Body composition derived from routine computed tomography imaging to improve chemotherapy dosing, toxicity estimation, and treatment response assessment in patients with gastrointestinal and pancreatic cancer: a proof-of-concept study

• Applicants: Professor Dr. Heiko Becker; Professor Dr. Michael Quante; Professor Dr. Jakob Weiss
• Subject Area: Radiology
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 584977163

Systemic therapies such as chemotherapy are a central part of cancer treatment, but they often cause serious side effects. These toxicities can force dose reductions or treatment discontinuation, even when the therapy itself is effective. Currently, chemotherapy doses are usually calculated using body surface area (BSA), which is based only on a patient’s height and weight. However, this method does not reliably reflect how individual patients process and tolerate cancer drugs, leading to a risk of overdosing some patients and underdosing others. Recent research shows that body composition especially the amount of skeletal muscle and different types of body fat provides more meaningful information for personalizing cancer treatment. Low muscle mass, in particular, has been linked to higher chemotherapy toxicity, more frequent dose reductions, and worse survival. Importantly, BSA does not correlate well with muscle mass, and patients with the same BSA can have very different body compositions. The influence of other body compartments, such as visceral and subcutaneous fat, on drug distribution and clearance is still not well understood. Although body composition can be measured on routine CT or MRI scans that many cancer patients already undergo, these measurements are not used in daily clinical practice due to time and technical limitations. As a result, valuable prognostic and predictive information remains largely unused. Recent advances in artificial intelligence now make it possible to automatically and efficiently extract body composition data from standard medical imaging. These AI-based methods offer a low-cost and scalable solution that could be seamlessly integrated into clinical workflows. With this project, we aim to close an important gap in personalized cancer therapy. We will use innovative AI techniques to automatically quantify body composition from routine imaging and investigate whether these measures can better predict chemotherapy-related toxicity than current clinical standards. By linking body composition to treatment side effects, dose adjustments, and patient survival, the study seeks to develop a more precise and individualized approach to chemotherapy dosing, ultimately improving treatment tolerability and patient outcomes.

• DFG Programme: Research Grants
• Co-Investigator: Dr. Marco Reisert