@kkli20111

PhD student at NUS School of Medicine, Computational Brain Imaging Group

 Singapore
Joined November 2021
  • Tweets5
  • Following111
  • Followers14
🧵 Struggling to replicate a proteomics biomarker study? Our new preprint quantifies one common cause: data leakage. With NO true signal, some runs with data leakage still reached AUCs near 0.8. w/ @_JakeVogel_, Caitlin Finney, Artur Shvetcov & @_neuroproteome
1
9
1
20
2,932
Hetu Li retweeted
Thomas Yeo@bttyeoJun 13
If you are attending @OHBM @OHBM_Trainees come by and check out our poster on how excessive censoring hurts parcellation and TMS target accuracy. Poster number 2399 Presenting on Wednesday, June 17 (12:45-13:45) and Thursday, June 18 (13:45-14:45)
Thomas Yeo@bttyeoMar 27
Do you censor high motion frames in fMRI? In two preprints by @twktan @mandyfmejia, we find that we may be censoring too much! doi.org/10.64898/2026.03.09.… arxiv.org/html/2603.07380v1 In personalized TMS, strict censoring leads to worse targets than no censoring even with high motion!
12
20
2,292
Hetu Li retweeted
Thomas Yeo@bttyeoJun 13
For those attending @OHBM @OHBM_Trainees, come check out our poster finding that functional MRI brain foundation models do not outperform ridge regression for phenotypic prediction Poster number 2269 Presenting on Wed, June 17 | 12:45-13:45 & Thurs, June 18 | 13:45-14:45
Thomas Yeo@bttyeoJun 11
The unreasonable effectiveness of ridge regression! We see that in functional MRI data as well.
3
17
1
58
6,979
Hetu Li retweeted
Thomas Yeo@bttyeoMay 22
Here's bonus slides on cross-validation tests, separate from our preprint. Covering: 1. paired (sign-flip) permutation test 2. label-swap permutation test 3. sample-level vs fold-averaged stats 4. a common misapplication of the corrected t-test 5. three bootstrap variants 1/N
Thomas Yeo@bttyeoMay 21
In a meta-analysis of 210 biomedical AI studies that statistically compared models under cross-validation, 97% used invalid statistical tests. Here's our new preprint doi.org/10.64898/2026.05.17.… led by @tianchuzeng @kkli20111 @ZShaoshi @ten_photos 1/N
1
24
1
45
7,899
Hetu Li retweeted
Thomas Yeo@bttyeoMay 21
In a meta-analysis of 210 biomedical AI studies that statistically compared models under cross-validation, 97% used invalid statistical tests. Here's our new preprint doi.org/10.64898/2026.05.17.… led by @tianchuzeng @kkli20111 @ZShaoshi @ten_photos 1/N
9
95
17
275
122,477
Load more