EMS-Resnet-DA Model With Perception Attention Mechanism for Effective Tissue Classification in Emphysema Diagnosis

Abstract The diagnosis of emphysema is generally done by processing the anatomical data from conventional CT scan image patterns. Even though various analysis methods were in practice to diagnosis this chronic disease, the pattern classification techniques and modeling of emphysematous tissues from CT images were reported with effective results. Multiscale residual network with data augmentation model (MS-ResNet-DA) followed by enhanced MS-ResNet-DA (EMS-ResNet-DA) model classify the Emphysema much effectively but both the models fails in multiscale classification of CLE and PLE. To overcome the mentioned limitation of classifying CLE and PLE effectively the EMS-ResNet-DA model with Perception Attention Mechanism (PAM) is proposed in this study. Parallel CNN architecture with LSTM is utilized for feature extraction of images with detailed information. The Perception Attention Mechanism is used to merge the two proposed neural network topologies. The results obtained are compared with the previous methodologies and proved for its effectiveness.