Dissecting first-episode psychosis heterogeneity with clustering analyses: A systematic review.
Luca Sperti, Alessandro Pigoni, Guido Nosari et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review examined the use of unsupervised machine learning clustering techniques in patients with first-episode psychosis (FEP). It analyzed 48 studies and found that while clustering based on cognitive, functional, immune, genetic, clinical, and imaging dimensions is promising, there is no consensus on cluster characteristics. A notable finding was the identification of a subgroup with greater impairment at onset.
Key findings
- 48 studies utilized unsupervised ML techniques to cluster FEP patients based on various dimensions.
- A two-cluster classification was suggested, with one subgroup showing higher inflammatory states or brain damage.
- A subgroup with greater cognitive, functional, or clinical impairment was consistently identified at the onset of psychosis.
Why it matters
Understanding the heterogeneity of first-episode psychosis through clustering can enhance diagnostic accuracy and inform tailored treatment strategies, potentially improving patient outcomes.
What to keep in mind
The review highlights a lack of reproducibility in clustering results across studies.