Evaluating Machine Learning Tools as Clinical Decision Support in Chest X-ray

Abstract

Chest X-rays are one of the most frequent imaging modalities and are consistently in demand within the field of radiology. With the rise of innovation in AI, this study evaluated the machine learning tool's performance as a clinical decision support tool to help reduce the radiologists' workload and improve clinical workflow. The machine learning focused on diagnosing the diseases cardiomegaly, pneumonia, and mass. A total of 56 chest X-ray radiographs were analyzed by the machine learning tool and were compared with those diagnoses made by a radiologist as the diagnostic gold standard. Additionally, the acceptability of the tool was evaluated using a Likert scale focusing on its functionality, reliability, usability, efficiency, and security. The results showed that the MLT demonstrated a good performance in detecting pneumonia, but had poor accuracy in detecting cardiomegaly and mass cases. Additionally, the acceptability survey tool showed an overall neutral rating from the radiologist. While the machine learning tool shows potential as a support, it is still unreliable and inaccurate for clinical use. This suggests the need for further improvement of the MLT in its algorithm design and training process to enhance its diagnostic accuracy and reliability.