MEDS-Net: Self-Distilled Multi-Encoders Network with Bi-Direction Maximum Intensity projections for Lung Nodule Detection
Early detection of lung cancer is an effective way to improve the survival rate of patients. It is a critical step to have accurate detection of lung nodules in computed tomography (CT) images for the diagnosis of lung cancer. However, due to the heterogeneity of the lung nodules and the complexity of the surrounding environment, it is a challenge to develop a robust nodule detection method. Numerous efforts have been made to develop an efficient Computer-aided detection (CADe) systems, albeit none compliance with the routine workflow of radiologists which limits the adaptability of such CADe system. To overcome this deficiency, in this study we propose a lung nodule detection scheme which fully incorporates the clinic workflow of radiologists. Particularly, we exploit Bi-Directional Maximum intensity projection (MIP) images of various thicknesses (i.e., 3, 5 and 10mm) along with 3D patch of CT scan, consisting of 10 adjacent slices to feed into self distillation based Multi-Encoders Network (MEDS-Net). The propose architecture first condense 3D patch input to three channels by using dense block which consists of dense units which effectively examines the nodule presence from 2D axial slices. This condensed information, along with the forward and backward MIP images is fed to three different encoders to learn the most meaningful representation which is forwarded into decoded block at various levels. At decoder block, we employ self distillation mechanism by connecting the distillation block which contains five lung nodule detectors. It helps to expedite the convergence and improves the learning ability of proposed architecture. Finally, the propose scheme reduces the false positives by complementing the main detector with auxiliary detectors. The proposed scheme has been rigorously evaluated on 888 scans of LUNA16 dataset and obtained a CPM score of 93.6%. The results demonstrate that incorporating of bi-direction MIP images enables MEDS-Net to effectively distinguish nodules from surroundings which helps to achieve the sensitivity of . false positives per scans with the sensitivity of 91.5% and 92.8% with the false positive rate of 0.25 and 0.5 per scan, respectively.
The pipeline consists of several steps.
- Image preprocessing (Lung segmentation using watershed algorithm, resizing and resampling of scans)
- Segmentation of nodule
- False positive reduction using texture and size based classification
The segmentation model is shown in figure:
