Changing Operating Points of AI Software Improves Pediatric Chest X-Ray Analysis

Background: Artificial intelligence (AI) methods to interpret chest x-rays in adults are being rapidly developed and are demonstrating impressive potential for efficient and accurate analysis. However, it remains uncertain whether this adult-oriented methodology can be applied to correctly interpret pediatric chest x-rays. Objective: To determine whether adult-focused AI methods used to evaluate adult chest x-rays can be adapted for use in accurately interpreting pediatric chest x-rays. Methods: With approval of the institutional review board, chest x-rays on patients aged 15%. Different operating points were evaluated to determine the most optimal setting for the analysis of pediatric chest x-rays in order to detect abnormalities such as pneumothorax, consolidation, nodule, and pleural effusion. The interpretation of an experienced pediatric radiologist was considered the "ground truth" for comparison of the AI software results. The pediatric chest x-rays were divided into subgroups more...