I still find it hard to imagine how people got around before GPS was a thing or how anyone could wait for days, if not weeks, to receive a response before telephones - in the not too distant past. Couple years later, would people wonder "How did people write and/or code before LLMs"? Back towards the end of 2020 as I was wrapping up my internship at Biogen (perhaps the best 6 months of my life, as of September 2024, will probably write a blog on it next time I want to procrastinate), I decided that I wanted to pursue a PhD developing interpretable, handcrafted imaging biomarkers. The struggle of writing the statement of purpose (SoP) was REAL (and I still do struggle with writing a lot)! Back then, Prof. Priya Narasimhan's tweetorials helped a lot - Link 1 , Link 2 . Perhaps the one tweet that really helped me get started was You can always edit a bad page. You can’t edit a blank page. — @jodipicoult — Prof. Priya Narasimhan (@YinzcamPriya) November 14, 2020 I spent se...
The field of radiomic biomarkers seems to be maturing with higher levels of evidence being generated through validation in randomized trials. However, these biomarkers, currently, are inductive - based on a retrospective training set, we hope the signature/model extrapolates to a new patient. It only takes one black swan (aka an unknown unknown) for the model to crash and burn. This is particularly a huge problem with deep learning approaches and I am therefore more bullish on hand-crafted radiomics. Within radiomics, I believe shape features are much more interpretable than texture features. I see the way forward as radiomic biomarkers being incorporated in the standard biomarker discovery process in earlier phases of trials. This will largely be a collaborative effort between pharmas and small/medium companies with expertise in the AI/radiomic side. However, the key point is that while prospective validation may be convincing, it might not provide sufficient immunity to black swans. ...
Comments
Post a Comment