Pneumonia Detection with Different Classifiers in the Hybrid Model Approach

Kılıç R., Sağlam H. K., Oral E. A., Özbek İ. Y.

4th International Conference on Advanced Engineering Technologies Ͷ , Bayburt, Türkiye, 28 - 30 Eylül 2022, ss.170-176

Yayın Türü: Bildiri / Tam Metin Bildiri
Basıldığı Şehir: Bayburt
Basıldığı Ülke: Türkiye
Sayfa Sayıları: ss.170-176
Atatürk Üniversitesi Adresli: Evet

Özet

Pneumonia is one of the dangerous diseases caused by bacteria or viruses, whose early diagnosis is important. Various methods are used to diagnose pneumonia: blood culture, sputum culture, fluid sample, bronchoscopy, pulse oximetry, and chest Xray. Chest X-ray is one of the most widely used methods of detecting pneumonia. Assisted diagnosis systems are being created for automatic pneumonia detection using chest X-ray images. In this study, the hybrid model approach in pneumonia detection is compared with different classifiers. For this purpose, X-Ray data set from pneumonia and non-pneumonia chest X-ray images was used. By taking the features from the penultimate fully connected layer of Alexnet, VGG16 and VGG19, which are transfer learning methods, classification is made with Naive Bayes, Decision tree, KNN and SVM classifier. In the classification made with the VGG16+DVM hybrid model, the accuracy was calculated as 97.61%, the sensitivity as 98.48%, the selectivity as 95.27% and the F1 score as 98.36%.