A Multiple Input Multiple Output Framework for the Automatic Optical Fractionator-based Cell Counting in Z-Stacks Using Deep Learning
Doctor of Philosophy (Ph.D.)
Degree Granting Department
Computer Science and Engineering
Dmitry Goldgof, Ph.D.
Lawrence O. Hall, Ph.D.
Peter R. Mouton, Ph.D.
Rangachar Kasturi, Ph.D.
Sudeep Sarkar, Ph.D.
Ashwin Parthasarathy, Ph.D.
Adaptive Stain Separation, Automatic Optical Fractionator, Automatic Unbiased Stereology, Microscopy z-stacks, MIMO U-Net
Quantifying cells in a defined region of biological tissue is critical for many clinical and preclinical studies, especially in pathology, toxicology, cancer, and behavior. Unbiased stereology is the state-of-art method for quantification of the total number and other morphometric parameters of stained objects in a defined region of biological tissue. As part of a program to develop accurate, precise, and more efficient automatic approaches for quantifying morphometric changes in biological tissue, our group has shown that both deep learning-based and hand-crafted algorithms can estimate the total number of histologically stained cells at their maximal profile of focus in extended depth of field (EDF) images. However, the majority of the previous approaches were designed for single-immunostained datasets. To extend the automatic approaches for dual stain (counterstained) tissue sections, the first contribution of this dissertation is an adaptive digital stain separation method. The information related to a primary stain of interest is extracted as a single-channel image which can be used as input for the previous methods for cell quantification. The proposed method was applied to two dual stain datasets. Automatic counting results using previous methods on the stain-separated images showed comparable results to manual counts.
The next contribution of this work is a disector-based Multiple Input and Multiple Output (MIMO) framework for automatic cell counting in a stack of z-axis images (also known as disector stacks). This DL-based digital automation of the ordinary optical fractionator ensures accurate counts through spatial separation of stained cells in the z-plane, thereby avoiding false negatives from overlapping cells in EDF images without the shortcomings of 3D and recurrent DL models. The contribution overcomes the issue of under-counting errors with EDF images due to overlapping cells in the z-plane (masking). We demonstrate the practical applications of these advances with automatic disector-based estimates of the total number of NeuN-immunostained neurons in a mouse neocortex on counterstained and single-immunostained datasets consisting of brightfield microscopy z-stacks. We also applied the proposed method to a set of confocal microscopy z-stacks from mouse neocortex for Rab5 endosome counting. In summary, this work provides the first demonstration of automatic estimation of a total cell number in tissue sections using a combination of deep learning and the disector-based optical fractionator method.
As part of the disector-based MIMO framework, we introduce a MIMO U-Net model to efficiently use a 2D U-Net for inter-image feature learning in a Multiple Input Multiple Output system that poses a binary segmentation task as a multi-class multi-label segmentation task. The proposed MIMO U-Net approach is also validated on a dataset from the Cell Tracking Challenge achieving comparable results to a compared method in which a U-Net architecture is equipped with integrated memory units.
To conclude, this dissertation advances the overall aim of automating unbiased stereology for estimating total numbers for different biological objects in stained tissue sections using deep learning with accuracy comparable to manual stereology, yet obtaining with higher throughput efficiency and reproducibility.
Scholar Commons Citation
Dave, Palak, "A Multiple Input Multiple Output Framework for the Automatic Optical Fractionator-based Cell Counting in Z-Stacks Using Deep Learning" (2022). USF Tampa Graduate Theses and Dissertations.