<![CDATA[Rhinosinusitis is an inflammatory disease affecting the mucosal lining of the nasal cavity (rhinitis) and paranasal sinuses (sinusitis), posing a significant public health challenge in Indonesia due to its high clinical and economic burden. This study aims to develop an advanced diagnostic method to assist healthcare professionals in accurately detecting sinusitis, thereby reducing diagnostic bias and improving patient outcomes. The proposed method involves a modified VGG16 architecture, where the traditional fully connected layers are replaced with Global Average Pooling (GAP) to minimize overfitting and computational complexity. By retaining the depth advantages of VGG16 while enhancing efficiency, this approach is tailored for medical image analysis. The dataset comprises 659 thermal images, evenly split between normal and sinusitis cases, which were preprocessed through cropping, global thresholding, and masking to improve feature extraction. The modified model incorporates additional convolutional layers (Conv3-4, Conv4-4, Conv5-4, and Conv5-5) to capture intricate spatial features, further boosting classification performance. Experimental results demonstrate an impressive accuracy of 95.67%, outperforming the standard VGG16 model, which achieved only 68.34% accuracy and exhibited overfitting. The study also highlights the effectiveness of thermal imaging as a non-invasive, cost-efficient alternative to conventional diagnostic methods like CT scans or MRIs.]]>
Copyrights © 2025
