Machine Learning for MDs Weekly Digest
What’s New in ML for MDs
Welcome to the ML for MDs Newsletter. The mission of ML for MDs is to connect physicians interested in machine learning. This newsletter provides the most relevant news, journal articles, and jobs at the intersection of medicine and machine learning.
Fun Facts
In honor of Sam Altman describing AI as a “printing press moment” during his Congressional testimony:
- The inventor of the printing press, Gutenberg, died poor; since so few people could read, he couldn’t sell his product.
- The Gutenberg Bible was printed in red ink, leading people to say it was written in human blood and created by witchcraft
This Week’s Top Stories
- The race to create low processing power generative AI models is on…Google announced that it ran a version of Palm2 on a Galaxy Samsung phone, and more are sure to follow. The head of DeepMind said their Gecko mobile model can produce about 10-15 words per second on a phone.
- Hollywood writers and now journalists are organizing and demanding protections related to AI. A company that hopes to replace healthcare workers with AI just got a $50m investment from a16z, and it’ll be interesting to see how that plays out. Traditionally physicians and healthcare workers haven’t organized efficiently, and liability/malpractice issues would need a significant revamping to replace the majority of healthcare workers.
Weekly summary
A dive into Code Free Deep Learning (CFDL) resources and two example use cases in radiology and pathology
Code Free Deep Learning allows non-programmers to run proof of concept studies or create straightforward tasks or personalized chatbots. Check out the amazing
Code Free Deep Learning Resources
- Free open-source Code Free Deep Learning platforms:
- Flowise – visual UI tool to build customize LLM using Langchain
- Orange – focused on data mining and data visualization
- Teachable Machine (Google) – For images and sounds
- Fremium Code Free Deep Learning platforms:
- Obviously AI – focused on predictive analytics
- StackAI – creates workflows connecting LLMs, vector databases, and data sources to produce APIs
- Commercially available Code Free Deep Learning platforms:
- Amazon Rekognition Custom Labels (Amazon)
- Apple Create ML (Apple)
- Baidu EasyDL (Baidu)
- Clarifai Train (Clarifai)
- Google Cloud AutoML Vision
- Huawei ModelArts ExeML (Huawei)
- MedicMind Deep Learning Training Platform (MedicMind)
- Microsoft Azure Custom Vision (Microsoft)
Code Free Deep Learning Use Cases
- Code free deep learning in radiology
- The authors wanted to use deep learning to create “models for disease classification, object detection, and segmentation on chest radiographs”
- Evaluated 6 code free deep learning (CFDL) platforms:
- Amazon, Clarifai, Google, Microsoft, Apple, MedicMind, which the authors trained on NIH and other free standard radiologic databases
- They trained the models to look for common pathologies (graph bottom right) and also segmentation for pneumonia and pneumothorax
- The models performed poorly, and all the solutions required some component of coding
- Code free deep learning in pathology
- Assessed Microsoft Custom Vision and Google AutoML for classification of histopathological images of diagnostic central nervous system tissue samples
- Both systems were able to perform binary classifications with high accuracy (from ∼80% to nearly 100%)
- Both platforms offered a relatively straightforward and intuitive user experience
- “Interestingly, it proved difficult to trace back misclassified images as both platforms would only display the image itself, but not the name or other data of the submitted image file”
Community News
- If you haven’t introduced yourself, please do so under the #intros channel.
Thanks for being a part of this community! As always, please let me know if you have questions/ideas/feedback.
Sarah
Sarah Gebauer, MD