Likelihood T. Prior

A Data Scientist is a device for turning coffee and data into better decisions.

Our CEO @davidwest_Irl shares a message on our FDA 510(k) clearance. Want to learn more? Join our webinar on March 8. bit.ly/3ume5Ag

Finally: some cutting-edge climate social science on NPR! The brilliant @triofrancos explains how we can avoid tons of dangerous, contested mining by pumping up public transit and driving normal size EVs. Your weekend dose of @cpluscp ☕️. npr.org/2023/01/29/115…

Mon collègue @CjBayesian a présenté une affiche sur l'#IA qui prédit la discordance diagnostique pour les cas de #mélanome lors du récent Digital Pathology & AI Congress (Europe) et l'a récapitulée sur le blog de @ProsciaInc : proscia.com/dissecting-dia…

Read more about our latest research wherein we demonstrate the how AI can identify melanoma cases with the highest diagnostic uncertainty. bit.ly/3BOLHHq @ProsciaInc @SeanGrullon @juliannalog @vvaughnage

From London to the blog - Sr. AI Scientist @CjBayesian recaps his poster presentation from #DigitalPathology & AI Congress: Europe. Get a look at our research demonstrating the promise of AI to identify melanoma cases w/ the highest diagnostic uncertainty. bit.ly/3BOLHHq

🚨 NEW 🚨 in @JCOCCI_ASCO: #MachineLearning algorithms to predict mortality or adverse events are often based on routinely-collected administrative #data. Big limitation. How does incorporating the #patient voice into predictions help? Answer: A lot! ascopubs.org/doi/full/10.12…

🚨NEW in @AMIAinformatics w/ Jinbo Chen @AmolNavathe @CjBayesian @k8courtright + colleagues @Penn🚨 How did the #COVID19 #pandemic impact healthcare #algorithms? We show that COVID utilization "shock"➡️long-lasting declines in algorithm performance🧵academic.oup.com/jamia/advance-…

Our deep learning model correctly predicts 82% of the discordance between pathologists.

We model each pathologist's probability of labeling a melanocytic lesion as malignant in order to determine how often we would expect pairs of pathologists to agree. Deviations from this expectation represent intra-pathologist diagnostic certainty that remains un-captured by AI.

Excited to share our team's work in London this week. We use AI to quantify implicit intra-pathologist discordance using only inter-pathologist data. This is key to understanding how much diagnostic signal the AI can capture!

Our work on predicting pathologist concordance of melanoma diagnosis detection is up! Helping identify the 'hard cases' will allow efficient routing to specialist, improving accuracy and reducing time-to-diagnosis.

Just announced! @unilabs has selected @ProsciaInc's DermAI to advance its aspiration of becoming the most digitally-driven diagnostic group. Read the full release for more. bit.ly/3ePH2Mw

Former lab-mate doing great work. Making my day :) Way to go, @ayazhyder !

Today! Senior AI Scientist Vaughn Spurrier takes the #ECDP2022 stage to present "Automated Quality Control of Whole Slide Images Using Artificial Intelligence." Catch his presentation in Ballroom II at 9:40 CEST.

Jensen's Ginequality.

🚨NEW🚨 in @PLOSONE led by @erichli1: Suppose a validated #MachineLearning algorithm identifies that your patient is at risk for a bad outcome. How do you respond? The answer: It depends on what type of doctor you are... journals.plos.org/plosone/articl…

Fighting sepsis is an ongoing battle and for patients with hospital acquired sepsis, capacity strain may contribute to antimicrobial delays. Learn more about this important underlying factor in new research from PAIR fellow @JenGinestraMD. @garyweissman pubmed.ncbi.nlm.nih.gov/35312462/

@banstweets @SEPTA I love this! A few years ago I made these for @katefcairns - one for each city we've lived in together. 🚇❤️

Colouring book self-supervised learning bayesianbiologist.com/2022/03/25/col…