Ensemble learning based Deep Learning Architecture for malignancy detection of mediastinal lymph nodes in CT images

The detection of malignancy in lymph nodes (LNs) in its early stages is important for appropriate treatment planning. The tumour growths are usually identified using invasive procedures like biopsy which are painful and complex in nature. Hence, in this study the authors have taken computed tomography images which are non-invasive in nature for detection of malignancy in LNs. For this purpose, the authors have proposed an ensemble learning based deep learning (DL) architecture. The proposed ensemble architecture comprises of various significant custom built DL models. The proposed architecture achieves accuracy – 98.56%, sensitivity – 99.09%, specificity - 98.04%, and AUC - 98.56%.