Katmanlı sinir ağları kullanılarak göğüs radyografilerinde göğüs tüpü tespiti için bir yaklaşım

We propose an approach to train a Convolutional Neural Network (CNN) to detect chest tubes present on radiographs. To better detect the chest tube skeleton as the final output, non-uniform rational B-spline (NURBS) curves are used to automatically fit to the CNN output. This is the first study conducted to automatically detect artificial objects in the lung region of chest radiographs. Other automatic detection schemes work on the mediastinum. According to our initial tests, we decided to use the Gradient Descent and Cross-Entropy algorithms with a sigmoid activation function with 5 layers. Our final CNN architecture contains 2, 32, 32, 128, and 1 nodes for the successive layers. Between layers, there are 32, 16, 128, and 1 links for each node in the layers. After a series of tuning tests, the learning rate was selected as 0.1. We evaluated the performance of the model using a pixel-based ROC analysis. Each true positive, true negative, false positive and false negative pixel is counted and used for calculating average accuracy, sensitivity, and specificity percentages. The results were 99.99 % accuracy, 59 % sensitivity, and 99.99 % specificity.