Ankush, MD

Cardiothoracic & Vascular Imaging 🫀🫁 🩻 · AI in Radiology 🤖

@wakehealthDR

[Ankush, MD @drankush]

Thank you @ASK_MSK You are so kind. I was myself busy with my next innovative project, that I forgot to even post this on my X.

If anyone needs full paper or any help setting it up, please feel free to reach out.

[Dr. Ameya Kawthalkar @ASK_MSK]

An amazing free and open source software to setup AI augmented radiology dictation:

VoxRad by Dr Ankush @drankush

Great example of radiologists innovating and building tools for radiologists!

Runs fully locally

1. Download Whisper (or other free, open source TTS AI model): fine tune it on your dictation and reports
2. Download a free open source LLM which will run on CPU: Phi or Gemma
3. Load them in VoxRad
4. Upload your templates into VoxRad

The app will generate a report based on your personal report templates with a short prompt like "MRI knee. ACL tear and bucket handle lateral meniscus tear."

Links in 🧵

[Ankush, MD @drankush]

I think you are trolling @AnthropicAI . It can't be used for medical research whatsoever.

[nature @Nature]

Calling all researchers using Anthropic's AI model Claude: how are you using the new Claude Fable 5 model in your research?

We want to hear about the most impressive things it's built for your research projects or left you asking what the fuss is all about. Can it do things you couldn't do before? Let us know.

[Omar Sanseviero @osanseviero]

Super excited about the release of the Colab CLI! 🔥

Enabling your agents to easily interact with Colab is about to unlock lots of fun experiments

https://t.co/xyQ6GFINtg

[Samuel Hume @DrSamuelBHume]

One of the most amazing things I’ve ever seen: a standing ovation for the full Daraxonrasib results

I feel inspired and energised, to put it mildly — we have a targeted therapy for pancreatic cancer now, and nothing is undruggable anymore

[Baris Turkbey MD @radiolobt]

More figures from this nice review👍

[theRadiologist @radiologistpage]

Everyone is willing radiologists into extinction but I’m willing to bet we’ll still be here in 10 years going strong (with AI making things easier for us). We’ll look back at previous years like radiology in hard mode

[Daniel Vargas, MD @DanielVargasMD]

This is a final boss type CXR!

There is Absence of the Right Clavicular Companion Shadow

This sign can indicate an underlying supraclavicular space-occupying lesion, typically lymphadenopathy

In this case, a man in his 40s, had lymphadenitis and R IJ thrombosis

#radiology

[Wil @WillyRontgen]

Another day, another Powerscribe One outage. A story as old as time. No matter what you do, do not let your group migrate to PSOne. @NuanceInc fix your cloud. @drsxr @drmoneymatters @heacockmd @benwhitemd

[Ankush, MD @drankush]

@_rohitreddy_ Have you run any validation tests/benchmarks using human experts yet. We have built something similar with the manuscript under round 2 review.

[Francis Deng, MD @francisdeng]

Med students may be interested to know that physician compensation varies by specialty, as does time worked per year (based on weeks time off a year and hours worked per week on).

[Ankush, MD @drankush]

What a great time we had at #SCMR26

[Ankush, MD @drankush]

Thank you @thoracicrad 😊🙏🏻🥳✌🏻

[Eric Topol @EricTopol]

The largest randomized trial of medical A.I.
—Over 100,000 women in Sweden
—radiologist + AI vs 2 radiologists, in follow-up
—AI added led to 29% more cancer detected, 44% reduced workload, and
—Less cancer dx in subsequent 2 years, and, when found, less aggressive
https://t.co/e1hY3F0cGo

[Society of Thoracic Radiology @thoracicrad]

Beautiful @RadioGraphics review on imaging of lung‑sparing surgery (https://t.co/HYR6LjoCDP). Covers preop CT for segmentectomy planning (nodule location, vasculature, fissures, emphysema/fibrosis) and postop complications like air leak and suture‑line granulomas.

[Eric Topol @EricTopol]

An assessment of LLMs for medical tasks
@NatureMedicine @drnigam
"Most models perform best in clinical note generation (0.74–0.85) and patient communication and education (0.76–0.89), moderately in medical research assistance (0.65–0.75) and clinical decision support (0.63–0.77), and worst in administration and workflow (0.53–0.63)."
https://t.co/Mwv6vYOad2

[Ankush, MD @drankush]

@GithubProjects A PowerPoint Add-in with 94,000+ free icons from 108 open-source libraries.

https://t.co/wDD5vKd6iX

[Tom Dörr @tom_doerr]

Curated list of research papers on agentic AI for healthcare

https://t.co/yynBTCrRTQ

[Simplifying AI @simplifyinAI]

Google has open-sourced the LangExtract Python library.

It turns unstructured documents into structured data, with clear source references for each result.

100% open-source.

[Or Hiltch @_orcaman]

Today we are launching @openwork_ai, an open-source (MIT-licensed) computer-use agent that’s fast, cheap, and more secure.

@openwork_ai  is the result of a short two-day hackathon our team decided to hack, which brings together some of our favorite open source AI modules into one powerful agent, to allow you to:

1. Bring your own model/API key (any provider and model supported by @opencode is supported by Openwork)

2. ~4x faster than Claude for Chrome/Cowork, and much more token-efficient, powered by dev-browser by @sawyerhood (legend)

3. More secure - contrary to Claude for Chrom/Cowork, does not leverage the main browser instance where you are logged into all services already. You login only to the services you need. This significantly reduces the risk of data loss in case of prompt injections, to which computer-use agents are highly exposed.

4. Free and 100% open-source!

You can download the DMG (macOS only for now) or fork the github repo via the link in bio (@openwork_ai).

Let us know what you think (or better, send a pull request)!

[Google Research @GoogleResearch]

Announcing our latest open medical AI models for developers: MedGemma 1.5, which is small enough to run offline & improves performance on 3D imaging (CT & MRI), & MedASR, a speech-to-text model for medical dictation. Both available on Hugging Face + Vertex AI. https://t.co/jPm4UodeAH

#MedGemma #HealthAI #GenerativeAI

[Philipp Schmid @_philschmid]

New Gemma model! @GoogleDeepMind MedGemma 1.5 4B is an updated 4B variant expanding support for 3D imaging, longitudinal analysis, and localization while improving text accuracy over MedGemma 1.

- Interprets 3D volumes for CT, MRI, and whole-slide histopathology.
- Achieves 38% intersection-over-union for X-ray anatomical localization.
- Analyzes longitudinal X-rays to compare current against historical scans.
- Records 89.6% accuracy on EHR understanding, a 22% improvement.

https://t.co/q08fl0xWrj