1: \begin{abstract}

2: Early detection of lung cancer is an effective way to improve the cure rate of patients. Therefore, it is crucial to accurately detect lung nodules in computed tomography (CT) images for early diagnosis of lung cancer. However, developing an accurate computer-aided detection (CADe) system for lung nodule detection is a challenging task owing to the complex lung's structure and the heterogeneity of the lung nodules. In this study, we propose a lung nodule detection scheme motivated by the clinical workflow of radiologists, which combines CT scan sub-volume with bidirectional maximum intensity projection (MIP) images in a single architecture. For this purpose, we designed a novel multi-encoder based network (MEDS-Net) that exploits self-distillation to effectively learn from three different types of inputs, including 3D sub-volume, forward, and backward MIP images. MEDS-Net exploits three different encoders to effectively extract the distinctive insights which are passed to the decoder block. Further, the decoder block leverages the self-distillation mechanism by connecting the distillation block, which contains four auxiliary lung nodule detectors. It helps to expedite the convergence and improves the learning ability of the proposed architecture. Finally, the proposed scheme reduces the false positives by complementing the main detector with auxiliary detectors. The proposed scheme has been rigorously evaluated on 888 scans of the LIDC-IDRI dataset and obtained a competition performance metric (CPM) score of 93.6\%. The results demonstrate that incorporating bidirectional MIP images enables MEDS-Net to accurately detect lung nodules and reduce the number of false positives using auxiliary detectors.

3: \end{abstract}

4: