Date of Award
Doctor of Philosophy
Arshad M. Khan
People who analyze images of biological tissue rely on the segmentation of structures as a preliminary step. In particular, laboratories studying the rat brain delineate brain regions to position scientific findings on a brain atlas to propose hypotheses about the rat brain and, ultimately, the human brain. Our work intersects with the preliminary step of delineating regions in images of brain tissue via computational methods.
We investigate pixel-wise classification or segmentation of brain regions using ten histological images of brain tissue sections stained for Nissl substance. We present a deep learning approach that uses the fully convolutional neural network, U-Net, to segment white matter regions. Then, we show that the segmentation of white matter can be achieved with fewer human labels to address the arduous task of obtaining human labels. We aim to assess the viability of segmenting brain structures from images alone with supervised and unsupervised approaches. Further, we determine how image and model scale and additional domain knowledge affect segmentation.
Experiments show convolutional models perform better when the data has enough resolution to distinguish cytoarchitectural patterns. At the same time, we show that larger input sizes can yield better segmentations with models designed to incorporate long-range dependencies. Moreover, additional domain knowledge, such as atlas-guided parcellations and region-based object detection, can improve segmentation results. Our work shows that segmenting brain regions is possible with histological images of brain tissue sections stained for Nissl substance.
Our efforts contribute to a long history of characterizing the brain. We continue working to streamline the reconstruction of brain maps and build accurate models of all brain systems.
Received from ProQuest
Arnal, Alexandro, "Region Detection & Segmentation of Nissl-Stained Rat Brain Tissue" (2022). Open Access Theses & Dissertations. 3643.