Clinical Decision Support for Early Diagnosis of Cardiomegaly by Using Deep Learning Techniques on Chest X-rays
50th Computing in Cardiology, CinC 2023, Georgia, Amerika Birleşik Devletleri, 1 - 04 Ekim 2023, cilt.50, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Cilt numarası: 50
- Doi Numarası: 10.22489/cinc.2023.042
- Basıldığı Şehir: Georgia
- Basıldığı Ülke: Amerika Birleşik Devletleri
- Gazi Üniversitesi Adresli: Evet
Özet
Computer-aided diagnosis (CAD) systems have been widely researched and used in the medical field since their introduction in the 1960s. In the field of neural networks, deep learning methods such as convolutional neural networks (CNNs) have shown promising and impressive results in many areas, including image recognition. A promising and popular CNN model called VGGNet-16 was used in the study to diagnose a condition called cardiomegaly, where a patient suffers from an enlarged heart, using chest X-rays. To further improve these procedures, three different model was composed to create an ensemble model and we got a unique detection model. A dataset of X-rays labelled 'Cardiomegaly' and 'No findings' was used to train the model with different values of hyper parameters for each training session. The results showed that the learning rate and number of epochs had the greatest impact on the performance of the model and can therefore be considered as the most important hyper parameters. It was also found that a lower value of the learning rate generally resulted in higher performance compared to higher values and that the combination of a low learning rate and a low batch size is preferable to achieve a higher performance of the model. In conclusion, our new ensemble model got a comparable performance with respect to the literature.