Fast Mitochondria Detection for Connectomics
High-resolution connectomics data allows for the identification of dysfunctional mitochondria which are linked to a variety of diseases such as autism or bipolar. However, manual analysis is not feasible since datasets can be petabytes in size. We present a fully automatic mitochondria detector based on a modified U-Net architecture that yields high accuracy and fast processing times. We evaluate our method on multiple real-world connectomics datasets, including an improved version of the EPFL mitochondria benchmark. Our results show a Jaccard index of up to 0.90 with inference times lower than 16ms for a 512 × 512px image tile. This speed is faster than the acquisition speed of modern electron microscopes, enabling mitochondria detection in real-time. Compared to previous work, our detector ranks first for real-time detection and can be used for image alignment. Our data, results, and code are freely available.
Paper
Vincent Casser, Kai Kang, Hanspeter Pfister, Daniel Haehn: Fast Mitochondria Segmentation for Connectomics. Medical Imaging with Deep Learning (MIDL'20), 2020.
Full text: https://arxiv.org/abs/1812.06024
Datasets
We release two new datasets: A re-annotation of the EPFL Hippocampus dataset, called Lucchi++, as well as the new high-resolution Kasthuri++. For Lucchi++, our experts re-annotated the two EPFL Hippocampus stacks. Our goal was to achieve consistency for all mitochondria membrane annotations and to correct any misclassifications in the ground truth labelings. For Kasthuri++, we use the mitochondria annotations of the 3-cylinder mouse cortex volume of Kasthuri et al. The tissue is dense mammalian neuropil from layers 4 and 5 of the S1 primary somatosensory cortex, acquired using serial section electron microscopy (ssEM). Similar to the EPFL Hippocampus dataset, we noticed membrane inconsistencies within the mitochondria segmentation masks in this data. We asked our experts to correct these shortcomings through re-annotation of two neighboring sub-volumes leveraging the same process described above for the Lucchi++ dataset. The stack dimensions are 1463×1613×85vx and 1334×1553×75vx with a resolution of 3×3×30nm per voxel.
Code
The code is available here.