Alex Kalinin
@alxndrkalinin.bsky.social
AI/ML for cell image analysis @biohub.org | prev Broad Institute, CUHK-SZ & UMich
We show how cubic can accelerate existing workflows by 10–1500×, including a 3D cell segmentation tutorial from CellProfiler. ⚡
5/6
5/6
October 23, 2025 at 6:01 PM
We show how cubic can accelerate existing workflows by 10–1500×, including a 3D cell segmentation tutorial from CellProfiler. ⚡
5/6
5/6
cubic keeps things familiar: swap imports, put your image on the device you want, and the same function names automatically dispatch to the right CPU/GPU implementation—optional acceleration with minimal refactoring. ✨
4/6
4/6
October 23, 2025 at 6:01 PM
cubic keeps things familiar: swap imports, put your image on the device you want, and the same function names automatically dispatch to the right CPU/GPU implementation—optional acceleration with minimal refactoring. ✨
4/6
4/6
The visual difference is clear: compared to a baseline (F-net), Spotlight sharply reduces artifacts, resulting in clearer nuclear boundaries and less segmentation artifacts, while preserving foreground textures.
4/5
4/5
July 10, 2025 at 6:52 PM
The visual difference is clear: compared to a baseline (F-net), Spotlight sharply reduces artifacts, resulting in clearer nuclear boundaries and less segmentation artifacts, while preserving foreground textures.
4/5
4/5
🔬🤖 Introducing Spotlight: virtual staining (VS) improved by focusing on cells
VS models often learn to predict both cells and noisy background, because training treats all pixels equally. We address this by explicitly training models to prioritize foreground.
1/5
VS models often learn to predict both cells and noisy background, because training treats all pixels equally. We address this by explicitly training models to prioritize foreground.
1/5
July 10, 2025 at 6:52 PM
🔬🤖 Introducing Spotlight: virtual staining (VS) improved by focusing on cells
VS models often learn to predict both cells and noisy background, because training treats all pixels equally. We address this by explicitly training models to prioritize foreground.
1/5
VS models often learn to predict both cells and noisy background, because training treats all pixels equally. We address this by explicitly training models to prioritize foreground.
1/5
🔬API-first feature extraction for image-based profiling workflows
If you need to obtain interpretable features from your segmented microscopy images, but want to do it in a fully automated way, we know the struggle.
1/6
If you need to obtain interpretable features from your segmented microscopy images, but want to do it in a fully automated way, we know the struggle.
1/6
July 8, 2025 at 7:22 PM
🔬API-first feature extraction for image-based profiling workflows
If you need to obtain interpretable features from your segmented microscopy images, but want to do it in a fully automated way, we know the struggle.
1/6
If you need to obtain interpretable features from your segmented microscopy images, but want to do it in a fully automated way, we know the struggle.
1/6
We approach this as an information retrieval problem: replicates of a phenotypically active perturbation should rank higher in similarity when retrieved against control replicates. We quantify this with a single, data-driven metric—mean average precision (mAP).
3/7
3/7
June 10, 2025 at 7:54 PM
We approach this as an information retrieval problem: replicates of a phenotypically active perturbation should rank higher in similarity when retrieved against control replicates. We quantify this with a single, data-driven metric—mean average precision (mAP).
3/7
3/7