Hierarchical Encoder-Decoder With Soft Label-Decomposition for Mitochondria Segmentation in EM Images

Semantic segmentation of mitochondria from electron microscopy (EM) images is an essential step to obtain reliable morphological statistics about mitochondria. However, automatically delineating plenty of mitochondria of varied shapes from complex backgrounds with sufficient accuracy is challenging. To address these challenges, we develop a hierarchical encoder-decoder network (HED-Net), which has a three-level nested U-shape architecture to capture rich contextual information. Given the irregular shape of mitochondria, we introduce a novel soft label-decomposition strategy to exploit shape knowledge in manual labels. Rather than simply using the ground truth label maps as the unique supervision in the model training, we introduce additional subcategory-aware supervision by softly decomposing each manual label map into two complementary label maps according to mitochondria's ovality. The three label maps are integrated with our HED-Net to supervise the model training. While the original label map guides the network to segment all the mitochondria of varied shapes, the auxiliary label maps guide the network to segment subcategories of mitochondria of circular shape and elliptic shape, respectively, which are much more manageable tasks. Extensive experiments on two public benchmarks show that our HED-Net performs favorably against state-of-the-art methods.

HIVE-Net: Centerline-aware hierarchical view-ensemble convolutional network for mitochondria segmentation in EM images

BACKGROUND AND OBJECTIVE

With the advancement of electron microscopy (EM) imaging technology, neuroscientists can investigate the function of various intracellular organelles, e.g, mitochondria, at nano-scale. Semantic segmentation of electron microscopy (EM) is an essential step to efficiently obtain reliable morphological statistics. Despite the great success achieved using deep convolutional neural networks (CNNs), they still produce coarse segmentations with lots of discontinuities and false positives for mitochondria segmentation.

METHODS

In this study, we introduce a centerline-aware multitask network by utilizing centerline as an intrinsic shape cue of mitochondria to regularize the segmentation. Since the application of 3D CNNs on large medical volumes is usually hindered by their substantial computational cost and storage overhead, we introduce a novel hierarchical view-ensemble convolution (HVEC), a simple alternative of 3D convolution to learn 3D spatial contexts using more efficient 2D convolutions. The HVEC enables both decomposing and sharing multi-view information, leading to increased learning capacity.

RESULTS

Extensive validation results on two challenging benchmarks show that, the proposed method performs favorably against the state-of-the-art methods in accuracy and visual quality but with a greatly reduced model size. Moreover, the proposed model also shows significantly improved generalization ability, especially when training with quite limited amount of training data. Detailed sensitivity analysis and ablation study have also been conducted, which show the robustness of the proposed model and effectiveness of the proposed modules.

CONCLUSIONS

The experiments highlighted that the proposed architecture enables both simplicity and efficiency leading to increased capacity of learning spatial contexts. Moreover, incorporating shape cues such as centerline information is a promising approach to improve the performance of mitochondria segmentation.