Malignant Mesothelioma Subtyping of Tissue Images via Sampling Driven Multiple Instance Prediction

Mark Eastwood; Silviu Tudor Marc; Xiaohong Gao; Heba Sailem; Judith Offman; Emmanouil Karteris; Angeles Montero Fernandez; Danny Jonigk; William Cookson; Miriam Moffatt; Sanjay Popat; Fayyaz Minhas; Jan Lukas Robertus; Martin Michalowski; Syed Sibte Raza Abidi; Samina Abidi

doi:10.1007/978-3-031-09342-5_25

Malignant Mesothelioma Subtyping of Tissue Images via Sampling Driven Multiple Instance Prediction

Mark Eastwood, Silviu Tudor Marc, Xiaohong Gao, Heba Sailem, Judith Offman, Emmanouil Karteris, Angeles Montero Fernandez, Danny Jonigk, William Cookson, Miriam Moffatt, Sanjay Popat, Fayyaz Minhas, Jan Lukas Robertus, Martin Michalowski, Syed Sibte Raza Abidi (Editor), Samina Abidi (Editor)

Comprehensive Cancer Centre

Research output: Contribution to conference types › Paper › peer-review

3 Citations (Scopus)

Abstract

Malignant Mesothelioma is a difficult to diagnose and highly lethal cancer usually associated with asbestos exposure. It can be broadly classified into three subtypes: Epitheliod, Sarcomatoid, and Biphasic. Early diagnosis and identification of the subtype informs treatment and can help improve patient outcome. However, the subtyping of malignant mesothelioma, and specifically the recognition of transitional features from routine histology slides has a high level of inter-observer variablity. In this work, we propose the first end-to-end multiple instance learning (MIL) approach for malignant mesothelioma subtyping. This uses an instance-based sampling scheme for training deep convolutional neural networks on this task that allows learning on a wider range of relevant instances compared to max or top-N based MIL approaches. The proposed MIL approach enables identification of malignant mesothelial subtypes of specific tissue regions. From this a continuous characterization of a sample according to predominance of sarcomatoid vs epithelioid regions is possible, thus avoiding the arbitrary and highly subjective categorisation by currently used subtypes. Instance scoring also enables studying tumor heterogeneity and identifying patterns associated with different subtypes. We have evaluated the proposed method on a dataset of 243 tissue micro-array cores with an AUROC of 0.87±0.04 for this task. The dataset and developed methodology is available for the community at: https://github.com/measty/PINS.

Original language	English
Pages	263-272
Number of pages	10
DOIs	https://doi.org/10.1007/978-3-031-09342-5_25
Publication status	Published - 9 Jul 2022

UN SDGs

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

Access to Document

10.1007/978-3-031-09342-5_25

Cite this

Eastwood, M., Marc, S. T., Gao, X., Sailem, H., Offman, J., Karteris, E., Fernandez, A. M., Jonigk, D., Cookson, W., Moffatt, M., Popat, S., Minhas, F., Robertus, J. L., Michalowski, M., Abidi, S. S. R. (Ed.), & Abidi, S. (Ed.) (2022). Malignant Mesothelioma Subtyping of Tissue Images via Sampling Driven Multiple Instance Prediction. 263-272. https://doi.org/10.1007/978-3-031-09342-5_25

@conference{16d9595b4a5a4e5cb641fce48cbaebea,

title = "Malignant Mesothelioma Subtyping of Tissue Images via Sampling Driven Multiple Instance Prediction",

abstract = "Malignant Mesothelioma is a difficult to diagnose and highly lethal cancer usually associated with asbestos exposure. It can be broadly classified into three subtypes: Epitheliod, Sarcomatoid, and Biphasic. Early diagnosis and identification of the subtype informs treatment and can help improve patient outcome. However, the subtyping of malignant mesothelioma, and specifically the recognition of transitional features from routine histology slides has a high level of inter-observer variablity. In this work, we propose the first end-to-end multiple instance learning (MIL) approach for malignant mesothelioma subtyping. This uses an instance-based sampling scheme for training deep convolutional neural networks on this task that allows learning on a wider range of relevant instances compared to max or top-N based MIL approaches. The proposed MIL approach enables identification of malignant mesothelial subtypes of specific tissue regions. From this a continuous characterization of a sample according to predominance of sarcomatoid vs epithelioid regions is possible, thus avoiding the arbitrary and highly subjective categorisation by currently used subtypes. Instance scoring also enables studying tumor heterogeneity and identifying patterns associated with different subtypes. We have evaluated the proposed method on a dataset of 243 tissue micro-array cores with an AUROC of 0.87±0.04 for this task. The dataset and developed methodology is available for the community at: https://github.com/measty/PINS.",

author = "Mark Eastwood and Marc, {Silviu Tudor} and Xiaohong Gao and Heba Sailem and Judith Offman and Emmanouil Karteris and Fernandez, {Angeles Montero} and Danny Jonigk and William Cookson and Miriam Moffatt and Sanjay Popat and Fayyaz Minhas and Robertus, {Jan Lukas} and Martin Michalowski and Abidi, {Syed Sibte Raza} and Samina Abidi",

note = "Funding Information: Acknowledgments. This work was conducted as part of the PRISM project, kindly funded by Cancer Research UK through the CRUK-STFC Early Detection Innovation Award. Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.",

year = "2022",

month = jul,

day = "9",

doi = "10.1007/978-3-031-09342-5_25",

language = "English",

pages = "263--272",

}

TY - CONF

T1 - Malignant Mesothelioma Subtyping of Tissue Images via Sampling Driven Multiple Instance Prediction

AU - Eastwood, Mark

AU - Marc, Silviu Tudor

AU - Gao, Xiaohong

AU - Sailem, Heba

AU - Offman, Judith

AU - Karteris, Emmanouil

AU - Fernandez, Angeles Montero

AU - Jonigk, Danny

AU - Cookson, William

AU - Moffatt, Miriam

AU - Popat, Sanjay

AU - Minhas, Fayyaz

AU - Robertus, Jan Lukas

AU - Michalowski, Martin

A2 - Abidi, Syed Sibte Raza

A2 - Abidi, Samina

N1 - Funding Information: Acknowledgments. This work was conducted as part of the PRISM project, kindly funded by Cancer Research UK through the CRUK-STFC Early Detection Innovation Award. Publisher Copyright: © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

PY - 2022/7/9

Y1 - 2022/7/9

N2 - Malignant Mesothelioma is a difficult to diagnose and highly lethal cancer usually associated with asbestos exposure. It can be broadly classified into three subtypes: Epitheliod, Sarcomatoid, and Biphasic. Early diagnosis and identification of the subtype informs treatment and can help improve patient outcome. However, the subtyping of malignant mesothelioma, and specifically the recognition of transitional features from routine histology slides has a high level of inter-observer variablity. In this work, we propose the first end-to-end multiple instance learning (MIL) approach for malignant mesothelioma subtyping. This uses an instance-based sampling scheme for training deep convolutional neural networks on this task that allows learning on a wider range of relevant instances compared to max or top-N based MIL approaches. The proposed MIL approach enables identification of malignant mesothelial subtypes of specific tissue regions. From this a continuous characterization of a sample according to predominance of sarcomatoid vs epithelioid regions is possible, thus avoiding the arbitrary and highly subjective categorisation by currently used subtypes. Instance scoring also enables studying tumor heterogeneity and identifying patterns associated with different subtypes. We have evaluated the proposed method on a dataset of 243 tissue micro-array cores with an AUROC of 0.87±0.04 for this task. The dataset and developed methodology is available for the community at: https://github.com/measty/PINS.

AB - Malignant Mesothelioma is a difficult to diagnose and highly lethal cancer usually associated with asbestos exposure. It can be broadly classified into three subtypes: Epitheliod, Sarcomatoid, and Biphasic. Early diagnosis and identification of the subtype informs treatment and can help improve patient outcome. However, the subtyping of malignant mesothelioma, and specifically the recognition of transitional features from routine histology slides has a high level of inter-observer variablity. In this work, we propose the first end-to-end multiple instance learning (MIL) approach for malignant mesothelioma subtyping. This uses an instance-based sampling scheme for training deep convolutional neural networks on this task that allows learning on a wider range of relevant instances compared to max or top-N based MIL approaches. The proposed MIL approach enables identification of malignant mesothelial subtypes of specific tissue regions. From this a continuous characterization of a sample according to predominance of sarcomatoid vs epithelioid regions is possible, thus avoiding the arbitrary and highly subjective categorisation by currently used subtypes. Instance scoring also enables studying tumor heterogeneity and identifying patterns associated with different subtypes. We have evaluated the proposed method on a dataset of 243 tissue micro-array cores with an AUROC of 0.87±0.04 for this task. The dataset and developed methodology is available for the community at: https://github.com/measty/PINS.

UR - http://www.scopus.com/inward/record.url?scp=85135087810&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-09342-5_25

DO - 10.1007/978-3-031-09342-5_25

M3 - Paper

SP - 263

EP - 272

ER -

Malignant Mesothelioma Subtyping of Tissue Images via Sampling Driven Multiple Instance Prediction

Abstract

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this