Maarten van Smeden
@MaartenvSmeden
Followers
40K
Following
28K
Media
1K
Statuses
20K
Is @maartenvsmeden.bsky.social
Utrecht
Joined December 2010
Our overview of and guidance for performance measures to evaluate medical AI is finally out! - Stop bashing AUROC - Calibration + clinical utility are key - Plot risk distributions - Classification measures are improper https://t.co/m3xrAHDzdg
thelancet.com
Numerous measures have been proposed to illustrate the performance of predictive artificial intelligence (AI) models. Selecting appropriate performance measures is essential for predictive AI models...
6
33
99
PROBAST+AI is out! This is "a comprehensive tool that allows to assess the trustworthiness, value, fairness, quality, risk of bias and applicability of prediction models" ( https://t.co/axY1alYIs9)
https://t.co/TTL6sS4bL7
bmj.com
The Prediction model Risk Of Bias ASsessment Tool (PROBAST) is used to assess the quality, risk of bias, and applicability of prediction models or algorithms and of prediction model/algorithm...
1
20
63
congrats @MaartenvSmeden for winning a VIDI (pretty much Netherland's most prestigious personal grant); great new for methods research on prediction models and their evaluation
Feeling grateful today for receiving a VIDI personal grant from @ZonMw This grant will allow us to develop better methods to ensure AI-based decision support systems in healthcare will remain safe and effective to use over time
1
2
43
Congratulations to all other VIDI grant recipients, in particular to my friend and close collaborator @laure_wynants (it is a good day for methods research!), and to my UMC Utrecht colleagues Hanneke Willemen, Albertien van Eerde, Neeltje Crombag and Chantal Tax
3
0
10
Feeling grateful today for receiving a VIDI personal grant from @ZonMw This grant will allow us to develop better methods to ensure AI-based decision support systems in healthcare will remain safe and effective to use over time
Over 86,7 million euros has been granted to 102 researchers from different domains. The grant creates the opportunity to develop an innovative line of research and expand their own research group. More information=> https://t.co/FJ1lNS79Ht
7
1
90
new PhD position with @MaartenvSmeden and me! are you: - interested in the intersection of "science" and deep learning? - keen to work with electrocardiography (ECG) data - eager to learn and be part of a vibrant data science team @UMCUtrecht? see
1
5
9
🚨Hiring a new fully funded PhD student🚨 We are searching for an enthusiastic PhD student to work on cutting edge methods research, working on methods to combine knowledge from cardioelectrophysiology, 'conventional’ ECG analysis and machine learning https://t.co/wJVsFEbQ6t
1
12
18
New blog article challenging medical researchers who are drawn to clustering patients to create so-called "phenotypes" to think again: https://t.co/50fqPRSLLl
@MaartenvSmeden @statsepi0 #Statistics #PrecisionMedicine @vandy_biostat
fharrell.com
Patient clustering, often described as the finding of new phenotypes, is being used with increasing frequency in the medical literature. Most of the applications of clustering of observations are not...
4
53
163
Must admit, I always envied this handle And also: ohhh nooooo....
4
1
36
My first blog in ~4 years (!) and perhaps my last tweet (moving to the SoMe that talks about the colour of the sky): this is about the unlikely mashup of an Aesop fable and a paper from @timpmorris.bsky.social, @maartenvsmeden.bsky.social and Tra Pham. https://t.co/waRvYabVSc
theupperquartile.wixsite.com
Aesop, Marginality Principle, Interactions and Children's Books. In the past few months / years I unfortunately found myself reading less and less. There are only two categories of books that I still...
0
1
4
The single best thing a student intending to use data should do is the following: Learn the difference between description, prediction, and causal inference. "I want to predict..." -- I am willing to bet you don't.
14
18
243
Ok, so it seems the academic conversation is moving towards the place that combines the color blue and the word sky. See you over there!
4
4
36
Given that your clinical prediction model counts as software as medical device, please do not say the model just needs (one) external validation before implementing it in practice
6
9
47
Once you realize p-values are probabilities relating to observing data rather than probabilities of an hypothesis being true, you are already doing significantly better than most people
37
371
3K
Mandatory periodic reminder. ➡️ Statistics: lines through data point cloud ➡️ Data science: lines through data point cloud ➡️ Machine learning: lines through data point cloud ➡️ AI: lines through data point cloud
5
32
146
Happy to present this NEW PAPER of the @stratosinit on 'Regression without regrets: Initial data analysis is a prerequisite for multivariable regression' https://t.co/wdeWBtYrb6
@f2harrell @MHuebnerPhD #stratosida
link.springer.com
BMC Medical Research Methodology - Statistical regression models are used for predicting outcomes based on the values of some predictor variables or for describing the association of an outcome...
1
23
73
Looking forward to talking about the blessings and curses of AI for clinical prediction modeling later today in Kopenhagen. And finally meeting @jessenleon in real life :) Slides ICYI:
3
8
53
Nice to see TRIPOD+AI reporting recommendations (link to guidance👇) mentioned in this @NICEComms position statement on the "Use of #AI in evidence generation" 😀 https://t.co/Fvts8JvF5a
NEW PAPER out today in @BMJ_latest TRIPOD+AI: reporting recommendations for studies developing or validating prediction models for use in healthcare that use #machinelearning methods https://t.co/ed9oiB1MVG
#ArtificialIntelligence #AIstandards #OpenAccess Please share 🙏
0
8
42