Radiomics & Non-Invasive Tumour Characterisation
Imaging biomarkers and radiomics pipelines for treatment monitoring and response prediction.
Overview
Radiomics extracts quantitative features from medical images (texture, shape, intensity statistics) that reflect underlying tumour biology invisible to the naked eye. When combined with machine learning classifiers, these pipelines provide actionable clinical predictions from routine CT or MRI acquisitions.
Applications
- Early response assessment — predicting chemotherapy response in sarcomas after 1–2 cycles from CT texture changes
- Renal tumours — non-invasive histological subtype and grade prediction
- Gliomas — longitudinal monitoring under bevacizumab
Key methodological contributions
- Reproducible, open-source segmentation and feature-extraction pipelines
- Robustness analysis across scanners and acquisition protocols
- Integration of radiomics features with clinical covariates in survival models
- Multi-centric prospective validation studies