[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

![rohanpaul_ai Avatar](https://lunarcrush.com/gi/w:24/cr:twitter::2588345408.png) Rohan Paul [@rohanpaul_ai](/creator/twitter/rohanpaul_ai) on x 73.6K followers
Created: 2025-07-09 23:16:50 UTC

The picture sorts the data first. On top you see X imaging streams—radiology, dermatology, digital pathology, ophthalmology—and X medical-text stream. Each arrow shows how those sources feed the rest of the stack.

The images go through MedSigLIP, a vision encoder that turns each scan or photo into a compact vector the language models can read.

Those vectors flow into MedGemma 4B Multimodal, a 4B-parameter model that handles both pictures and words in a single forward pass.

For text-only work there is a larger 27B-parameter MedGemma model that skips the image part and focuses on language reasoning.

![](https://pbs.twimg.com/media/Gvc7PzLakAAyoiY.png)

XXXXX engagements

![Engagements Line Chart](https://lunarcrush.com/gi/w:600/p:tweet::1943086976170942974/c:line.svg)

**Related Topics**
[scan](/topic/scan)
[arrow](/topic/arrow)
[dermatology](/topic/dermatology)

[Post Link](https://x.com/rohanpaul_ai/status/1943086976170942974)