Frequently asked questions

Only corneal cross-sectional images in DICOM format (AS-OCT B-scans). Whether a file is DICOM is checked from its content, not its extension, so exports without an extension can be uploaded too. Files that are not DICOM, the device's alignment photographs and images of other structures cannot be analyzed.

The application was developed and tested with exports from the Heidelberg ANTERION. Images from other devices have not been validated; results for such images may not be reliable.

Up to 10 when selecting files one by one. For more, use the folder upload; every file in the folder is analyzed in turn. The total size of one upload is limited to 2 GB.

Not strictly. The application skips files it cannot read as DICOM or recognize as corneal sections; the skipped files are shown in an on-screen warning and listed in the key file of the folder package. The analysis takes longer, though, so sorting them out beforehand is more comfortable.

Not recommended. The browser adds the files of the subfolders to the upload as well, and the application tries to analyze all of them; non-corneal files are skipped, but the analysis takes longer. Put all DICOM files into one folder and leave no other folder inside it.

Folder selection works in current desktop browsers (Chrome, Edge, Firefox, Safari). Some mobile browsers do not support it and the button opens a file picker instead; in that case use a computer or upload the files one by one.

One image takes about 20–35 seconds on the server; the first analysis adds up to half a minute for loading the model. The 6 meridians of one eye finish in 2–3 minutes, a folder of 42 files in 12–25 minutes. When the analysis starts, the screen shows an estimate for that moment and updates it as the work progresses.

The server runs one analysis at a time. If another user's analysis is running, yours is queued; the screen shows how many analyses are ahead of you and when yours is expected to start. Just wait without closing the tab.

The analysis stops and no result is downloaded. You need to upload the images again and start over. Nothing is kept on the server.

Check it when all uploaded images are different meridians of the same eye; the Excel file then comes as a single-eye report that combines the meridians. Leave it unchecked for images of different eyes or different persons; each eye becomes a separate row.

The Excel file then combines different eyes as if they were meridians of a single eye, and the results are misleading. Uncheck the box and repeat the analysis. The segmented images are correct in both cases.

It analyzes an anonymized sample section stored on the server so that you can try the application without an image of your own. Use it to see what the results screen and the downloaded files look like.

In folder mode RESULTS.zip downloads automatically when the analysis finishes. It contains three things: images/ (per image: analysis crop, full frame, overlay, mask and zone-line images), measurements.xlsx (eye summary in single-eye mode; otherwise the combined report, file list and eye summaries) and key.xlsx (which code belongs to which patient). When files are uploaded one by one, the same contents are obtained with the download buttons on the screen: eye summary, combined report, images and separate results.

Your browser may have blocked the download. Check the download notice at the right end of the address bar and allow it, or use the “Download RESULTS.zip” button shown in the Download section of the left panel. Try this before refreshing the page; otherwise the analysis starts over.

Each row belongs to one image or one eye. The sheets hold the layer thicknesses (epithelium, Bowman's layer, stroma, Descemet–endothelium and total) by ring zone, keratometry and apex values, selected biomarkers and the keratoconus probability. How zones and thicknesses are defined is explained on the measurement definitions page.

Where the center of the ring zones is placed: the vertex of the conic fitted to the anterior surface, or the thinnest point. The default is the vertex; when the thinnest point is chosen, the zones are redrawn accordingly. The Excel file contains separate zone sheets for both centers. Both definitions are explained on the measurement definitions page.

The percentage on the screen is the probability given for your single uploaded image by the model that combines the three measurement groups (A, B, C). The model was developed at a single centre with a single device type and validated in an independent set from the same centre. The result is based on one image; it does not replace clinical assessment, topography or the physician's judgement. CorneApp is a research prototype.

The warning appears when the uploaded image is in the narrow format. For this format the probability calibration was derived from limited independent data (38 eyes). The measurements can be used; interpret the probability with caution and prefer current wide-format scans where possible.

Not exactly; the measurement definitions may differ. In wide-format scans the differences between the application and the device software were small for the thinnest-point and total thickness; epithelial thickness may differ by a few micrometres. Bowman's layer and the Descemet–endothelium complex are not reported by the device, so there is no direct comparison for them.

No. The images are processed only for the duration of the analysis; when the session ends, the temporary files and the results package on the server are deleted. Nothing is stored permanently on the server.

No. The identity fields inside the DICOM files are read and replaced with codes such as Patient 1, Patient 2 or, in single-eye mode, Eye 1, Eye 2 before anything is written. The codes are resolved only in key.xlsx; that file contains personal data, so remove it from the package and keep it to yourself when sharing results.

First make sure the file is a corneal DICOM (open it in a DICOM viewer). If the problem persists, send a bug report through the form on the contact page; do not include patient names or identifiers in your message.

With the TR / EN switch in the top band. The user guide, the measurement definitions and this page are also available in both languages.