Skip to main content

    Pathology troubleshooting

    Fix empty tile sets, missing slide URLs, unsupported workflows, and slide-loading failures.

    When a pathology run fails or returns fewer results than you expected, the cause is usually the slide reference, the quality filters, or the workflow you chose. This page covers the common failure modes and how to resolve them.

    "Missing required parameter 'slide_url'"

    The job needs an S3 URI to the slide as its one required input.

    • Upload the whole slide image to workspace storage first, then reference it by its s3://… URI.
    • Confirm the URI is well-formed — the tool rejects paths that do not start with s3://.

    The tile set is empty or tiny

    Tile extraction discards any tile that fails the tissue or blur filter, so a filter set too strict for your slide yields few or no tiles.

    • Lower tissue_threshold (default 0.5) if much of the slide is faintly stained or sparse.
    • Lower blur_threshold (default 100.0, Laplacian variance) if in-focus regions are being dropped as blurry.
    • Check the magnification — tiling at 40x over a small image can leave too little area per tile.

    "Unsupported workflow"

    The processor runs tile_extraction, tissue_detection, and feature_extraction. If you request a workflow it cannot execute, it raises an unsupported-workflow error.

    • Pick one of the three operational workflows for slide processing.
    • For tissue-typing or cell-level tasks, start from tile extraction or feature extraction and analyze downstream.

    The slide will not open

    The tool opens SVS, NDPI, and pyramidal TIFF via histolab/OpenSlide, falling back to PIL for standard images.

    • Confirm the file is a supported format (SVS, TIFF, NDPI, PNG, JPG).
    • Corrupt or truncated slide files fail to load — re-upload from the source.
    • Standard PNG/JPG images work but do not carry pyramidal magnification metadata, so the magnification setting has no effect on them.

    The job runs a long time

    Runs are asynchronous and can take 10–60 minutes; very large slides process tile by tile with a heartbeat.

    • Cap the workload with max_tiles (default 1000) to bound tile extraction.
    • Ensure sufficient storage for intermediate tiles — large WSI files are processed in tiles.

    Feature extraction returns no features

    • If tile extraction found no tiles, feature extraction has nothing to process — fix the tile filters first (see above).
    • Confirm save_features is on so the features.npy artifact is uploaded.

    Guest access denied

    Pathology analysis is an authenticated, professional-tier tool. If you hit a guest-access-denied response, sign in with an account that has access.

    Was this page helpful?

    On this page

    On this page

    Loading...