UCL Discovery
UCL home » Library Services » Electronic resources » UCL Discovery

Using machine learning techniques on digitised histopathological images to distinguish lipoma (LP) from atypical lipomatous tumours (ALT)

Chai, Binghao; (2023) Using machine learning techniques on digitised histopathological images to distinguish lipoma (LP) from atypical lipomatous tumours (ALT). Doctoral thesis (Ph.D), UCL (University College London). Green open access

[thumbnail of Binghao Chai PhD Final Thesis.pdf]
Preview
Text
Binghao Chai PhD Final Thesis.pdf - Submitted Version

Download (9MB) | Preview

Abstract

Lipomas are benign neoplasms of fat and are amongst the most common tumours with an estimated incidence rate of 1 per 1000 people per year (Rydholm & Berg 1983, Weiss et al. 2007). They need to be distinguished from atypical lipomatous tumours (ALT) which are malignant and rare and present both a clinical and histological challenge. The distinction is made by assessing nuclear and subtle architectural features which require reviewing multiple sections and the use of ancillary genetic testing. Benign fatty tumours, therefore, present a considerable workload in the general pathology setting and the distinction from malignancy often require specialist review. This problem could be addressed by automated whole slide image (WSI) analysis, however, there is a lack of existing tools for this task. The paucicellular nature of the fatty tissues also presents a computational challenge which is addressed in this study. A tile-based deep learning workflow is developed to quantitatively analyse fatty tumour WSIs. A total of 206 internal and 402 external slides were employed for the training, validation and testing of the convolutional neural networks. A semi-automatic nucleus annotating workflow is also proposed, and a nucleus-level lipoma and ALT dataset is developed for the tasks of fatty tumour nucleus classification and segmentation. A lipoma and ALT nucleus detector is trained on top of this dataset. The tile-based workflow and the nucleus detector are then integrated for the slide-level classification of lipoma and ALT. The pipeline achieved 70.31% overall accuracy with an AUC of 0.777 for slide-level diagnostic classification using the real-world dataset. The results suggest that WSI analysis of fatty tumours is feasible but presents a unique set of challenges.

Type: Thesis (Doctoral)
Qualification: Ph.D
Title: Using machine learning techniques on digitised histopathological images to distinguish lipoma (LP) from atypical lipomatous tumours (ALT)
Open access status: An open access version is available from UCL Discovery
Language: English
Additional information: Copyright © The Author 2023. Original content in this thesis is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) Licence (https://creativecommons.org/licenses/by-nc/4.0/). Any third-party copyright material present remains the property of its respective owner(s) and is licensed under its existing terms. Access may initially be restricted at the author’s request.
UCL classification: UCL
UCL > Provost and Vice Provost Offices > UCL BEAMS
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Computer Science
URI: https://discovery.ucl.ac.uk/id/eprint/10170583
Downloads since deposit
33Downloads
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item