My PhD focused on CT-based habitat imaging to quantify intratumor heterogeneity in colorectal liver metastases: whether spatial patterns within a tumor can predict treatment response and survival in patients receiving anti-angiogenic therapy.
The thesis introduces a reproducible, biologically grounded approach to CT habitat computation. Key findings: handcrafted radiomics features outperform deep learning for habitat computation, and rim entropy (a measure of heterogeneity at the tumor boundary) predicts bevacizumab response in metastatic colorectal cancer.