Siamese neural network distinguishes pneumonia, TB, and aspergillosis on chest X-rays with 98.72% accuracy

A novel Siamese CNN distinguished aspergillosis, pneumonia, tuberculosis, and normal lungs on chest X-rays with 98.72% accuracy, outperforming standard CNN and ResNet-50 in a dataset of 7,200 images. A web interface supports clinical upload.

• The Siamese CNN achieved 98.72% accuracy for four-class classification (aspergillosis, pneumonia, tuberculosis, normal) on chest X-rays.
• The dataset comprised 7,200 images; various training-testing ratios were used, but no external validation was reported.
• An intuitive web interface was developed to allow healthcare providers to upload and analyze medical images.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.