Organoids Segmentation using Self-Supervised Learning: How Complex Should the Pretext Task Be?

Woude, Bart van der (2023) Organoids Segmentation using Self-Supervised Learning: How Complex Should the Pretext Task Be? Bachelor's Thesis, Artificial Intelligence.

Preview	Text Thesis___Organoids_segmentation_using_SSL__how_complex_should_the_pretext_task_be_-34.pdf Download (11MB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (129kB)

Abstract

Deep learning methods are well-suited for biomedical image analysis, this research focuses specifically on progenitor liver organoid segmentation. Most popular supervised-learning approaches, however, require large annotated data sets that are time-consuming and costly to create. Self-supervised learning (SSL) has proven to be a viable method for increasing downstream performance, through pre-training models on a pretext task. However, the literature is not conclusive on how to choose the best pretext task. This research sheds light on how the complexity of the pretext task affects organoid segmentation performance, in addition to understanding whether a self-prediction or innate relationship SSL strategy is best suited for organoid segmentation. Eight novel self-prediction distortion methods were implemented, creating a number of simple and complex pretext tasks. Two well-known innate relationship pretext tasks, Jigsaw and Predict rotation, were implemented in order to compare strategies. Results showed that complexity of the pretext tasks do not correlate with segmentation performance. However, complex models (average F1-score = 0.862) consistently, albeit with a small effect size, outperform simple tasks (average F1-score = 0.848) possibly due to acquiring a wider variety of learned features after pretext learning despite not being necessarily more complex. The small effect size and high standard deviations of F1-score segmentation performances make the results non-conclus

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Schomaker, L.R.B. and Haja, A.
Degree programme:	Artificial Intelligence
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	04 Sep 2023 13:26
Last Modified:	04 Sep 2023 13:26
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/31365

Actions (login required)

View Item