Jatin Jassi
Technical · AI / ML

Sparse Random Projection for Medical Image Analysis

A feature-extraction pipeline that compresses high-dimensional medical imaging while keeping diagnostic signal.

Python / ML / DL

Overview

Built a feature-extraction pipeline using Sparse Random Projection to compress three-dimensional medical imaging into interpretable two-dimensional slices, reducing dimensionality by 65% while keeping diagnostically important features. Engineered OpenCV preprocessing, including automated cropping and noise reduction, to standardise quality across more than 10,000 heterogeneous images.

Combined global features (pixel-intensity histograms), local features (texture gradients) and pattern-based features (pixel-difference matrices) to improve interpretability for clinician review. The pipeline scales to full 3D volumetric analysis with under 10% code change, and its sparse matrix operations made feature extraction 35% faster than a PCA equivalent.

Contact

Full-time unrestricted working rights, Subclass 485 Temporary Graduate Visa, valid until January 2028.