We’re excited to share our latest work published in Technology in Cancer Research & Treatment: “Deep Learning-Based Auto-Segmentation for Liver Yttrium-90 Selective Internal Radiation Therapy” — a collaboration between Jun Li, Rani Anne, and myself.
This study introduces a deep learning (DL) model built on the 3D U-Net architecture, developed to automatically segment the liver in CT scans for patients undergoing Y-90 Selective Internal Radiation Therapy (SIRT). Accurate liver segmentation is a critical step for calculating Y-90 dosage, traditionally done manually — a time-consuming and subjective process.
Our DL-based pipeline:
- Outperformed Atlas-based methods (DSC: 0.94 vs. 0.83)
- Achieved near-perfect agreement in dose calculation (RA ~1.00)
- Was deployed clinically using a seamless DICOM workflow
- Processed each case in under 2 minutes
This work demonstrates the clinical viability of AI-assisted planning in interventional radiology, particularly for liver-directed therapies.
Qualia Radiomics 