View Proposal

---

Proposer

Marta Vallejo

Title

AI-Powered Registration of Cellular ALS Images using Pretrained and Contrastive Learning Models

Goal

You’ll apply deep learning to ALS images in collaboration with UK and Spanish teams, aiming for a journal publication.

Description

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease characterised by the progressive loss of motor neurons, leading to muscle weakness, paralysis, and ultimately respiratory failure. Most patients die within two to four years of symptom onset, though the clinical course varies widely. In addition to motor impairment, cognitive deficits, often associated with TDP-43 proteinopathy, are increasingly recognised as a core feature of the disease. Despite its severity, early diagnosis remains difficult due to clinical heterogeneity and the absence of definitive biomarkers. This project will utilise a new dataset of high-resolution cellular images provided by the Spanish National Research Council (CSIC). These images display multiple cells in grey, with nuclei stained in blue using DAPI. Currently, cell and nucleus contours are manually delineated, a labour-intensive and time-consuming process. The project aims to automate this annotation pipeline using semi-supervised learning and transfer learning techniques, thereby reducing reliance on extensive manual labelling. Specifically, the student will explore consistency regularisation methods (e.g. Mean Teacher, FixMatch), self-training strategies, and contrastive pretraining approaches (e.g. SimCLR, MoCo) on unlabelled data to enhance segmentation performance. Pre-trained encoders from biomedical models, such as BioViT, Cellpose, or UNet++ with ImageNet or Cellpose weights, will be evaluated for model initialisation. The ultimate goal is to accelerate cell and nucleus segmentation while improving the reliability and scalability of image-based analysis pipelines for ALS pathology research.

Resources

A real clinical dataset with some annotations

Background

Interest in the topic and techniques. Good programming skills. Willingness to learn. Interest in healthcare.

Url

Difficulty Level

Variable

Ethical Approval

Full

Number Of Students

2

Supervisor

Marta Vallejo

Keywords

cellular imaging, image segmentation, semi-supervised learning, deep learning, contrastive learning, transfer learning, cell registration, biomedical image analysis

Degrees

Bachelor of Science in Computer Science
Master of Science in Artificial Intelligence
Master of Science in Data Science
Master of Science in Software Engineering
Bachelor of Science in Computing Science
Bachelor of Science in Statistical Data Science
BSc Data Sciences