Using Convolutional Neural Networks for the Classification of Suboptimal Chest Radiographs.

The article discusses the use of DenseNet121 and YOLOv8 neural networks in classifying suboptimal chest X-rays to minimize delays in patient care and diagnosis. The study included 3,587 patients and utilized a dataset of 10,000 chest X-rays for training and validation. Both AI models showed strong capabilities in differentiating between optimal and suboptimal X-rays, with DenseNet121 achieving an AUROC of 0.97 and YOLOv8 scoring 0.95. The research suggests that these models could provide feedback to radiographers to enhance image quality, although the alignment between radiologists and models varied due to the lack of clinical indications. [Extracted from the article]