✦ Automated segmentation and classification of single cells
Accurately distinguishing between apoptosis and pyroptosis is crucial for understanding regulated cell death and its implications for immunity and disease. Deeptosis integrates Cellpose for robust cell segmentation with a Vision Transformer (ViT) model to achieve high-accuracy phenotypic classification. Deeptosis supports all standard image formats, including .png, .jpg/.jpeg, .tif/.tiff, and .bmp, and with no restrictions on image size.
✦ Deeptosis
✦ Important Note
♦ Job Identifier (JID):
Each submission is assigned a unique Job Identifier (JID). Use this ID to track the processing status and to download the corresponding results package.
♦ Input:
1. Upload Image(s): All standard microscopy image formats, including .png, .jpg/.jpeg, .tif/.tiff, and .bmp, are supported. Multiple cells in one image are allowed. Very large images may take longer to process.
2. Parameter: Select which classes to display in the annotated result (e.g. "All classes", "Apoptosis only", or "Pyroptosis only"). This does not change the model prediction itself; it only controls what is rendered on the output image/table.
3. Threshold: Confidence score cutoff for visualization (Default = 0.50). Predictions with a confidence lower than this value will be hidden from the annotated image and summary table.
♦ Console:
1. Clear: Remove the current input and output panels and reset the page to the initial state.
2. Example: Load a built-in demo image and run the full pipeline to see how segmentation, classification, and annotation work.
3. Submit: Send the current input to the backend and generate the annotated image plus CSV summary.