Histo-Cloud

Histo-Cloud performs convolutional neural network-based segmentation of whole slide images (WSIs) on cloud infrastructure as an extension of HistomicsTK to enable quantitative analysis of histological structures in nephropathology and other tissues.

Key Features:

• Cloud-Based Computation: Executes WSI segmentation and analysis on cloud servers to handle large-scale datasets.
• Advanced Segmentation Capabilities: Uses a convolutional neural network (CNN) tailored for WSI segmentation to identify glomeruli, interstitial fibrosis, tubular atrophy, and vascular structures in renal and non-renal WSIs.
• Feature Extraction: Extracts quantitative features from segmented regions for downstream machine learning and analysis.
• Scalability and Adaptability: Scales to large WSI volumes and adapts to segment diverse histological structures across staining methods.
• Transfer Learning Best Practices: Incorporates transfer learning best practices to improve performance across different datasets and conditions.
• Integration with HistomicsTK: Extends HistomicsTK to support computational histology workflows.

Scientific Applications:

• Nephropathology: Enables quantitative analysis of kidney histology in human and animal models focusing on glomerular, tubular, interstitial, and vascular features.
• Disease Research: Supports studies of diabetic nephropathy and HIV-associated nephropathy by enabling detailed analysis of glomerular features in murine models.

Methodology:

Deploys a convolutional neural network (CNN) for WSI segmentation, applies transfer learning for cross-dataset performance, executes computations on cloud servers, and extracts quantitative features from segmented regions.