CoChat

Effect of data leakage in brain MRI classification using 2D convolutional neural networks

Open access
Yagis, Ekin, Atnafu, Selamawet Workalemahu, García Seco de Herrera, Alba, Marzi, Chiara, Scheda, Riccardo, Giannelli, Marco, Tessa, Carlo, Citi, Luca, Diciotti, Stefano · 2021 · Scientific Reports

Published version of a preprint

The DOI matches its Crossref record; 79 of the 81 references it cites were found in a registry or an index.

Abstract

In recent years, 2D convolutional neural networks (CNNs) have been extensively used to diagnose neurological diseases from magnetic resonance imaging (MRI) data due to their potential to discern subtle and intricate patterns. Despite the high performances reported in numerous studies, developing CNN models with good generalization abilities is still a challenging task due to possible data leakage introduced during cross-validation (CV). In this study, we quantitatively assessed the effect of a data leakage caused by 3D MRI data splitting based on a 2D slice-level using three 2D CNN models to classify patients with Alzheimer’s disease (AD) and Parkinson’s disease (PD). Our experiments showed that slice-level CV erroneously boosted the average slice level accuracy on the test set by 30% on Open Access Series of Imaging Studies (OASIS), 29% on Alzheimer’s Disease Neuroimaging Initiative (ADNI), 48% on Parkinson’s Progression Markers Initiative (PPMI) and 55% on a local de-novo PD Versilia dataset. Further tests on a randomly labeled OASIS-derived dataset produced about 96% of (erroneous) accuracy (slice-level split) and 50% accuracy (subject-level split), as expected from a randomized experiment. Overall, the extent of the effect of an erroneous slice-based CV is severe, especially for small datasets.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41598-021-01681-w

Cites

79 of 79 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Schematic diagram of the overall T1-weighted MRI data processing and validation scheme. First, a preprocessing stage included co-registration to a standard space, skull-stripping and slices selection based on entropy calculation. Then, CNNs model’s training and validation have been performed on each dataset in a nested CV loop using two different data split strategies: (a) subject-level split, in which all the slices of a subject have been placed either in the training or in the test set, avoiding any form of data leakage; (b) slice-level split, in which all the slices have been pooled together before CV, then split randomly into training and test set.
Figure 1 Schematic diagram of the overall T1-weighted MRI data processing and validation scheme. First, a preprocessing stage included co-registration to a standard space, skull-stripping and slices selection based on entropy calculation. Then, CNNs model’s training and validation have been performed on each dataset in a nested CV loop using two different data split strategies: (a) subject-level split, in which all the slices of a subject have been placed either in the training or in the test set, avoiding any form of data leakage; (b) slice-level split, in which all the slices have been pooled together before CV, then split randomly into training and test set.CC BY · PubMed Central
Figure 2 The two different networks based on the VGG16 architecture are shown. Each colored block of layers illustrates a series of convolutions. (a) The first model, named as VGG16-v1 consists of five convolutional blocks followed by three fully connected layers. Only the last three fully connected layers are fine-tuned. (b) On the other hand, the second model, VGG16-v2, has five convolutional blocks followed by a global average pooling layer, and all the layers are fine-tuned.
Figure 2 The two different networks based on the VGG16 architecture are shown. Each colored block of layers illustrates a series of convolutions. (a) The first model, named as VGG16-v1 consists of five convolutional blocks followed by three fully connected layers. Only the last three fully connected layers are fine-tuned. (b) On the other hand, the second model, VGG16-v2, has five convolutional blocks followed by a global average pooling layer, and all the layers are fine-tuned.CC BY · PubMed Central
Figure 3 A modified ResNet-18 architecture with an average pooling layer at the end is shown. The upper box represents a residual learning block with an identity shortcut. Each layer is denoted as (filter size, # channels); layers labeled as “freezed” indicates that the weights are not updated during backpropagation, whereas when they are labeled as “fine-tuned” they are updated. The identity shortcuts can be directly used when the input and output are of the same dimensions (solid line shortcuts) and when the dimensions increase (dotted line shortcuts). ReLU rectified linear unit.
Figure 3 A modified ResNet-18 architecture with an average pooling layer at the end is shown. The upper box represents a residual learning block with an identity shortcut. Each layer is denoted as (filter size, # channels); layers labeled as “freezed” indicates that the weights are not updated during backpropagation, whereas when they are labeled as “fine-tuned” they are updated. The identity shortcuts can be directly used when the input and output are of the same dimensions (solid line shortcuts) and when the dimensions increase (dotted line shortcuts). ReLU rectified linear unit.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

1 retracted

Show all 79 references

Research works printed but not linked: 4

  1. Simonyan, K. & Zisserman, A. Very deep convolutional networks for large-scale image recognition. Preprint at http://arXiv.org/1409.1556 (2015).
  2. Lin, M., Chen, Q. & Yan, S. Network in network. Preprint at http://arXiv.org/1312.4400 (2014).
  3. TensorFlow Developers. TensorFlow (2021).
  4. Brett, M. et al. Nipy/nibabel: 2.3.3 (2019).

Cited by

See all 195

Metadata

Authors
Yagis, E., Atnafu, S. W., García Seco de Herrera, A., Marzi, C., Scheda, R., Giannelli, M., Tessa, C., Citi, L., & Diciotti, S.
Affiliations
  • University of Bologna (IT)
  • Azienda Ospedaliera Universitaria Pisana (IT)
  • Ospedale Versilia (IT)
  • University of Essex (GB)
Corresponding author
Diciotti, Stefano
Venue
Scientific Reports
Published
2021-11-19
Volume
11
Issue
1
Pages
22544-22544
ISSN
2045-2322 (ISSN-L)
OpenAlex
W3214734349
Cited by
161
References
81
Citations by year
2026: 61 · 2025: 32 · 2024: 26 · 2023: 25 · 2022: 16
FWCI
12.02
Type
journal article
Language
en
Open access
Licence
Locations
  • Scientific Reports (publishedVersion, cc-by)
  • Research Square (acceptedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • UNED repository (publishedVersion, other-oa)
  • Open Access at Essex (University of Essex) (acceptedVersion, other-oa)
  • Archivio istituzionale della ricerca (Alma Mater Studiorum Università di Bologna) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Florence Research (University of Florence) (submittedVersion)
Funders
  • European Union's Horizon 2020 research and innovation programme (824153 “POTION”)
  • NVIDIA Academic GPU Grant Program
Topics
  • Health Sciences › Medicine › Psychiatry and Mental health › Dementia and Cognitive Impairment Research
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Functional Brain Connectivity Studies
  • Health Sciences › Health Professions › Health Information Management › Artificial Intelligence in Healthcare
Keywords
data leakage, cross-validation, 2D convolutional neural network, brain MRI classification, Alzheimer's disease, Parkinson's disease, model generalization, neurological disease diagnosis, small datasets
MeSH terms
Deep Learning, Aged, Aged, 80 and over, Alzheimer Disease, Brain, Cross-Sectional Studies, Female, Humans, Image Interpretation, Computer-Assisted, Magnetic Resonance Imaging, Male, Middle Aged, Parkinson Disease, Predictive Value of Tests, Reproducibility of Results, Case-Control Studies, Neural Networks, Computer, Neuroimaging

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Effect of data leakage in brain MRI classification using 2D convolutional neural networks. https://app.cochat.ai/papers/10.1038/s41598-021-01681-w
The paper
Yagis, E., Atnafu, S. W., García Seco de Herrera, A., Marzi, C., Scheda, R., Giannelli, M., Tessa, C., Citi, L., & Diciotti, S. (2021). Effect of data leakage in brain MRI classification using 2D convolutional neural networks. Scientific Reports. https://doi.org/10.1038/s41598-021-01681-w

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.