Master’s thesis · Completed

Predicting mucus plugs from lung CT scans

For my master’s thesis, I built a pipeline from raw lung CT images to estimates of mucus plug burden. I compared CNN models using fivefold cross-validation, evaluated predictions at patient level, and examined data leakage, uncertainty and generalisation. The thesis received an A. A paper based on the thesis is being prepared for submission to a suitable conference.

  • Python
  • PyTorch
  • Computer vision
  • Research
  • Cross-validation
  • Data pipelines
  • Model evaluation
AI-generated illustration inspired by lung CT imaging
AI-generated illustration of Lung CT images

From images to estimates

My work covered the pipeline from raw lung CT images to estimates of mucus plug burden, including model training and comparison.

Evaluation

I used fivefold cross-validation and evaluated predictions at patient level. The research examined data leakage, uncertainty and generalisation.

Research outcome

The completed thesis received an A. The card image is an AI-generated illustration, not a patient scan or a model result.

Publication in progress

A paper based on the thesis is being prepared for submission to a suitable conference.