Factors determining generalization in deep learning models for scoring COVID-CT images

The COVID-19 pandemic has inspired unprecedented data collection and computer vision modelling efforts worldwide, focused on the diagnosis of COVID-19 from medical images. However, these models have found limited, if any, clinical application due in part to unproven generalization to data sets beyon...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Michael James Horry, Subrata Chakraborty, Biswajeet Pradhan, Maryam Fallahpoor, Hossein Chegeni, Manoranjan Paul
Formato:	article
Lenguaje:	EN
Publicado:	AIMS Press 2021
Materias:	covid-19 scoring computed tomography deep learning external validation model generalization image pre-processing Biotechnology TP248.13-248.65 Mathematics QA1-939
Acceso en línea:	https://doaj.org/article/1294cd35fd75485a91ae3b3e461cd519
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:1294cd35fd75485a91ae3b3e461cd519
record_format	dspace
spelling	oai:doaj.org-article:1294cd35fd75485a91ae3b3e461cd5192021-11-29T05:55:34ZFactors determining generalization in deep learning models for scoring COVID-CT images10.3934/mbe.20214561551-0018https://doaj.org/article/1294cd35fd75485a91ae3b3e461cd5192021-10-01T00:00:00Zhttps://www.aimspress.com/article/doi/10.3934/mbe.2021456?viewType=HTMLhttps://doaj.org/toc/1551-0018The COVID-19 pandemic has inspired unprecedented data collection and computer vision modelling efforts worldwide, focused on the diagnosis of COVID-19 from medical images. However, these models have found limited, if any, clinical application due in part to unproven generalization to data sets beyond their source training corpus. This study investigates the generalizability of deep learning models using publicly available COVID-19 Computed Tomography data through cross dataset validation. The predictive ability of these models for COVID-19 severity is assessed using an independent dataset that is stratified for COVID-19 lung involvement. Each inter-dataset study is performed using histogram equalization, and contrast limited adaptive histogram equalization with and without a learning Gabor filter. We show that under certain conditions, deep learning models can generalize well to an external dataset with F1 scores up to 86%. The best performing model shows predictive accuracy of between 75% and 96% for lung involvement scoring against an external expertly stratified dataset. From these results we identify key factors promoting deep learning generalization, being primarily the uniform acquisition of training images, and secondly diversity in CT slice position.Michael James HorrySubrata ChakrabortyBiswajeet PradhanMaryam Fallahpoor Hossein ChegeniManoranjan PaulAIMS Pressarticlecovid-19 scoringcomputed tomographydeep learningexternal validationmodel generalizationimage pre-processingBiotechnologyTP248.13-248.65MathematicsQA1-939ENMathematical Biosciences and Engineering, Vol 18, Iss 6, Pp 9264-9293 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	covid-19 scoring computed tomography deep learning external validation model generalization image pre-processing Biotechnology TP248.13-248.65 Mathematics QA1-939
spellingShingle	covid-19 scoring computed tomography deep learning external validation model generalization image pre-processing Biotechnology TP248.13-248.65 Mathematics QA1-939 Michael James Horry Subrata Chakraborty Biswajeet Pradhan Maryam Fallahpoor Hossein Chegeni Manoranjan Paul Factors determining generalization in deep learning models for scoring COVID-CT images
description	The COVID-19 pandemic has inspired unprecedented data collection and computer vision modelling efforts worldwide, focused on the diagnosis of COVID-19 from medical images. However, these models have found limited, if any, clinical application due in part to unproven generalization to data sets beyond their source training corpus. This study investigates the generalizability of deep learning models using publicly available COVID-19 Computed Tomography data through cross dataset validation. The predictive ability of these models for COVID-19 severity is assessed using an independent dataset that is stratified for COVID-19 lung involvement. Each inter-dataset study is performed using histogram equalization, and contrast limited adaptive histogram equalization with and without a learning Gabor filter. We show that under certain conditions, deep learning models can generalize well to an external dataset with F1 scores up to 86%. The best performing model shows predictive accuracy of between 75% and 96% for lung involvement scoring against an external expertly stratified dataset. From these results we identify key factors promoting deep learning generalization, being primarily the uniform acquisition of training images, and secondly diversity in CT slice position.
format	article
author	Michael James Horry Subrata Chakraborty Biswajeet Pradhan Maryam Fallahpoor Hossein Chegeni Manoranjan Paul
author_facet	Michael James Horry Subrata Chakraborty Biswajeet Pradhan Maryam Fallahpoor Hossein Chegeni Manoranjan Paul
author_sort	Michael James Horry
title	Factors determining generalization in deep learning models for scoring COVID-CT images
title_short	Factors determining generalization in deep learning models for scoring COVID-CT images
title_full	Factors determining generalization in deep learning models for scoring COVID-CT images
title_fullStr	Factors determining generalization in deep learning models for scoring COVID-CT images
title_full_unstemmed	Factors determining generalization in deep learning models for scoring COVID-CT images
title_sort	factors determining generalization in deep learning models for scoring covid-ct images
publisher	AIMS Press
publishDate	2021
url	https://doaj.org/article/1294cd35fd75485a91ae3b3e461cd519
work_keys_str_mv	AT michaeljameshorry factorsdetermininggeneralizationindeeplearningmodelsforscoringcovidctimages AT subratachakraborty factorsdetermininggeneralizationindeeplearningmodelsforscoringcovidctimages AT biswajeetpradhan factorsdetermininggeneralizationindeeplearningmodelsforscoringcovidctimages AT maryamfallahpoor factorsdetermininggeneralizationindeeplearningmodelsforscoringcovidctimages AT hosseinchegeni factorsdetermininggeneralizationindeeplearningmodelsforscoringcovidctimages AT manoranjanpaul factorsdetermininggeneralizationindeeplearningmodelsforscoringcovidctimages
_version_	1718407569680105472

Factors determining generalization in deep learning models for scoring COVID-CT images

Ejemplares similares