FPGA-Based Pulmonary Disease Detection Using Lightweight CNN

Artificial intelligence plays a significant yet challenging role in the automatic detection of pulmonary diseases, aiding radiologists in early diagnosis and treatment initiation. This letter presents a lightweight convolutional neural network architecture for pulmonary disease detection. The proposed model achieved an accuracy of 97.72% for lung cancer detection and 98.19% for tuberculosis–pneumonia detection, with only 1779 parameters. The field programmable gate array implementation of the proposed model outperformed earlier designs by delivering a throughput of 2109.7 inferences per second at 250 MHz. In addition, the proposed design offers better computational efficiency with minimal power consumption and optimized resource utilization.