mlsanityData to 7 Oct 2026

Preventing dataset shift from breaking machine-learning biomarkers

arXiv 2107.09947 ↗First posted 21 Jul 2021cs.LG math.ST q-bio.QMcited by 14

Not on the map. Why?

Abstract (arXiv)

Machine learning brings the hope of finding new biomarkers extracted from cohorts with rich biomedical measurements. A good biomarker is one that gives reliable detection of the corresponding condition. However, biomarkers are often extracted from a cohort that differs from the target population. Such a mismatch, known as a dataset shift, can undermine the application of the biomarker to new individuals. Dataset shifts are frequent in biomedical research, e.g. because of recruitment biases. When a dataset shift occurs, standard machine-learning techniques do not suffice to extract and validate biomarkers. This article provides an overview of when and how dataset shifts breaks machine-learning extracted biomarkers, as well as detection and correction strategies.

Key facts. Cited by 14 papers in the map (data through 7 Oct 2026). Builds on 15 of its 93 references that are in the map. First posted on arXiv 21 Jul 2021.

Builds on

15 of 93 references.

References · 15

  1. 2015Learning Transferable Features with Deep Adaptation Networkscited by 1,323
  2. 2017Adversarial Discriminative Domain Adaptationcited by 1,309
  3. 2020Out-of-Distribution Generalization via Risk Extrapolation (REx)cited by 737
  4. 2015Return of Frustratingly Easy Domain Adaptationcited by 389
  5. 2018Fairness Without Demographics in Repeated Loss Minimizationcited by 363
  6. 2018Learning Models with Uniform Performance via Distributionally Robust Optimizationcited by 334
  7. 2017Data Augmentation Generative Adversarial Networkscited by 278
  8. 2019Hidden Stratification Causes Clinically Meaningful Failures in Machine Learning for Medical Imagingcited by 215
  9. 2017Emergence of Invariance and Disentanglement in Deep Representationscited by 213
  10. 2019A review of domain adaptation without target labelscited by 154
  11. 2016Variance-based regularization with convex objectivescited by 153
  12. 2020Distributionally Robust Losses for Latent Covariate Mixturescited by 57
3 more
  1. 2019Machine Learning with Multi-Site Imaging Data: An Empirical Study on the Impact of Scanner Effectscited by 40
  2. 2020Risk of Training Diagnostic Algorithms on Data with Demographic Biascited by 23
  3. 2020Kernel Distributionally Robust Optimizationcited by 6

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 1; 2022: 2; 2023: 3; 2024: 3; 2025: 2; 2026: 3.

14 papers.

Sort:

  1. 2023Generative models improve fairness of medical classifiers under distribution shiftscited by 90
  2. 2022How Robust is Unsupervised Representation Learning to Distribution Shift?cited by 24
  3. 2024Geodesic Optimization for Predictive Shift Adaptation on EEG datacited by 9
  4. 2021Statistical quantification of confounding bias in predictive modellingcited by 6
  5. 2023Deployment of Image Analysis Algorithms under Prevalence Shiftscited by 4
  6. 2023FetMRQC: a robust quality control system for multi-centric fetal brain MRIcited by 4
  7. 2025Advances in Automated Fetal Brain MRI Segmentation and Biometry: Insights from the FeTA 2024 Challengecited by 4
  8. 2024Physics-informed and Unsupervised Riemannian Domain Adaptation for Machine Learning on Heterogeneous EEG Datasetscited by 3
  9. 2024Evaluating Synthetic Data Generation for Domain Generalization in Fetal Brain MRI Segmentationcited by 3
  10. 2022Improving generalization of machine learning-identified biomarkers with causal modeling: an investigation into immune receptor diagnosticscited by 1
  11. 2025Automatic quality control in multi-centric fetal brain MRI super-resolution reconstructioncited by 1
  12. 2026Tracking the Discriminative Axis: Dual Prototypes for Test-Time OOD Detection Under Covariate Shiftcited by 0
2 more
  1. 2026Development and evaluation of CADe systems in low-prevalence setting: The RARE25 challenge for early detection of Barrett's neoplasiacited by 0
  2. 2026Prior-matched evaluation of operational Earth-observation classifiers: a three-number reporting method demonstrated on Sentinel-1 internal-wave detectioncited by 0