Applying likelihood-based out-of-distribution detection to malaria microscopy using Deep Diffusion Models

Joe Goodier; Richard W. Bowman; Pietro Cicuta; Joe Knapper; Samuel Mcdermott; Joram Mduda; Catherine Mkindi; Joel  Collins; Julian Stirling; William Wadsworth; Boyko Vodenicharski; Jessica Nicholson; Neill Campbell

Applying likelihood-based out-of-distribution detection to malaria microscopy using Deep Diffusion Models

Joe Goodier, Richard W. Bowman, Pietro Cicuta, Joe Knapper, Samuel Mcdermott, Joram Mduda, Catherine Mkindi, Joel Collins, Julian Stirling, William Wadsworth, Boyko Vodenicharski, Jessica Nicholson, Neill Campbell

Research output: Chapter or section in a book/report/conference proceeding › Book chapter

Abstract

Malaria is a serious febrile illness affecting nearly a quarter of a billion people per year, and responsible for half a million deaths. The gold standard for malaria diagnosis is microscopic examination of blood films. Poor slide quality and suboptimal imaging can lead to misdiagnosis and are common problems when collecting training data for diagnostics. We propose leveraging Out-of-Distribution detection to improve diagnostics by employing probabilistic generative models to detect deviations from healthy
samples. We use a class-conditioned diffusion model to detect potentially suboptimal and pathological images in a Giemsa-stained, thin-film microscopy dataset. This is achieved by using a Deep Denoising Diffusion Model to build a Diffusion Classifier model. Our results demonstrate the effectiveness of this approach, offering a promising avenue for enhancing malaria detection and triaging care in resource-limited settings

Original language	English
Title of host publication	Frontiers of Medical Technology
Editors	Moi Hoon Yap, Timothy Cootes, Reyer Zwiggelaar, Neil Reeves
Place of Publication	London, U. K.
Publisher	Frontiers Media
Chapter	Set 2
Pages	123-129
Number of pages	6
ISBN (Print)	9782832512449
Publication status	Published - 7 Oct 2024

Funding

The authors acknowledge support from the Wellcome Collaborative Award in Science (224390/Z/21/Z) and Women In STEM Hub (WISH) seed funding from Kingston University.

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being

Access to Document

https://www.frontiersin.org/books/Medical_Image_Understanding_and_Analysis/12759

Centre for the Analysis of Motion, Entertainment Research and Applications (CAMERA) - 2.0
Campbell, N. (PI), Cosker, D. (PI), Bilzon, J. (CoI), Campbell, N. (CoI), Cazzola, D. (CoI), Colyer, S. (CoI), Cosker, D. (CoI), Lutteroth, C. (CoI), McGuigan, P. (CoI), O'Neill, E. (CoI), Petrini, K. (CoI), Proulx, M. (CoI) & Yang, Y. (CoI)
Engineering and Physical Sciences Research Council
1/11/20 → 31/10/25
Project: Research council
Centre for the Analysis of Motion, Entertainment Research and Applications (CAMERA)
Cosker, D. (PI), Bilzon, J. (CoI), Campbell, N. (CoI), Cazzola, D. (CoI), Colyer, S. (CoI), Fincham Haines, T. (CoI), Hall, P. (CoI), Kim, K. I. (CoI), Lutteroth, C. (CoI), McGuigan, P. (CoI), O'Neill, E. (CoI), Richardt, C. (CoI), Salo, A. (CoI), Seminati, E. (CoI), Tabor, A. (CoI) & Yang, Y. (CoI)
Engineering and Physical Sciences Research Council
1/09/15 → 28/02/21
Project: Research council

Cite this

Goodier, J., Bowman, R. W., Cicuta, P., Knapper, J., Mcdermott, S., Mduda, J., Mkindi, C., Collins, J., Stirling, J., Wadsworth, W., Vodenicharski, B., Nicholson, J., & Campbell, N. (2024). Applying likelihood-based out-of-distribution detection to malaria microscopy using Deep Diffusion Models. In M. H. Yap, T. Cootes, R. Zwiggelaar, & N. Reeves (Eds.), Frontiers of Medical Technology (pp. 123-129). Article 3 Frontiers Media. https://www.frontiersin.org/books/Medical_Image_Understanding_and_Analysis/12759

Applying likelihood-based out-of-distribution detection to malaria microscopy using Deep Diffusion Models. / Goodier, Joe; Bowman, Richard W.; Cicuta, Pietro et al.
Frontiers of Medical Technology. ed. / Moi Hoon Yap; Timothy Cootes; Reyer Zwiggelaar; Neil Reeves. London, U. K.: Frontiers Media, 2024. p. 123-129 3.

Research output: Chapter or section in a book/report/conference proceeding › Book chapter

Goodier, J, Bowman, RW, Cicuta, P, Knapper, J, Mcdermott, S, Mduda, J, Mkindi, C, Collins, J, Stirling, J, Wadsworth, W, Vodenicharski, B, Nicholson, J & Campbell, N 2024, Applying likelihood-based out-of-distribution detection to malaria microscopy using Deep Diffusion Models. in MH Yap, T Cootes, R Zwiggelaar & N Reeves (eds), Frontiers of Medical Technology., 3, Frontiers Media, London, U. K., pp. 123-129. <https://www.frontiersin.org/books/Medical_Image_Understanding_and_Analysis/12759>

@inbook{6c93214d39554dfd92e755e04d3f862e,

title = "Applying likelihood-based out-of-distribution detection to malaria microscopy using Deep Diffusion Models",

abstract = "Malaria is a serious febrile illness affecting nearly a quarter of a billion people per year, and responsible for half a million deaths. The gold standard for malaria diagnosis is microscopic examination of blood films. Poor slide quality and suboptimal imaging can lead to misdiagnosis and are common problems when collecting training data for diagnostics. We propose leveraging Out-of-Distribution detection to improve diagnostics by employing probabilistic generative models to detect deviations from healthy samples. We use a class-conditioned diffusion model to detect potentially suboptimal and pathological images in a Giemsa-stained, thin-film microscopy dataset. This is achieved by using a Deep Denoising Diffusion Model to build a Diffusion Classifier model. Our results demonstrate the effectiveness of this approach, offering a promising avenue for enhancing malaria detection and triaging care in resource-limited settings",

author = "Joe Goodier and Bowman, \{Richard W.\} and Pietro Cicuta and Joe Knapper and Samuel Mcdermott and Joram Mduda and Catherine Mkindi and Joel Collins and Julian Stirling and William Wadsworth and Boyko Vodenicharski and Jessica Nicholson and Neill Campbell",

year = "2024",

month = oct,

day = "7",

language = "English",

isbn = "9782832512449",

pages = "123--129",

editor = "Yap, \{Moi Hoon\} and Timothy Cootes and Reyer Zwiggelaar and Neil Reeves",

booktitle = "Frontiers of Medical Technology",

publisher = "Frontiers Media",

}

TY - CHAP

T1 - Applying likelihood-based out-of-distribution detection to malaria microscopy using Deep Diffusion Models

AU - Goodier, Joe

AU - Bowman, Richard W.

AU - Cicuta, Pietro

AU - Knapper, Joe

AU - Mcdermott, Samuel

AU - Mduda, Joram

AU - Mkindi, Catherine

AU - Collins, Joel

AU - Stirling, Julian

AU - Wadsworth, William

AU - Vodenicharski, Boyko

AU - Nicholson, Jessica

AU - Campbell, Neill

PY - 2024/10/7

Y1 - 2024/10/7

N2 - Malaria is a serious febrile illness affecting nearly a quarter of a billion people per year, and responsible for half a million deaths. The gold standard for malaria diagnosis is microscopic examination of blood films. Poor slide quality and suboptimal imaging can lead to misdiagnosis and are common problems when collecting training data for diagnostics. We propose leveraging Out-of-Distribution detection to improve diagnostics by employing probabilistic generative models to detect deviations from healthy samples. We use a class-conditioned diffusion model to detect potentially suboptimal and pathological images in a Giemsa-stained, thin-film microscopy dataset. This is achieved by using a Deep Denoising Diffusion Model to build a Diffusion Classifier model. Our results demonstrate the effectiveness of this approach, offering a promising avenue for enhancing malaria detection and triaging care in resource-limited settings

AB - Malaria is a serious febrile illness affecting nearly a quarter of a billion people per year, and responsible for half a million deaths. The gold standard for malaria diagnosis is microscopic examination of blood films. Poor slide quality and suboptimal imaging can lead to misdiagnosis and are common problems when collecting training data for diagnostics. We propose leveraging Out-of-Distribution detection to improve diagnostics by employing probabilistic generative models to detect deviations from healthy samples. We use a class-conditioned diffusion model to detect potentially suboptimal and pathological images in a Giemsa-stained, thin-film microscopy dataset. This is achieved by using a Deep Denoising Diffusion Model to build a Diffusion Classifier model. Our results demonstrate the effectiveness of this approach, offering a promising avenue for enhancing malaria detection and triaging care in resource-limited settings

M3 - Book chapter

SN - 9782832512449

SP - 123

EP - 129

BT - Frontiers of Medical Technology

A2 - Yap, Moi Hoon

A2 - Cootes, Timothy

A2 - Zwiggelaar, Reyer

A2 - Reeves, Neil

PB - Frontiers Media

CY - London, U. K.

ER -

Applying likelihood-based out-of-distribution detection to malaria microscopy using Deep Diffusion Models

Abstract

Funding

UN SDGs

Access to Document

Fingerprint

Projects

Centre for the Analysis of Motion, Entertainment Research and Applications (CAMERA) - 2.0

Centre for the Analysis of Motion, Entertainment Research and Applications (CAMERA)

Cite this