Abstract:The development of mouse spermatozoa is a continuous process from spermatogonia, spermatocytes, spermatids to mature sperm. Those developing germ cells (spermatogonia, spermatocyte, and spermatids) together with supporting sertoli cells are all enclosed inside seminiferous tubules of the testis, their identification is key to testis histology and pathology analysis. Automated segmentation of all these cells is a challenging task because of their dynamical changes in different stages. The accurate segmentation of testicular cells is critical in developing computerized spermatogenesis staging. In this paper, we present a novel segmentation model, SED-Net, which incorporates a squeeze-and-excitation (SE) module and a dense unit. The SE module optimizes and obtains features from different channels, whereas the dense unit uses fewer parameters to enhance the use of features. A human-in-the-loop strategy, named deep interactive learning, is developed to achieve better segmentation performance while reducing the workload of manual annotation and time consumption. Across a cohort of 274 seminiferous tubules from stages VI to VIII, the SED-Net achieved a pixel accuracy of 0.930, a mean pixel accuracy of 0.866, a mean intersection over union of 0.710, and a frequency weighted intersection over union of 0.878, respectively, in terms of four types of testicular cell segmentation. There is no significant difference between manual annotated tubules and segmentation results by SED-Net in cell composition analysis for tubules from stages VI to VIII. In addition, we performed cell composition analysis on 2346 segmented seminiferous tubule images from 12 segmented testicular section results. The results provided quantitation of cells of various testicular cell types across 12 stages. The rule reflects the cell variation tendency across 12 stages during development of mouse spermatozoa. The method could enable us to not only analyze cell morphology and staging during the development of mouse spermatozoa but also potentially could be applied to the study of reproductive diseases such as infertility.

Deep SED-Net with interactive learning for multiple testicular cell types segmentation and cell composition analysis in mouse seminiferous tubules

Rapid detection of mouse spermatogenic defects by testicular cellular composition analysis via enhanced deep learning model

Deep learning-based method for segmenting epithelial layer of tubules in histopathological images of testicular tissue

Deep Learning Based Multi-Organ Segmentation And Metastases Segmentation In Whole Mouse Body And The Cryo-Imaging Cancer Imaging And Therapy Analysis Platform (Citap)

Automated Deep Learning Model for Sperm Head Segmentation, Pose Correction, and Classification

Automatic segmentation of Sperm's parts in microscopic images of human semen smears using concatenated learning approaches

Advancements in automated sperm morphology analysis: a deep learning approach with comprehensive classification and model evaluation

Seminiferous Tubule Molecular Imaging for Evaluation of Male Fertility: Seeing is Believing.

Germline stem cell lineage tracing by free-floating immunofluorescent assay of mouse seminiferous tubule.

Artificial intelligence interpretation of touch print smear cytology of testicular specimen from patients with azoospermia

Multidimensional Morphological Analysis of Live Sperm Based on Multiple-target Tracking

A modified U-Net to detect real sperms in videos of human sperm cell

CP-Net: Instance-aware part segmentation network for biological cell parsing

Automated Sperm Morphology Analysis Based on Instance-Aware Part Segmentation

Establishment and Characterization of Testis Organoids with Proliferation and Differentiation of Spermatogonial Stem Cells into Spermatocytes and Spermatids

An automatic classification method of testicular histopathology based on SC-YOLO framework

An Accurate Nuclei Segmentation Algorithm in Pathological Image Based on Deep Semantic Network

A Comprehensive Roadmap of Murine Spermatogenesis Defined by Single-Cell RNA-Seq

3D Segmentation of Mice Gland Based on Ensemble Learning

FastCellpose: A Fast and Accurate Deep-Learning Framework for Segmentation of All Glomeruli in Mouse Whole-Kidney Microscopic Optical Images

Deep learning enabled multi-organ segmentation of mouse embryos