At first glance, medicine and astrophysics seem worlds apart. Yet, technologies from astrophysics have repeatedly enhanced medical diagnostics. For instance, photonic algorithms from telescopes have refined breast cancer detection. Now, these fields unite again to tackle Alzheimer's disease.
A deep learning algorithm, originally developed in astrophysics to analyze vast telescope data interactions, has been adapted to patient records. It diagnoses Alzheimer's earlier and more effectively than general practitioners (GPs).
This innovation, born from collaboration between the University of Brighton, Sussex Medical School (BSMS), and astrophysicists at the University of Sussex, promises to transform early dementia detection in primary care. In the UK, only two-thirds of dementia patients receive a formal diagnosis, often too late in the disease progression.
Led by Elizabeth Ford, the research team analyzed GP records to identify 70 indicators of dementia onset, spanning the five years before diagnosis.
Partnering with astrophysicists, they tested various machine learning models to detect patterns in pre-diagnosis clinical data. The top-performing model flagged 70% of cases before GPs did, though it also produced some false positives.
Ford explains: "Patients show a diverse array of indicators before dementia diagnosis. GPs struggle to connect these dots, but a computer model can do so more efficiently. Early detection could profoundly improve patient care."
Related: How can astronomy help fight cancer?
"These results are promising but raise important questions for GPs and patients about integrating such technology into clinical practice," Ford adds. "As AI evolves, we must discuss our comfort with machines assessing risks for life-altering conditions like dementia."
Astrophysicist Seb Oliver notes: "Collaborating with Ford's team has been rewarding. Statistical methods like AI and deep learning uncover insights from complex data—be it space telescope images or patient records. Applying these in medicine highlights unique challenges and opportunities."
