Physics / mechanism
Digital pathology converts glass histology slides into high-resolution whole-slide images (WSIs) via brightfield or fluorescence scanning at 20×–40× magnification (0.25–0.5 µm/pixel). Scanners use line-scan CMOS sensors, precision z-stack autofocus, and stitching algorithms to produce 1–4 GB TIFF/SVS files per slide. AI inference layers (CNNs, vision transformers) then perform cell segmentation, biomarker quantification, and grading. Current SOTA: Philips IntelliSite, Leica Aperio GT 450 (400 slides/day throughput), Hamamatsu NanoZoomer. FDA cleared WSI platforms now accepted as primary diagnosis. Compute bottleneck is shifting from scanning to GPU-accelerated inference at scale.
Competitive landscape
Incumbents: Philips, Leica Biosystems, Roche (iScan), Hamamatsu. AI software layer: Paige.AI (FDA-cleared for prostate cancer), PathAI, Aiforia, Owkin. Competing diagnostic modalities include multiplexed immunofluorescence (Akoya CODEX, 40+ markers vs. standard IHC 4–6), spatial transcriptomics (10x Visium, higher molecular resolution but 10–100× cost), and flow cytometry for liquid biopsy. Key differentiator axes:
Companies using
Connected ideas
Sources
Frontier (open questions)
- To be added.