Nature Communications: Longitudinal imaging improves treatment response prediction

Our team published a study in Nature Communications showing that longitudinal imaging can be combined with multi-task deep learning to improve treatment response prediction in rectal cancer.

The model jointly learns tumor segmentation and response prediction from paired pre-treatment and post-treatment MRI scans, enabling deeper use of temporal imaging information during therapy assessment.

Key points:

• Trained on 2,568 MRI scans from 321 rectal cancer patients.
• Achieved strong multi-institution validation performance for pathologic complete response prediction.
• Further improved accuracy when imaging features were integrated with blood-based tumor markers.

Paper: Predicting treatment response from longitudinal images using multi-task deep learning