Fig. 2: Workflow for HRD tumor classification and treatment response prediction using CT image. | npj Precision Oncology

Fig. 2: Workflow for HRD tumor classification and treatment response prediction using CT image.

From: Predicting homologous recombination deficiency and treatment responses using a computed tomography-based foundation model: a preclinical study

Fig. 2: Workflow for HRD tumor classification and treatment response prediction using CT image.

a Study design for animal experiment: mice bearing isogenic xenografts (HRD/HRP) were subjected to CP-506 (i.e., hypoxia-activated DNA-crosslinking agent) and control (vehicle) treatments. In the main experiment, CT scans were acquired before and after treatment, and DNA damage intensity was quantified in tumor tissues using γ-H2AX immunohistochemical staining 48 h post-treatment. In a parallel experiment, tumor volumes were measured regularly until tumors quadrupled in volume (event-free survival, EFS)51. b Radiomic pipeline for HRD classification and treatment response prediction: A pre-trained foundation model (FM) was employed to derive high-dimensional features from the CT scans. A shallow classifier was utilized to determine HRD tumor status. Subsequently, HRD-related features extracted from the final layer of the classifier were used to predict treatment responses (i.e., DNA damage and EFS).

Back to article page