DSSViT: Multi‐Scale Adaptive Fusion Vision Transformer With Dense Feature Reuse for Robust Pneumonia Detection in Chest Radiography DOI
Jinhui Huang

International Journal of Imaging Systems and Technology, Journal Year: 2025, Volume and Issue: 35(3)

Published: May 1, 2025

ABSTRACT Accurate pneumonia diagnosis using chest x‐rays (CXR) remains a critical challenge due to the need for precise extraction of fine‐grained local features and effective multi‐scale spatial pattern recognition. While Vision Transformer (ViT) models have demonstrated strong performance in medical imaging, they often struggle with these aspects, limiting their effectiveness clinical applications. This study proposes Dense‐SEA ViT (DSSViT), an enhanced architecture, address limitations by improving feature representation information capture detection. DSSViT integrates DenseNet121 as extractor enhance reuse improve flow, thereby compensating ViT's weakness capturing low‐level visual details. Additionally, Squeeze‐Excitation Adaptive Fusion (SEA) mechanism is introduced calibrate channel attention enable adaptive fusion, enhancing model's ability extract diagnostic while reducing noise interference. The proposed architecture was evaluated on X‐ray dataset classification. Experimental results demonstrate that achieves superior capability, leading test accuracy 97.69%, outperforming baseline such EfficientNet (93.90%) VGG19 (96.57%). These findings suggest promising approach automated settings.

Language: Английский

DSSViT: Multi‐Scale Adaptive Fusion Vision Transformer With Dense Feature Reuse for Robust Pneumonia Detection in Chest Radiography DOI
Jinhui Huang

International Journal of Imaging Systems and Technology, Journal Year: 2025, Volume and Issue: 35(3)

Published: May 1, 2025

ABSTRACT Accurate pneumonia diagnosis using chest x‐rays (CXR) remains a critical challenge due to the need for precise extraction of fine‐grained local features and effective multi‐scale spatial pattern recognition. While Vision Transformer (ViT) models have demonstrated strong performance in medical imaging, they often struggle with these aspects, limiting their effectiveness clinical applications. This study proposes Dense‐SEA ViT (DSSViT), an enhanced architecture, address limitations by improving feature representation information capture detection. DSSViT integrates DenseNet121 as extractor enhance reuse improve flow, thereby compensating ViT's weakness capturing low‐level visual details. Additionally, Squeeze‐Excitation Adaptive Fusion (SEA) mechanism is introduced calibrate channel attention enable adaptive fusion, enhancing model's ability extract diagnostic while reducing noise interference. The proposed architecture was evaluated on X‐ray dataset classification. Experimental results demonstrate that achieves superior capability, leading test accuracy 97.69%, outperforming baseline such EfficientNet (93.90%) VGG19 (96.57%). These findings suggest promising approach automated settings.

Language: Английский

Citations

0