Mitigating Bias in Radiology Machine Learning: 2. Model Development

Kuan Zhang; Bardia Khosravi; Sanaz Vahdati; Shahriar Faghani; Fred Nugen; Seyed Moein Rassoulinejad-Mousavi; Mana Moassefi; Jaidip Manikrao M. Jagtap; Yashbir Singh; Pouria Rouzrokh; Bradley J. Erickson

doi:10.1148/ryai.220010

Mitigating Bias in Radiology Machine Learning: 2. Model Development

Kuan Zhang, Bardia Khosravi, Sanaz Vahdati, Shahriar Faghani, Fred Nugen, Seyed Moein Rassoulinejad-Mousavi, Mana Moassefi, Jaidip Manikrao M. Jagtap, Yashbir Singh, Pouria Rouzrokh, Bradley J. Erickson

Radiology

Research output: Contribution to journal › Article › peer-review

Abstract

There are increasing concerns about the bias and fairness of artificial intelligence (AI) models as they are put into clinical practice. Among the steps for implementing machine learning tools into clinical workflow, model development is an important stage where different types of biases can occur. This report focuses on four aspects of model development where such bias may arise: data augmentation, model and loss function, optimizers, and transfer learning. This report emphasizes appropriate considerations and practices that can mitigate biases in radiology AI studies.

Original language	English (US)
Article number	e220010
Journal	Radiology: Artificial Intelligence
Volume	4
Issue number	5
DOIs	https://doi.org/10.1148/ryai.220010
State	Published - Sep 2022

Keywords

Bias
Deep Learning
Machine Learning
Model
Radiology

ASJC Scopus subject areas

Radiological and Ultrasound Technology
Radiology Nuclear Medicine and imaging
Artificial Intelligence

Access to Document

10.1148/ryai.220010

Cite this

@article{903734c902204171b6a56378d2d676a2,

title = "Mitigating Bias in Radiology Machine Learning: 2. Model Development",

abstract = "There are increasing concerns about the bias and fairness of artificial intelligence (AI) models as they are put into clinical practice. Among the steps for implementing machine learning tools into clinical workflow, model development is an important stage where different types of biases can occur. This report focuses on four aspects of model development where such bias may arise: data augmentation, model and loss function, optimizers, and transfer learning. This report emphasizes appropriate considerations and practices that can mitigate biases in radiology AI studies.",

keywords = "Bias, Deep Learning, Machine Learning, Model, Radiology",

author = "Kuan Zhang and Bardia Khosravi and Sanaz Vahdati and Shahriar Faghani and Fred Nugen and Rassoulinejad-Mousavi, {Seyed Moein} and Mana Moassefi and Jagtap, {Jaidip Manikrao M.} and Yashbir Singh and Pouria Rouzrokh and Erickson, {Bradley J.}",

note = "Publisher Copyright: {\textcopyright} RSNA, 2022.",

year = "2022",

month = sep,

doi = "10.1148/ryai.220010",

language = "English (US)",

volume = "4",

journal = "Radiology: Artificial Intelligence",

issn = "2638-6100",

publisher = "Radiological Society of North America Inc.",

number = "5",

}

TY - JOUR

T1 - Mitigating Bias in Radiology Machine Learning

T2 - 2. Model Development

AU - Zhang, Kuan

AU - Khosravi, Bardia

AU - Vahdati, Sanaz

AU - Faghani, Shahriar

AU - Nugen, Fred

AU - Rassoulinejad-Mousavi, Seyed Moein

AU - Moassefi, Mana

AU - Jagtap, Jaidip Manikrao M.

AU - Singh, Yashbir

AU - Rouzrokh, Pouria

AU - Erickson, Bradley J.

N1 - Publisher Copyright: © RSNA, 2022.

PY - 2022/9

Y1 - 2022/9

N2 - There are increasing concerns about the bias and fairness of artificial intelligence (AI) models as they are put into clinical practice. Among the steps for implementing machine learning tools into clinical workflow, model development is an important stage where different types of biases can occur. This report focuses on four aspects of model development where such bias may arise: data augmentation, model and loss function, optimizers, and transfer learning. This report emphasizes appropriate considerations and practices that can mitigate biases in radiology AI studies.

AB - There are increasing concerns about the bias and fairness of artificial intelligence (AI) models as they are put into clinical practice. Among the steps for implementing machine learning tools into clinical workflow, model development is an important stage where different types of biases can occur. This report focuses on four aspects of model development where such bias may arise: data augmentation, model and loss function, optimizers, and transfer learning. This report emphasizes appropriate considerations and practices that can mitigate biases in radiology AI studies.

KW - Bias

KW - Deep Learning

KW - Machine Learning

KW - Model

KW - Radiology

UR - http://www.scopus.com/inward/record.url?scp=85139096480&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85139096480&partnerID=8YFLogxK

U2 - 10.1148/ryai.220010

DO - 10.1148/ryai.220010

M3 - Article

AN - SCOPUS:85139096480

SN - 2638-6100

VL - 4

JO - Radiology: Artificial Intelligence

JF - Radiology: Artificial Intelligence

IS - 5

M1 - e220010

ER -

Mitigating Bias in Radiology Machine Learning: 2. Model Development

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this