Computer Learns How to Diagnose Patients
Reported September 14, 2009
(Ivanhoe Newswire) -- Mayo Clinic
researchers say their new "teachable" software system mimics the human brain
and may help diagnose cardiac infections without an invasive exam.
Developers call their program ďartificial neural network" (ANN) because it
mimics the brain's cognitive function and reacts differently to situations
depending on its accumulated knowledge. In this case, the ANN underwent
three training sessions to learn how to evaluate numerous symptoms it would
ANN's focus is to diagnose endocarditis, an infection in the heart's valves
or chambers. Endocarditis is serious and deadly for patients with implanted
medical devices, killing as many as one in every five even with aggressive
treatment and device removal. Diagnosis is invasive and risky, inserting a
probe down the esophagus.
ANN tested 189 Mayo Clinic patients with
device-related endocarditis diagnosed between 1991 and 2003. The best
trained ANN correctly diagnosed 72 of 73 implant-related infections and 12
of 13 endocarditis cases.
Developers hope ANN will save patients from needless invasive procedures,
avoiding the discomfort, risks and costs.
SOURCE: Presented at the Interscience Conference on Antimicrobial Agents and
Chemotherapy (ICAAC) in San Francisco, September 12, 2009