Gaetano Scebba

UNIQUE: A Framework for Uncertainty Quantification Benchmarking #MachineLearning pubs.acs.org/doi/10.1021/ac… @GScebba @foo_fighterin @Novartis #JCIM Vol64 Issue22 #ApplicationNote

As part of our new data-centric AI focus (vanderschaar-lab.com/data-centric-a…), we also introduce DC-Check: an actionable guide for practitioners & researchers to practically engage with DCAI for all stages of the ML pipeline: Data, Training, Testing, Deployment. vanderschaar-lab.com/dc-check/

Towards "deep" evidence-based medicine.

MIMIC-IV was published this week. The core dataset has been out for a while, but we've just published the deidentified free-text clinical notes: 300,000+ discharge summaries and 2.5 million radiology reports! nature.com/articles/s4159…

Weakly supervised deep-learning models for the analysis of whole-slide images from tumour biopsies perform better at prognostic tasks if the models incorporate context from the local microenvironment. rdcu.be/cTU7L [N&V] rdcu.be/cTU55 [Paper]

A phase 3 clinical trial evaluating the impact of a digital nurse navigator-led system in patients receiving oral #cancer treatment shows that remote monitoring with a smartphone results in improved patient experience & optimizes resources @GustaveRoussy nature.com/articles/s4159…

Great collaboration with stellar teams at @derbalgrist @Unispital_USZ @mhsl_ethz @jia_zhang93 Sabrina Catanzaro, Carina Mihai, Oliver Distler, Martin Berli, @karliwalti @skintegrityCH

We propose Detect-and-Segment, a two stage #DeepLearning approach to produce #wound #segmentation maps with high #generalization capabilities. The DS approach showed high performance on #OutOfDistribution testing and enabled the reduction of segmentation labels used for training.

What is the simplest yet robust and automatic way to segment #wound images? Read our latest work on robust wound image segmentation with deep neural networks! Paper link: doi.org/10.1016/j.imu.…

MedMNIST v2 is a large-scale MNIST-like collection of standardized biomedical images, consisting of ~708K 2D images and ~10K 3D images in total. Supports research purposes in biomedical image analysis, computer vision and machine learning. paperswithcode.com/dataset/medmni…

Introducing SeqSNR, a new multi-task learning system that captures the complexities of a realistic ICU setting to predict adverse events, accounting for competing risks, along with the interdependencies between organ systems. goo.gle/3x2wHjH

Soon you’ll be able to use Google Fit on Pixel to measure your heart rate and respiratory rate using just your phone’s camera. Learn more about this update, announced at today’s #TheCheckUp event → goo.gle/36IhOsA

In a major scientific breakthrough, the latest version of #AlphaFold has been recognised as a solution to one of biology's grand challenges - the “protein folding problem”. It was validated today at #CASP14, the biennial Critical Assessment of protein Structure Prediction (1/3)

If you're looking for Ph.D. positions in AI this year, consider applying to the new Fellowship program of the recently launched ETH AI Center: ai.ethz.ch/education/phd-… @ETH_en

Stanford has now conducted >360,000 video visits this year and utilization remains quite high since start of the pandemic. The digital age of healthcare delivery has begun. @StanfordMed #digitalhealth

@GiuliaDaPoian @snsf_ch Great Giulia!! Congratulations!!

"I think telehealth is here to stay. ... But I think we’re discovering that, even as we go back to in-person visits, there’s a lot we’re going to retain in the virtual world." -Dean Lloyd Minor on #COVID19 and the future of medicine. stan.md/30NdYf6

Working at the intersection of #AI and #Health? Here some overview articles I found useful: (1) Comparison of deep learning performance against health-care professionals (Lancet Digital Oncology 2019): thelancet.com/pdfs/journals/… (Note: major criticism of validation standards)