Accurate detection of prostate and kidney cancer relies on cross-sectional imaging, health records, and biomarkers, but current methods do not effectively utilise the vast amounts of available multimodal health data. This leads to over- and underdiagnoses, overtreatment, and missed diagnoses.
Multimodal AI models have the potential to improve diagnosis and therapy by finding latent relationships in complex data. However, challenges such as lack of access to data sources, lack of joint validation, multilingual patient representations, different privacy regulations, and lack of real-world validation hinder implementation into clinical practice and affect trust in AI.
COMFORT will develop trusted AI-empowered healthcare solutions by unlocking the full potential of data-based solutions in the fight against urologic cancers.
This will be achieved in three interconnected stages:
Stage 1 – Data Collection and Preparation
In the initial stage, the COMFORT teams will collect and annotate data and will deliver one of the largest and most comprehensive databases which is tailored specifically to the challenges of developing AI models for detecting and diagnosing urologic cancers.
Stage 2 – Development
During stage 2, we will develop reliable multimodal AI models and implement novel trans-modality data fusion techniques. This will lead to AI models that mimic physician decision pathways by processing information from multiple sources, producing accurate predictions of cancer types, progression, and prognosis. These innovative model architectures, designed in cooperation between physicians, computer scientists, and patients, will deeply influence future research.
Stage 3 – Deployment
Lastly, COMFORT will produce the first multi-national, prospective evaluation of AI models in a real-world clinical setting and will provide new insights into the usefulness, impact, and acceptance of AI in routine clinical practice.