Clinical outcomes modelling
Current focusWe develop and validate predictive models of treatment response and normal tissue toxicity in patients treated with proton therapy. By integrating dosimetric data, quantitative imaging biomarkers, and clinical follow-up, we aim to understand the biological mechanisms underlying treatment outcomes and support therapeutic decisions. Current work focuses on statistical modelling of treatment response, integrating radiomic feature extraction, habitat imaging, and mixed-effects models applied to longitudinal patient data for both tumour response and normal tissue complications.
Research topics
Imaging biomarkers
We extract quantitative features from CT and MR imaging to characterise tissue properties beyond visual inspection. Radiomic pipelines capture texture, shape, and intensity descriptors from tumour volumes, while habitat imaging uses multiparametric MRI to map spatially distinct sub-regions with different biological behaviours — identifying areas of potential radio-resistance or early response.
Dosimetric analysis
Dose-volume metrics and spatial dose distributions are integrated with clinical and imaging data to understand the relationship between delivered dose and treatment outcomes. We work with both physical dose and biologically weighted quantities relevant to proton therapy, including LET-weighted dose and RBE-adjusted distributions.
Statistical modelling of outcomes
Imaging and dosimetric inputs feed into statistical models of treatment response and normal tissue toxicity. We use NTCP models calibrated on clinical data, survival analysis frameworks including Cox regression for time-to-event endpoints, and generalised linear mixed-effects models (GLMM) for longitudinal patient data — accounting for within-patient correlation across repeated assessments.
Funding
Modeling of proton RBE variations in patients
In radio-oncology, predictive algorithms are becoming established for estimating cancer cure rates or risks of complications. A key aspect of accurately predicting treatment outcome is understanding the biological effects that radiation has on cancer cells and normal tissues. As proton therapy becomes an accessible option for patients in hospital settings, the progressive availability of clinical follow-up data makes the engineering of predictive models a key research field. This project focuses on specific aspects of validation and refinement of predictive technologies applied to proton therapy, while improving our knowledge of RBE in patients.
Cancer radiomics targeted by pencil beam scanning proton therapy for deformable tumours
Tumours located in thoracic and abdominal sites present unique challenges for radiotherapy because of their anatomical complexity, biological heterogeneity, and non-stationary motion dynamics. The aim of this project was to improve the quality of radiation therapy by precisely delivering the dose to radio-resistant tumour locations, using personalised proton therapy that integrates radiomics, tumour geometry, and motion — an early example of combining imaging biomarkers with treatment planning in this setting.
Public outreach
DeepHealth Movement — Engineering the Future of Health
Our work on pencil beam scanning proton therapy and personalised treatment planning was featured in the DeepHealth Movement exhibition, organised by the Personalized Health and Related Technologies (PHRT) initiative of the ETH Domain. The exhibition presents cutting-edge health research to the general public, translating scientific concepts into accessible visual narratives. Our contribution — Hypoxia Guided Proton Therapy — illustrates how functional imaging and biological modelling combine to personalise radiation dose for individual patients.
The exhibition was presented at the Swiss Pavilion, Expo 2025 Osaka, and is touring Switzerland including the Stapferhaus and Empa. Photos by Adrian Notz.
Hypoxia-guided proton therapy for NSCLC — ResearchPod podcast
A science communication episode by ResearchPod covering our work on hypoxia-guided dose escalation with proton therapy in advanced-stage non-small cell lung cancer. The episode explains how PET imaging and radiobiological modelling can be combined to personalise treatment and improve outcomes while reducing toxicity.
Listen on ResearchPod →Publications
NTCP modeling for high-grade temporal radionecroses in a large cohort of patients receiving pencil beam scanning proton therapy for skull base and head and neck tumors
The impact of organ motion and the appliance of mitigation strategies on the effectiveness of hypoxia-guided proton therapy for non-small cell lung cancer
Assessment of radiation-induced optic neuropathy in a multi-institutional cohort of chordoma and chondrosarcoma patients treated with proton therapy
Combining clinical and dosimetric features in a PBS proton therapy cohort to develop a NTCP model for radiation-induced optic neuropathy
Investigating the potential of proton therapy for hypoxia-targeted dose escalation in non-small cell lung cancer