APPLICATION OF COST-SENSITIVE CONVOLUTIONAL NEURAL NETWORK FOR PNEUMONIA DETECTION

Abstract

Pneumonia is a disease caused by a viral, bacterial, or fungal infection. In the diagnostic process of pneumonia, one approach is to use X-ray images. One of the existing problems is the lack of qualified and experienced medical personnel to recognize the X-ray images that have been taken. For this reason, an alternative is needed to detect pneumonia. Existing research shows that the use of convolutional neural networks can effectively detect pneumonia X-ray images. However, one of the problems is that this approach focuses a lot on accuracy without considering performance criteria such as sensitivity and specificity. To solve this problem, a cost-sensitive based approach has been proposed. In this study, a convolutional neural network-based model was created and trained using a cost-sensitive and non-cost sensitive approach. From the results obtained, it is seen that the model made still has a comparatively low level of accuracy. However, it is found that training with a cost-sensitive approach is able to improve performance on the specificity side, although at the expense of performance on the sensitivity side.