What Researchers Did
Researchers used iterative consultation between two independent AI systems (Claude and Perplexity) to analyze complex clinical neuroimaging and patient data from a 63-year-old woman.
What They Found
Initial AI analyses of neuroimaging reports, cognitive testing, and clinical data diverged substantially (45-60 percentage-point difference in probability estimates). Through autonomous error detection and cross-validation over 5-7 iterative cycles, the systems converged to a consensus (<10 percentage-point difference). They discovered a 3.5% increase in total intracranial volume (indicating measurement artifact), an 11-month temporal gap between cognitive testing and MRI, and concluded modest real improvements (2-4%) embedded within measurement artifact (3-5%).
Canadian Relevance
This study has no direct Canadian connection.
Study Limitations
A significant limitation is that this study is based on a single case, which limits the generalizability of its findings.