Multi-view Modelling of Longitudinal Health Data for Improved Prognostication of Colorectal Cancer Recurrence

Danliang Ho, Mehul Motani
Proceedings of the 8th Machine Learning for Healthcare Conference, PMLR 219:265-284, 2023.

Abstract

Colorectal cancer (CRC) is a leading cause of cancer-related deaths worldwide, with a high incidence of recurrence following surgical resection. Accurate prognostication of recurrence risk is essential to identify patients who may benefit from adjuvant therapies and improve their health outcomes. In our study, we propose a novel approach to CRC recurrence prognostication using multi-view deep learning. Our proposed approach, Fusion with Multi-view Attention (FMA), integrates static and longitudinal data from heterogeneous healthcare records, and learns complex interactions between data views to predict recurrence and time-to-recurrence. Our model achieves an AUROC score of 0.97, and precision, sensitivity and specificity scores of 0.80, 0.90 and 0.95 respectively, outperforming all-known published results based on the commonly-used CEA prognostic marker, as well as state-of-the-art CRC recurrence prognostication models. We show through a sensitivity analysis that incorporating multiple data views improves model performance significantly compared to using only a single view. We also show that our model accurately stratifies patients into risk groups that are associated with the actual 5-year recurrence-free survival, paving the way towards better identification of high-risk patients who may benefit from adjuvant therapies. Our proposed approach demonstrates the potential of multi-view modelling to push state-of-the-art in CRC recurrence prognostication and could contribute towards more personalised patient management and follow-up in the clinic.

Cite this Paper

BibTeX
@InProceedings{pmlr-v219-ho23a, title = {Multi-view Modelling of Longitudinal Health Data for Improved Prognostication of Colorectal Cancer Recurrence}, author = {Ho, Danliang and Motani, Mehul}, booktitle = {Proceedings of the 8th Machine Learning for Healthcare Conference}, pages = {265--284}, year = {2023}, editor = {Deshpande, Kaivalya and Fiterau, Madalina and Joshi, Shalmali and Lipton, Zachary and Ranganath, Rajesh and Urteaga, Iñigo and Yeung, Serene}, volume = {219}, series = {Proceedings of Machine Learning Research}, month = {11--12 Aug}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v219/ho23a/ho23a.pdf}, url = {https://proceedings.mlr.press/v219/ho23a.html}, abstract = {Colorectal cancer (CRC) is a leading cause of cancer-related deaths worldwide, with a high incidence of recurrence following surgical resection. Accurate prognostication of recurrence risk is essential to identify patients who may benefit from adjuvant therapies and improve their health outcomes. In our study, we propose a novel approach to CRC recurrence prognostication using multi-view deep learning. Our proposed approach, Fusion with Multi-view Attention (FMA), integrates static and longitudinal data from heterogeneous healthcare records, and learns complex interactions between data views to predict recurrence and time-to-recurrence. Our model achieves an AUROC score of 0.97, and precision, sensitivity and specificity scores of 0.80, 0.90 and 0.95 respectively, outperforming all-known published results based on the commonly-used CEA prognostic marker, as well as state-of-the-art CRC recurrence prognostication models. We show through a sensitivity analysis that incorporating multiple data views improves model performance significantly compared to using only a single view. We also show that our model accurately stratifies patients into risk groups that are associated with the actual 5-year recurrence-free survival, paving the way towards better identification of high-risk patients who may benefit from adjuvant therapies. Our proposed approach demonstrates the potential of multi-view modelling to push state-of-the-art in CRC recurrence prognostication and could contribute towards more personalised patient management and follow-up in the clinic.} }
Endnote
%0 Conference Paper %T Multi-view Modelling of Longitudinal Health Data for Improved Prognostication of Colorectal Cancer Recurrence %A Danliang Ho %A Mehul Motani %B Proceedings of the 8th Machine Learning for Healthcare Conference %C Proceedings of Machine Learning Research %D 2023 %E Kaivalya Deshpande %E Madalina Fiterau %E Shalmali Joshi %E Zachary Lipton %E Rajesh Ranganath %E Iñigo Urteaga %E Serene Yeung %F pmlr-v219-ho23a %I PMLR %P 265--284 %U https://proceedings.mlr.press/v219/ho23a.html %V 219 %X Colorectal cancer (CRC) is a leading cause of cancer-related deaths worldwide, with a high incidence of recurrence following surgical resection. Accurate prognostication of recurrence risk is essential to identify patients who may benefit from adjuvant therapies and improve their health outcomes. In our study, we propose a novel approach to CRC recurrence prognostication using multi-view deep learning. Our proposed approach, Fusion with Multi-view Attention (FMA), integrates static and longitudinal data from heterogeneous healthcare records, and learns complex interactions between data views to predict recurrence and time-to-recurrence. Our model achieves an AUROC score of 0.97, and precision, sensitivity and specificity scores of 0.80, 0.90 and 0.95 respectively, outperforming all-known published results based on the commonly-used CEA prognostic marker, as well as state-of-the-art CRC recurrence prognostication models. We show through a sensitivity analysis that incorporating multiple data views improves model performance significantly compared to using only a single view. We also show that our model accurately stratifies patients into risk groups that are associated with the actual 5-year recurrence-free survival, paving the way towards better identification of high-risk patients who may benefit from adjuvant therapies. Our proposed approach demonstrates the potential of multi-view modelling to push state-of-the-art in CRC recurrence prognostication and could contribute towards more personalised patient management and follow-up in the clinic.
APA
Ho, D. & Motani, M.. (2023). Multi-view Modelling of Longitudinal Health Data for Improved Prognostication of Colorectal Cancer Recurrence. Proceedings of the 8th Machine Learning for Healthcare Conference, in Proceedings of Machine Learning Research 219:265-284 Available from https://proceedings.mlr.press/v219/ho23a.html.

Related Material