AI Algorithm Shows Promise Detecting Chest X-ray Abnormalities But Lacks Accuracy for Key Diagnosis
-
Radiology is an important medical imaging field, but there is a shortage of radiologists to interpret scans. Artificial intelligence (AI) tools may help address this need.
-
An AI algorithm called ChestEye was validated on 278 chest x-rays from a primary care center, comparing its diagnoses to a radiologist's.
-
Overall accuracy was high at 0.95, but sensitivity for detecting specific conditions was lower at 0.48, indicating the algorithm may miss some abnormalities.
-
The algorithm had good specificity at 0.98, accurately identifying normal images without abnormalities. This could help triage and prioritize abnormal scans.
-
Additional real-world training is needed to improve ChestEye's sensitivity for primary care conditions like bronchial wall thickening and chronic lung abnormalities. External validation is crucial before implementing AI tools.