Skip to content
SDGs Center UDINUS

Evidence record · UDINUS · SDGs

The Impact of Squeeze-and-Excitation Blocks on CNN Models and Transfer Learning for Pneumonia Classification Using Chest X-ray Images

Rektorat UDINUS 2025 Published 2026-10-30

Evidence record · UDINUS · SDGs

Unit
Rektorat UDINUS
Period
2025
Goals
Record
7f5cf7c5-c447-4098-a671-8efa21eb3cfe
Published
2026-10-30

Universitas Dian Nuswantoro conducted research evaluating Squeeze-and-Excitation (SE) Block integration in deep learning models for pneumonia classification using chest X-ray images. The activity involved UDINUS researchers Yunan, Muhammad; Marjuni, Aris; Affandy; Soeleman, Mochamad Arief; and Firdaus, Iqbal. The research addressed pneumonia as one of the leading causes of death due to respiratory tract infections, especially among children and the elderly. It also considered the importance of early detection because manual interpretation of chest X-ray images is often subjective and error-prone.

The activity evaluated a custom Convolutional Neural Network (CNN) model and three transfer learning architectures: MobileNetV2, VGG16, and InceptionV3. The dataset contained 5,856 images taken from the Chest X-ray Images (Pneumonia) dataset on Kaggle. The images underwent preprocessing, undersampling, and augmentation. Each model was tested in two configurations, without an SE Block and with an SE Block. Evaluation used accuracy, precision, recall, F1-score, and test loss metrics.

The findings showed that SE Block integration improved the accuracy of most models. Custom CNN accuracy increased from 95.17% to 95.88%, MobileNetV2 accuracy increased from 97.18% to 97.59%, and VGG16 accuracy increased from 96.88% to 97.69%. InceptionV3 accuracy also increased from 94.06% to 94.16%, although its test loss increased. The research described SE Block as strengthening the model's emphasis on important features through an inter-channel recalibration mechanism, especially in MobileNetV2 and VGG16.

This UDINUS research relates to the stated health-related goal by examining pneumonia classification and early detection methods. Its findings support development of a more accurate, efficient, and adaptive deep learning-based pneumonia diagnosis system, including implementation in healthcare facilities with limited resources.

Reference https://www.scopus.com/pages/publications/105027907609?origin=resultslist