Dermatologist and Imaging Informaticist @MSKCancerCenter, Board of Directors @SIIM_tweets, 🇦🇷🦇 ⚽ fan, formerly @dukebme and #columbiaderm, views my own
Important report out of the @theNASEM: Rethinking Race and Ethnicity in Biomedical Research. This document may be huge in guiding future work across biomedicine. Take a look here: https://t.co/vGO7oxNF9e
New paper from @SimpsonLabUW @DermatologyUW is online @JCI_insight🍾We found ERK hyperactivation disrupts cell-cell adhesion in epidermis, linked this pathway to both drug-induced & spontaneous Grover disease & identified MEK inhibition as a novel therapy: https://t.co/Ko48RnTxNp
Brilliant talk by @Dr_vron on #SMR24 on artificial and augmented intelligence in the detection of melanoma. Her key points to improve: larger/broader data sets, real world human-AI collaboration studies, integrating series of data over time as we do in the clinic. @SocietyMelanoma@MSKCancerCenter
@DrCristianMD@TitusBrinker@Nature@DKFZ Absolutely thrilled to see public models continuing to be tested and validated!! I think it further underlines the value of public/publicly available and testable models @ISICArchive. Congratulations to the authors and research team!
Thank you for the shoutout, Dr. Meskó!
ICYMI, we published a synthesis of 141 binding policies applicable to AI in healthcare and population health in the EU and 10 European countries:
https://t.co/eexXvoo7os
@SciencePartisan@littmath I was able to do it with regular magna tiles - strategy was to get all 20 attached to each other with the 5 pieces at each corner but not the correct angles and kind of bend them together gently at the end. It was fun!
@predict_addict@kaggle@ISICArchive Would love to hear more about your thoughts - no matter what, these models need to be prospectively tested in their intended use setting, and the calibration measured in that context. In that sense, the pAUC shouldn't create any difference in risk from any other metrics.
This is great. While this is specifically about students, I find microaggressions often happen with other members of the care team, and it's very helpful in the 1:1 debrief to also work on identifying preferences for future situations.
When I say "sunhat" I don't just mean straw hats like the one on the left (I don't like straw, it's not my vibe), I *also* mean floppy bucket-style hats like the ones on the right (specifically in cotton/linen).
One of the key outputs of the @HIMSS@SIIM_Tweets Enterprise Imaging Community is their freely available #openaccess whitepapers.
Over the past decade, the community has published 1️⃣7️⃣ whitepapers on a variety of topics.
https://t.co/j1Av8SkZ8k
I'm shopping for a sunhat, and it should go without saying, but, I MUST emphasize: do NOT buy a something meant to protect you from the sun/heat/sweating made out of polyester, lol.
I am not anti-polyester. But I am when it comes to that. You might as well wear a plastic bag.
Just in! Weir et al. propose considerations that researchers must take in order to improve device robustness to skin tone bias. Scan the QR code below or follow the link to learn more!
https://t.co/hcOXIMHK9B
There is so much work to do to evaluate the potential for bias in medical technologies and I'm so grateful for the amazing collaborators and our patients who participate in this research as we are trying to do better.
Excited to share the results of a fun multidisciplinary collaboration! We know that skin tone measurement is important to evaluate technologies across domains - and we just published the results of a review of those methods! https://t.co/JYkpiVUTdP
Yes. I understand that it can be very challenging to evaluate AI and especially LLMs, but our priority is protection of patients and improvement of health. If we can't figure out whether something will help patients, and it has a large potential for harm, we should not use it.