Barriers and Challenges in Clinical Application of Machine Learning Algorithms

According to the World Health Organization’s 2022 report, pneumonia is the primary cause of death for children, with over five million under-five deaths, while Nigeria, India, Pakistan, the Democratic Republic of the Congo, and Ethiopia account for half of these deaths. Currently, this disease is diagnosed manually by a radiologist using a chest x-ray image. However, through this method, the identification of pneumonia is difficult due to the overlapping features of various conditions of the lower respiratory tract. Recently, there have been plenty of machine learning models developed to diagnose pneumonia. Nevertheless, none of these models were used for support in the diagnosis of pneumonia in a healthcare setting. By collecting chest CXR images from local healthcare facilities as well as global sources and creating a CNN model, the main barriers and challenges associated with machine learning algorithms in clinical applications were identified in this work. The key barriers and challenges were identified as data, algorithms and clinical utilities, strategy and financing, and technology. The association between clinical diagnosis and machine learning algorithms was also investigated, including the radiologist’s attitude toward machine learning, performance analysis on pneumonia detection, limitations of clinical diagnosis, limitations of machine learning diagnosis, and machine learning-aided diagnosis. Finally, a clinical diagnosis of pneumonia supported by machine learning algorithms is recommended as the only viable option for accurate treatment.