This project addresses the clinical challenges of diagnosing Heart Failure with Preserved Ejection Fraction (HFpEF), a complex syndrome affecting up to 50% of heart failure patients. HFpEF is associated with risk factors and comorbidities, often linked to inflammation and immune degeneration, yet lacks a definitive diagnostic marker. Current diagnosis relies on manually calculated probabilistic scores combining clinical, laboratory, and echocardiographic data, a process requiring specialized expertise and limiting timely diagnosis.
The project leverages advanced machine-learning algorithms to develop diagnostic tools that assist clinicians in recognizing HFpEF. Using patient data collected in clinical settings, the approach employs Natural Language Processing (NLP) models to extract and transform unstructured information from electronic health records. These models are trained to identify relevant keywords and concepts, making the data accessible for quantitative analysis.
This innovative approach aims to overcome current diagnostic limitations, enabling faster and more accurate patient management. By reducing the reliance on expert manual scoring, these tools promote early and personalized diagnoses, significantly enhancing the quality of care. The project’s outcomes promise to alleviate the burden on clinicians and improve patient outcomes through precision diagnostics and tailored treatment strategies.