Drag & drop your IHC image

or video for stitching here

(or click the upload button)

Upload your IHC image

or video for stitching

was not fully uploaded. Reselect the same saved file through Browse/Files to continue where it stopped.

Upload Complete!

Please select the image resolution
and click begin to start processing.

Estimated time:
(Time can vary based on server load)

Upload complete

Choose the settings, then begin stitching.

Video Resolution
Eyepiece number On the eyepiece you look into: 10x/20 → 20. Not sure: estimated from the video, approximate.

Uploading video

Keep this page open and the screen on.
Stitching starts when the upload ends.

Please Wait

Something went wrong!

DeepLIIF (Deep Learning Inferred ImmunoFluorescence) is a virtual restaining and quantification AI model for standard immunohistochemistry (IHC) images.

Upload an image above or process one of these sample images by clicking on it:

Sample Image
512 x 512  |  Ki67 Marker  |  40x
Image from BCDataset
Sample Image
512 x 512  |  Ki67 Marker  |  40x
Image from BCDataset
Sample Image
513 x 513  |  Ki67 Marker  |  40x
Image from DeepSlides
Sample Image
513 x 513  |  Ki67 Marker  |  40x
Image from DeepSlides
Sample Image
512 x 512  |  Ki67 Marker  |  40x
Image from DeepLIIF
Sample Image
512 x 512  |  Ki67 Marker  |  40x
Image from DeepLIIF
Sample Image
512 x 512  |  Ki67 Marker  |  40x
Image from DeepLIIF
Sample Image
512 x 512  |  Ki67 Marker  |  40x
Image from DeepLIIF
Sample Image
2100 x 1300  |  Ki67 Marker  |  20x
Image from neuroendocrine_
Sample Image
2100 x 1300  |  Ki67 Marker  |  20x
Image from neuroendocrine_
Sample Image
2202 x 2204  |  CD3 Marker  |  40x
Image from Pan-Tumor
Sample Image
2255 x 2246  |  CD3 Marker  |  40x
Image from Pan-Tumor
Sample Image
2212 x 2215  |  CD3 Marker  |  40x
Image from Pan-Tumor
Sample Image
2216 x 2215  |  CD3 Marker  |  40x
Image from Pan-Tumor
Sample Image
512 x 512  |  HER2 Marker  |  20x
Image from DeepLIIF
Sample Image
512 x 512  |  HER2 Marker  |  20x
Image from DeepLIIF
Sample Image
512 x 512  |  HER2 Marker  |  20x
Image from DeepLIIF

The DP4ALL (Digital Pathology for All) Initiative allows FREE whole-slide image stitching from uploaded videos, which can be viewed and shared using our cloud-based viewer.

Videos can be acquired using either an attached microscope camera or via a cell phone camera mounted on an adapter.

During scanning • Clean slide and microscope lens. • Use a stage and move slowly. • Make brief but complete pauses at every section of the sample to avoid motion artifacts. • Have significant overlap between pauses so that there is sufficient context for stitching frames together. • Methodical and consistent movement over the sample. For example, smoothly sweeping row by row from top to bottom. • Try not to backtrack to the same field. This tends to happen at the end of a fragment of tissue. • Keep tissue in good focus throughout video. • Do not overly compress video, which can lead to noticeable artifacts. • Do not switch between magnifications during a single video. • Keep a 20X scan to about 6 minutes per slide. At 1080p that is about 330 MB, so 10 GB of free space holds about 30 slides. • File size limit is 3 GB. Setup if using a cell phone • Turn off any autofocus feature on the cell phone camera. • Mount the cell phone onto the microscope using an adapter. • Position the slide at your starting point prior to recording. • Select video mode and keep the camera at 1x. Do not zoom in: the dark ring around the field must stay visible, because it is used to measure size. • Turn up microscope light to highest setting - make sure the field is as evenly white as possible. The dark ring around the field is expected; avoid shadows inside it. • Record at 1080p HD, 30 frames per second. Do not use 4K or 60 fps: they make the video 2 to 7 times larger. • Use the smaller video format: on an iPhone, Camera > Formats > High Efficiency; on Android, HEVC or H.265 (sometimes called High efficiency video) in the camera's video settings, if offered. • On an iPhone, turn off HDR Video (Settings > Camera > Record Video), and any HDR video option on Android. HDR videos stitch with faded colors. • On the upload page, pick the number printed on the eyepiece you look into (10x/20 means 20), or Not sure if there is none. After recording • Keep the upload page open and the screen on until the upload is complete; plugging in the phone helps. • Once the slide appears on DeepLIIF, the video can be deleted from the phone. Then empty Recently Deleted (iPhone) or the Trash (Android) to free the space. • If the upload stops, open the page again in the same browser and pick the same video: it continues where it stopped.
How much will this cost? This is for free, with no cost to the users. The Nadeem Lab will incur all costs and pay for it via various sources. Who will pay for the support and usage? Costs will be covered by Dr. Nadeem's grants, MSK, and other philanthropic organizations/foundations. We will maintain long-term support for this. Is DeepLIIF more cost/storage efficient than other scanners/digitization infrastructure? Yes, DeepLIIF is more cost and memory efficient and no physical scanner is needed. There are no maintenance, service, or upgrade fees nor any support required. No expensive enterprise hardware storage is needed with the compulsion to delete older slides before scanning new slides. No separate viewer is needed. DeepLIIF has a slide viewer built into our cloud platform, providing instantaneous scanning, viewing, and sharing of the links/URLs for telepathology and consultancy. What skillset is needed for the DeepLIIF virtual slide scanner? None. We have heard from colleagues in Tanzania and Nigeria that because physical scanners are expensive, only skilled pathologists or physicians can touch them. This is in contrast to our free software driven cloud solution which does not need any special skillset to operate. Any person (educated or not) can be employed to create the videos and generate the corresponding links which can be shared with pathologists/physicians via WhatsApp or email as needed. This can open up a new job opportunity for people without any previous education, and no special training or background is needed. How does DeepLIIF handle (patient/specimen/user) data? No PHI information is recorded or stored. Pathologists or individuals scanning the slides will upload only the pixel-level videos without any patient or specimen identifiers and an anonymized URL link to the slide will be generated. The user will need to maintain this information on their end with any relevant information indexed with our generated URLs. The URL is only accessible to the person who uploaded the video and with people to whom this link was shared (these private links are not publicly available). Does DeepLIIF allow downloading of stitched slides? No. We have not enabled this feature to safeguard the data. If you want access to your stitched whole slide images for research or education, please send us the links for the stitched slides and we can share the whole slide images on your preferred sharing platform. What types of cameras are supported? We support both microscope and cell phone cameras (iPhone or Android devices) with mov, mp4, or avi formats. We will be happy to support any other formats as needed. Adapters are required for cell phone cameras. What types of specimens or stains are supported? The stitching process is specimen and stain agnostic. All types of cytology/hemepath/surgpath specimens are supported along with H&E, IHC, or special stains. What resolution/magnification is supported? We support 10X, 20X, or 40X magnification with the caveat that the entire video needs to be recorded in one resolution (the user cannot switch between one or the other during traversal/recording). Obviously, 10X is faster to record and stitch whereas 20X for finer details takes longer and might require low-resolution spatial context to navigate. We are working with our pathologist colleagues to figure out which use cases require the higher 40X or lower 10X resolutions. For 20X or 40X, the user can also just record the video for a region of interest rather than the whole tissue section to save time and effort. Does this support only core biopsies or resections as well? We support both. In resection or cytology specimens, for example, a user can record for just a region of interest rather than the whole tissue section. How does DeepLIIF handle patchy internet connections? The upload continues by itself after the connection drops. Keep the page open and the screen on until it finishes. If the page was closed, open DP4ALL again and pick the same video (on an iPhone, from Files) to continue where it stopped. Once the upload is complete you can copy the page link and leave: stitching runs on our server, and the link opens the finished slide. I generated stitched whole slide image links for individual serial sections (e.g. H&E and IHC) for the same patient and I would like to view them in the DeepLIIF viewer with a file tree view. Is this possible? Yes. For example you generated a URL for H&E (https://deepliif.org/wsi/viewer?ids=video1) and for PAS (https://deepliif.org/wsi/viewer?ids=video2) for the same patient. You can just add the "ids" from the PAS link to the end of H&E with a ";" to generate the tree with serial sections for the same patient (https://deepliif.org/wsi/viewer?ids=video1;video2). Similar to physical scanners, does this require thorough calibration to work for different stains (IHC, special stains, etc)? No. We follow a What You See Is What You Get paradigm where what you see under the microscope is what you get in the final stitched whole slide image. No calibration is needed like in physical scanners, which are notorious for automated processing that can confound the results if you are not careful (e.g., removal of HER2-low signal in IHC slides).

[Nature Machine Intelligence'22]

[CVPR'22]

[MICCAI'23]

[Histopathology'23]

[MICCAI'24]