Glo-In-One

Glo-In-One performs quantitative detection, segmentation, and classification of glomeruli from high-resolution whole slide imaging (WSI) to enable quantitative analysis in digital renal pathology.

Key Features:

• Holistic glomerular analysis: Outputs WSI-level multi-class circle glomerular detection results, segmented glomerular image patches, and lesion type annotations.
• Fine-grained Global Glomerulosclerosis (GGS) characterization: Classifies GGS into assessed-solidified-GSS (related to hypertension), disappearing-GSS (resulting from contiguous fibrotic interstitium), and obsolescent-GSS (associated with aging).
• Self-supervised deep learning: Incorporates self-supervised deep learning trained with a large-scale collection of 30,000 unlabeled glomerular images obtained via web image mining to reduce dependency on annotated data.
• Performance metrics: Reports average precision of 0.627 for glomerular detection using circle representations and a patch-wise Dice similarity coefficient of 0.955 for segmentation.
• Outputs: Produces detection results, segmented image patches with masks, and lesion classifications as analytical outputs.

Scientific Applications:

• Computer-assisted diagnosis: Provides quantitative glomerular detection, segmentation, and lesion classification to support diagnostic workflows in digital renal pathology.
• Renal pathology research: Enables large-scale quantitative analyses of glomerular morphology and GGS subtypes to support studies of kidney disease mechanisms.

Methodology:

Processes whole slide images using self-supervised deep learning trained with 30,000 unlabeled glomerular images obtained via web image mining to produce WSI-level multi-class circle detections, segmented glomerular image patches with masks, and lesion classifications.