-01-06T15:19:18Z Publisher: Wiley Published DOI:10.1002/mp.16188 Terms of use: This article is made available under terms and conditions as specified in the corresponding bibliographic description in the repository Publisher copyright Wiley postprint/Author's Accepted Manuscript This is the peer reviewed version of the above quoted article, which has been published in final form at [http://dx.doi.org/10.1002/mp.16188.This article may be](http://dx.doi.org/10.1002/mp.16188.This article may be) used for non-commercial purposes in accordance with Wiley Terms and Conditions for Use of Self-Archived Versions. (Article begins on next page) 18 September 2024 1 AAPM Task Group Report 273: Recommendations on Best 2 Practices for AI and Machine Learning for Computer-Aided 3 Diagnosis in Medical Imaging 4 5 Lubomir Hadjiiski 6 Department of Radiology, University of Michigan, Ann Arbor, Michigan, USA 7 8 Kenny Cha 9 U.S. Food and Drug Administration, Silver Spring, Maryland, USA 10 11 Heang-Ping Chan 12 Department of Radiology, University of Michigan, Ann Arbor, Michigan, USA 13 14 Karen Drukker 15 Department of Radiology, University of Chicago, Chicago, Illinois, USA 16 17 Lia Morra 18 Department of Control and Computer Engineering, Politecnico di Torino, Torino, Italy 19 20 Janne J. Näppi 21 3D Imaging Research, Department of Radiology, Massachusetts General Hospital and Harvard 22 Medical School, Boston, Massachusetts, USA 23 24 Berkman Sahiner 25 U.S. Food and Drug Administration, Silver Spring, Maryland, USA 26 27 Hiroyuki Yoshida 28 3D Imaging Research, Department of Radiology, Massachusetts General Hospital and Harvard 29 Medical School, Boston, Massachusetts, USA 30 31 Quan Chen 32 Department of Radiation Medicine, University of Kentucky, Lexington, Kentucky, USA 33 34 Thomas M. Deserno 35 Peter L. Reichertz Institute for Medical Informatics of TU Braunschweig and Hannover Medical 36 School, Braunschweig, Germany 37 38 Hayit Greenspan 39 Department of Biomedical Engineering, Faculty of Engineering, Tel Aviv University, Tel Aviv, 40 Israel & Department of Radiology, Ichan School of Medicine, Mt Sinai, NYC, NY, USA 41 42 43 Henkjan Huisman 44 Radboud University Medical Center, Radboud Institute for Health Sciences, Nijmegen, The 45 Netherlands 46 47 Zhimin Huo 48 Tencent America, Palo Alto, CA 49 50 Richard Mazurchuk 51 Division of Cancer Prevention, National Cancer Institute, National Institutes of Health, 52 Bethesda, Maryland, USA 53 54 Nicholas Petrick 55 U.S. Food and Drug Administration, Silver Spring, Maryland, USA 56 57 Daniele Regge 58 Radiology Unit, Candiolo Cancer Institute, FPO-IRCCS, Candiolo 59 Department of Surgical Sciences, University of Turin, Turin, Italy 60 61 Ravi Samala 62 U.S. Food and Drug Administration, Silver Spring, Maryland, USA 63 64 Ronald M. Summers 65 Radiology and Imaging Sciences, National Institutes of Health Clinical Center, Bethesda, 66 Maryland, USA 67 68 Kenji Suzuki 69 Institute of Innovative Research, Tokyo Institute of Tech

FAQs

What topics are covered under the report’s guidance for data?

The report covers data collection and case sampling, public databases, ethics considerations (including de-identification), diversity and inclusion, quality considerations, plus data augmentation and data harmonization.

How does the report address reference standards and annotations?

It discusses objective versus subjective reference standards, multiple annotation granularities (entire image, region-based, pixel-based), methods for acquiring annotations (expert labels, electronic health record, crowd sourcing, phantoms, weak/noisy labels), and how true positives are defined.

Which machine learning approaches are included in model development recommendations?

Model development guidance includes data sampling strategies and learning supervision levels, spanning supervised, semi-supervised, self-supervised, and unsupervised learning.

Aug 4, 2026

Upload by Jiven