_version_ 1866911783638269952
author Rikhye, Rajeev V.
Loh, Aaron
Hong, Grace Eunhae
Singh, Preeti
Smith, Margaret Ann
Muralidharan, Vijaytha
Wong, Doris
Sayres, Rory
Phung, Michelle
Betancourt, Nicolas
Fong, Bradley
Sahasrabudhe, Rachna
Nasim, Khoban
Eschholz, Alec
Mustafa, Basil
Freyberg, Jan
Spitz, Terry
Matias, Yossi
Corrado, Greg S.
Chou, Katherine
Webster, Dale R.
Bui, Peggy
Liu, Yuan
Liu, Yun
Ko, Justin
Lin, Steven
author_facet Rikhye, Rajeev V.
Loh, Aaron
Hong, Grace Eunhae
Singh, Preeti
Smith, Margaret Ann
Muralidharan, Vijaytha
Wong, Doris
Sayres, Rory
Phung, Michelle
Betancourt, Nicolas
Fong, Bradley
Sahasrabudhe, Rachna
Nasim, Khoban
Eschholz, Alec
Mustafa, Basil
Freyberg, Jan
Spitz, Terry
Matias, Yossi
Corrado, Greg S.
Chou, Katherine
Webster, Dale R.
Bui, Peggy
Liu, Yuan
Liu, Yun
Ko, Justin
Lin, Steven
contents Recently, there has been great progress in the ability of artificial intelligence (AI) algorithms to classify dermatological conditions from clinical photographs. However, little is known about the robustness of these algorithms in real-world settings where several factors can lead to a loss of generalizability. Understanding and overcoming these limitations will permit the development of generalizable AI that can aid in the diagnosis of skin conditions across a variety of clinical settings. In this retrospective study, we demonstrate that differences in skin condition distribution, rather than in demographics or image capture mode are the main source of errors when an AI algorithm is evaluated on data from a previously unseen source. We demonstrate a series of steps to close this generalization gap, requiring progressively more information about the new source, ranging from the condition distribution to training data enriched for data less frequently seen during training. Our results also suggest comparable performance from end-to-end fine tuning versus fine tuning solely the classification layer on top of a frozen embedding model. Our approach can inform the adaptation of AI algorithms to new settings, based on the information and resources available.
format Preprint
id arxiv_https___arxiv_org_abs_2402_15566
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Closing the AI generalization gap by adjusting for dermatology condition distribution differences across clinical settings
Rikhye, Rajeev V.
Loh, Aaron
Hong, Grace Eunhae
Singh, Preeti
Smith, Margaret Ann
Muralidharan, Vijaytha
Wong, Doris
Sayres, Rory
Phung, Michelle
Betancourt, Nicolas
Fong, Bradley
Sahasrabudhe, Rachna
Nasim, Khoban
Eschholz, Alec
Mustafa, Basil
Freyberg, Jan
Spitz, Terry
Matias, Yossi
Corrado, Greg S.
Chou, Katherine
Webster, Dale R.
Bui, Peggy
Liu, Yuan
Liu, Yun
Ko, Justin
Lin, Steven
Image and Video Processing
Computer Vision and Pattern Recognition
Machine Learning
Recently, there has been great progress in the ability of artificial intelligence (AI) algorithms to classify dermatological conditions from clinical photographs. However, little is known about the robustness of these algorithms in real-world settings where several factors can lead to a loss of generalizability. Understanding and overcoming these limitations will permit the development of generalizable AI that can aid in the diagnosis of skin conditions across a variety of clinical settings. In this retrospective study, we demonstrate that differences in skin condition distribution, rather than in demographics or image capture mode are the main source of errors when an AI algorithm is evaluated on data from a previously unseen source. We demonstrate a series of steps to close this generalization gap, requiring progressively more information about the new source, ranging from the condition distribution to training data enriched for data less frequently seen during training. Our results also suggest comparable performance from end-to-end fine tuning versus fine tuning solely the classification layer on top of a frozen embedding model. Our approach can inform the adaptation of AI algorithms to new settings, based on the information and resources available.
title Closing the AI generalization gap by adjusting for dermatology condition distribution differences across clinical settings
topic Image and Video Processing
Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2402.15566