Research — 2025

Skin-Lesion ClassificationNeural Architecture Search, Medical Imaging

A professor asked me to look at a skin-cancer classification project that had stopped moving. The model wasn't broken — it was being optimised for the wrong number.

Role
Lead — inherited and rebuilt the pipeline
Client
Research · University of East London
Year
2025
Discipline
Research

The problem

The pipeline was tuned for overall accuracy on 27,000+ ISIC dermoscopy images. On a heavily imbalanced dataset that is close to meaningless: a model can look excellent while missing most malignant lesions. Recall on the class that actually matters sat at 49%. In clinical screening, a false negative is the expensive error.

The approach

I rebuilt the pipeline around recall rather than accuracy. NASNetMobile transfer learning with Keras Tuner architecture search, focal loss to stop the majority class dominating the gradient, class-weighted training, and stratified augmentation so validation splits kept the real class distribution instead of flattering the metrics.

The outcome

Recall on malignant lesions went from a 49% baseline to 77.7%, at 0.86 AUC across 27,000+ images — a model that is now useful as a screening aid rather than a well-scored classifier that misses half the cases.

Results

By the numbers.

Measured, not estimated
49% → 77.7%
Malignant-lesion recall
0.86
AUC
27k+
Dermoscopy images
Stack

What it's built on.

6 components
Python TensorFlow/Keras NASNetMobile Keras Tuner focal loss ISIC dataset
Contact

Let's build
something honest.