1. Background
Anhedonia, the reduced capacity to experience pleasure, is a core feature across depression, psychosis and bipolar disorder, and is closely linked to poorer recovery and quality of life. Yet clinicians still lack reliable tools to identify which patients show biologically distinct forms of anhedonia or to forecast who will benefit from particular treatments. This project leverages de-identified NHS clinical records at King’s (CRIS), linked where available to structural brain scans and routine blood tests, to study anhedonia across diagnostic boundaries in real-world care.
2. Novelty & Importance
Most studies focus on a single data type in a single disorder. Our work takes a more holistic view by carefully integrating clinical trajectories, brain structure information and commonly collected blood measures. By combining these complementary signals with modern machine-learning methods, we aim to uncover reproducible patterns that are clinically meaningful without disclosing implementation details that could be copied out of context. The goal is not a “black box,” but practical, transparent insights that can support earlier, more personalised decisions—ultimately improving outcomes while using data already available in routine services.
3. Aims & Objectives
(1) Derive data-driven subtypes of anhedonia across diagnoses using linked clinical, imaging and blood information, with rigorous quality control and privacy safeguards.
(2) Test whether these subtypes help predict important outcomes (e.g., symptom change, relapse risk, functioning) beyond standard clinical assessments.
(3) Examine whether patterns are stable over time and robust across different clinical settings, including independent, ethically approved datasets outside our institution.
(4) Produce clear, clinician-facing summaries (e.g., risk/subtype reports) and governance-compliant resources that enable responsible evaluation in healthcare environments.