Abstract: Pathology The diagnosis of pathology in a chest X-ray is generally complicated, even for experienced practitioners. A system that can automatically diagnose the findings in images obtained through X-rays of the chest can be useful in the medical examination of the patient as there is a shortage of experienced doctors. Classifying the chest X-ray is a multi-label classification task as a patient may have multiple diseases. In this research, we aim to develop an algorithm using deep learning techniques to identify the condition in the chest X-ray with high accuracy. In this research, we fine-tuned a pre-trained CNN architecture named DenseNet-121 to extract the features from the chest X-ray and to classify the extracted features into the pathology. The weights of the model are initially set with the weights of a model that is trained on ImageNet. Then the model is trained on a sample of the "ChestXray14" dataset.
Keywords: Chest X-ray; CNN; CAD; DenseNet.
|
DOI:
10.17148/IJARCCE.2022.11443
