Automatic Detection of Bacilli Bacteria from Ziehl-Neelsen Sputum Smear Images

Manual bacilli detection from Zeihl-Neelsen (ZN) stain images is tedious results in error due to bacteria size and lack of trained experts. Bacilli detection is often a complex task due to their numbers and stain particles. Automatic detection models are the best solution to increase the accuracy of bacilli detection. In the proposed work bacilli detection model using Deep Convolution Neural Network (CNN) is proposed. Preprocessing and segmentation are also explored in the present study. A model such as VGG16, ResNet50, and SqueezeNet are explored. A comparison study is carried to analyze the performance metrics. A proposed model using SqueezeNet as a classifier gives an overall accuracy of 97%.