Use of artificial intelligence for the preoperative diagnosis of pulmonary lesions

The relatively new field of artificial intelligence has spawned a variety of techniques associated with computer-assisted diagnosis. These techniques have been applied to the diagnosis of pulmonary lesions, but previous reports have focused on medical rather than surgical populations and the results have been evaluated using only retrospective patient surveys. We used a Bayesian algorithm to develop a diagnostic computer model for prospectively evaluating patients undergoing thoracotomy for suspected pulmonary malignancy. Patients who had a preoperative diagnosis were not included. Preoperative clinical and radiographic parameters for 100 consecutive patients were prospectively entered into the diagnostic model, which then categorized the lesion as benign or malignant. The computer predictions agreed with the final histological diagnosis in 95 of the 100 patients. The sensitivity was 96% and the specificity was 89% for this prospective series. These results indicate that the computer-assisted diagnosis of pulmonary lesions may have a role in this clinical setting.