Differential diagnosis of Cervical Lymph Nodes in CT images using modified VGG-Net

Differential diagnosis of benign and malignant cervical lymph nodes (LNs) play an important role in the treatment planning of patients suffering from Head and Neck Cancer. Invasive pathological tests for detecting malignancy are painful and complex procedures. Computed tomography (CT) is a widely used and preferred non-invasive imaging modality for imaging assessment of all types of cancer related diseases. Manual assessment of CT scans is a time consuming and error- prone task. Hence, in this work authors have presented deep learning (DL) based computer-aided detection (CAD) approach for differential diagnosis of benign and malignant cervical LNs. In the proposed approach, VGG-Net is modified using Spatial Squeeze and Excitation (SSE), and residual concept without increasing the computation burden. Further, the inclusion of SSE block at various depths has been also studied. Moreover, a comparative study between the proposed and the state-of-art DL approaches is also presented. The achieved best results are sensitivity = 97.02%, specificity = 92%, accuracy = 96%, and area under curve = 94.22%.