How to use
- Decide which model to use:
Low magnification (5×) - overall disease pattern
High magnification (40×) - individual aggregates
- Take snapshots:
Extract square snapshots from the region of interest at the correct magnification
- Prepare images:
Save snapshots into a single folder
Ensure images match the selected model:
• Low magnification - representative field of the region
• High magnification - one aggregate per image
- Upload images:
Select the folder to upload and multiple images will be processed automatically
- Interpret outputs:
The model returns class probabilities (%) for each image
(Higher values indicate stronger morphological similarity to a trained class)
- Export results:
Download predictions as an Excel file for further analysis
* Models were trained on AT8-stained tau pathology images from representative frontal cortex regions, but may generalise to other tau immunostains and brain regions.
* Predictions are probabilistic outputs from a supervised convolutional neural network and should be interpreted alongside full neuropathological assessment, not used as a standalone diagnostic tool.
