Jeffrey L. Sponsler
Expert system, Artificial intelligence, Neurology, Stroke
Decision support for stroke diagnosis is important given the complexity of diagnosis. The project goal was to extend the prototype StrokeDx program with new diagnoses including multiple sclerosis, Brown-Sequard, Weber, Millard-Gubler, and thalamic stroke. Benchmark data files were created to contain symptoms/signs for each new syndrome. Rules were encoded for new diagnoses. StrokeDx employs logic programming to compute a confidence factor for a diagnosis. Each diagnostic rule base was applied to all benchmark datasets. Previous diagnoses included frontal stroke, occipital stroke, Wallenberg syndrome, CADASIL and radial neuropathy. The sensitivity of each diagnostic rule set (for the corresponding benchmark) was 100%. Total diagnosis count is currently 10. The StrokeDx development toolset is extensible and when applied to diagnostic benchmarks is accurate.
Important Links:
Go Back