A compact and interpretable multi-source framework for heterogeneous medical image classification.
Paper
Loaded from Europe PMC by your browser; OSCR does not store it: doi.org · Europe PMC
The paper is loaded when this pane is shown.
The authors' code
Markdown · 225 lines · 7.1 KB · MIT
- # ML-ConvNet: A Compact and Interpretable Multi-Source Framework for Heterogeneous Medical Image Classification
- [](https://www.python.org/)
- [](https://pytorch.org/)
- [](LICENSE)
- Official repository for the paper:
- > **A Compact and Interpretable Multi-Source Framework for Heterogeneous Medical Image Classification**
- > Williams Ayivi, Xiaoling Zhang, Wisdom Xornam Ativi, Francis Sam
- > University of Electronic Science and Technology of China | Vienna University of Economics and Business
- > *Scientific Reports* (under review)
- ---
- ## Overview
- ML-ConvNet is an ultra-lightweight convolutional architecture (~4.2K parameters) designed for
- heterogeneous medical image classification across independent multi-source imaging cohorts.
- The framework operates without requiring paired or co-registered multi-modal data, making it
- suitable for realistic fragmented clinical data ecosystems.
- ### Key components
- | Component | Description | Parameters |
- |-----------|-------------|------------|
- | MBRConv | Multi-Branch Re-parameterized Convolution | ~912 per block |
- | IWO | Incremental Weight Optimization strategy | 0 (training only) |
- | FST | Feature Self-Transformation module | ~72 |
- | HDPA | Hierarchical Dual-Path Attention | ~60 |
- | LVW | Local Variance Weighted loss | 0 (training only) |
- **Total inference parameters:** 4,230 (for K=4 classes)
- **FLOPs:** 924M at 512x512 input resolution
- **Inference latency:** 2.3 ms (CPU) | 0.21 ms (GPU) | 6.1 ms (Snapdragon 888)
- ---
- ## Datasets
- Three publicly available benchmark datasets are used. Each is evaluated independently.
- | Modality | Dataset | Classes | Samples | Imbalance |
- |----------|---------|---------|---------|-----------|
- | Brain MRI | [Nickparvar Brain Tumor MRI](https://doi.org/10.34740/kaggle/dsv/14832123) | 4 (Glioma, Meningioma, Pituitary, Normal) | 7,023 | Balanced |
- | Lung CT | [IQ-OTH/NCCD](https://doi.org/10.34740/kaggle/ds/672399) | 3 (Malignant, Benign, Normal) | 1,097 | High (benign minority) |
- | Chest X-ray | [Guangzhou Pediatric Pneumonia](https://data.mendeley.com/datasets/rscbjbr9sj/2) | 2 (Pneumonia, Normal) | 5,856 | Moderate |
- Download each dataset and place under `data/` as follows:
- data/
- ├── mri/
- │ ├── Training/
- │ │ ├── glioma/
- │ │ ├── meningioma/
- │ │ ├── notumor/
- │ │ └── pituitary/
- │ └── Testing/
- ├── ct/
- │ ├── Malignant cases/
- │ ├── Benign cases/
- │ └── Normal cases/
- └── xray/
- ├── train/
- │ ├── PNEUMONIA/
- │ └── NORMAL/
- └── test/
- ---
- ## Installation
- ```bash
- git clone https://github.com/W-ayivi/ML-ConvNet.git
- cd ML-ConvNet
- pip install -r requirements.txt
- ```
- ---
- ## Repository Structure
- ML-ConvNet/
- ├── models/
- │ ├── mbrconv.py # Multi-Branch Re-parameterized Convolution + IWO
- │ ├── fst.py # Feature Self-Transformation module
- │ ├── hdpa.py # Hierarchical Dual-Path Attention
- │ └── mlconvnet.py # Full network definition (skeleton)
- ├── losses/
- │ └── lvw_loss.py # Local Variance Weighted loss
- ├── utils/
- │ ├── reparameterize.py # Structural re-parameterization utility
- │ └── preprocessing.py # Modality-specific preprocessing pipelines
- ├── evaluation/
- │ ├── metrics.py # Classification metrics
- │ └── intervals.py # CI, PI, TI computation
- ├── data/
- │ └── dataset.py # Dataset loaders
- ├── train.py # Training script
- ├── evaluate.py # Evaluation script
- ├── requirements.txt
- └── README.md
- ---
- ## Preprocessing
- All inputs are standardized to 512x512x3. Single-channel acquisitions (MRI, CT)
- are replicated across three channels. No modality identity label is passed as a
- conditioning signal at any stage.
- | Modality | Pipeline |
- |----------|----------|
- | MRI | N4ITK bias correction, skull stripping, in-mask z-score, 3-channel replication |
- | CT | HU windowing [-1000, 400], scale to [0,1], lesion-centric sampling |
- | X-ray | CLAHE, median filtering, global z-score (fallback min-max), pad to square |
- On-the-fly augmentation during training: elastic warping, random rotation, Gaussian noise.
- Minibatches are constructed via round-robin sampling across cohorts to maintain
- approximate modality balance. Missing modality inputs are replaced by zero tensors.
- ---
- ## Training
- ```bash
- python train.py --modality mri --data_dir data/mri --epochs 100 --batch_size 32 --lr 1e-4
- python train.py --modality ct --data_dir data/ct --epochs 100 --batch_size 32 --lr 1e-4
- python train.py --modality xray --data_dir data/xray --epochs 100 --batch_size 32 --lr 1e-4
- ```
- ### Training configuration
- | Parameter | Value |
- |-----------|-------|
- | Optimizer | Adam |
- | Learning rate | 1e-4 |
- | Batch size | 32 |
- | Epochs | 100 |
- | Weight initialization | Xavier uniform |
- | Validation | 10-fold stratified cross-validation (patient-level) |
- | Framework | PyTorch v2.x, Python 3.12.3 |
- | GPU | NVIDIA RTX 4060 (8GB VRAM) |
- Structural re-parameterization is applied after training completion within each fold.
- Numerical equivalence between training-phase and inference-phase outputs is verified
- with maximum absolute deviation below 1e-6.
- ---
- ## Evaluation
- ```bash
- python evaluate.py --modality mri --checkpoint checkpoints/mri_fold1.pth --data_dir data/mri
- ```
- Metrics reported per fold: Accuracy, Precision, Recall, F1-score, MCC, Cohen's kappa, AUC.
- Aggregate statistics: mean +/- SD, 95% CI, 95% PI, 95/95 TI across 10 folds.
- ---
- ## Results
- ### MRI Brain Tumor
- | Metric | Mean +/- SD | 95% CI |
- |--------|-------------|--------|
- | Accuracy | 0.982 +/- 0.002 | [0.981, 0.983] |
- | Macro F1 | 0.980 +/- 0.002 | [0.979, 0.981] |
- | MCC | 0.973 +/- 0.002 | [0.972, 0.974] |
- | AUC | 0.996 +/- 0.001 | [0.995, 0.997] |
- ### CT Lung Nodule
- | Metric | Mean +/- SD | 95% CI |
- |--------|-------------|--------|
- | Accuracy | 0.949 +/- 0.003 | [0.947, 0.951] |
- | Macro F1 | 0.933 +/- 0.003 | [0.931, 0.935] |
- | MCC | 0.918 +/- 0.003 | [0.916, 0.920] |
- | AUC | 0.972 +/- 0.002 | [0.971, 0.973] |
- ### Chest X-ray
- | Metric | Mean +/- SD | 95% CI |
- |--------|-------------|--------|
- | Accuracy | 0.958 +/- 0.002 | [0.957, 0.959] |
- | Macro F1 | 0.950 +/- 0.002 | [0.949, 0.951] |
- | MCC | 0.940 +/- 0.003 | [0.938, 0.942] |
- | AUC | 0.978 +/- 0.002 | [0.977, 0.979] |
- ---
- ## Citation
- ```bibtex
- @article{ayivi2025mlconvnet,
- title={A Compact and Interpretable Multi-Source Framework for Heterogeneous Medical Image Classification},
- author={Ayivi, Williams and Zhang, Xiaoling and Ativi, Wisdom Xornam and Sam, Francis},
- journal={Scientific Reports},
- year={2025},
- note={Under review}
- }
- ```
- ---
- ## Acknowledgements
- The core architectural design of ML-ConvNet is inspired by the MobileIE framework
- (Yan et al., ICCV 2025). We thank the authors for making their work publicly available.
- ---
- ## License
- MIT License. See LICENSE for details.
- ---
- ## Contact
- Williams Ayivi — [email hidden]
README.md at commit c7c3a6c, under MIT · at the source
Overview
- School of Information and Communication Engineering, University of Electronic Science and Technology of China, Chengdu, 611731 China
- Department of Information Systems and Operations Management, Vienna University of Economics and Business, 1020 Vienna, Austria
- School of Computer Science and Technology, University of Electronic Science and Technology of China, Chengdu, 611731 China
- School of Information and Software Engineering, University of Electronic Science and Technology of China, Chengdu, 611731 China
- NOMATEN CoE, National Centre for Nuclear Research, 05-400 Otwock, Poland
- Scientific Research Center, Baku Engineering University, AZ0101 Baku, Azerbaijan
Abstract
The abstract is not reproduced here: the paper's license (CC BY-NC-ND) does not allow it. Read it in the paper, at the publisher or on Europe PMC.
Repository
Its files are read in the Code ↔ Paper reader above.
W-ayivi/ML-ConvNet
c7c3a6c92cf485a23473e477af51f754a0272341, 22 March 2026Availability: 1 check, the latest on 28 September 2026: the link answers
- 28 September 2026: the link answers
1 file
- README.md, Text, 225 lines
Code availability statement
The paper has a code availability statement. Its license (CC BY-NC-ND) does not allow reproducing it here; in short, from what the harvester recognized in it:
- it points to the authors' code: W-ayivi/
ML-ConvNet
Read it in the paper: doi.org/10.1038/s41598-026-51120-x.
Tracing map
Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.
What the map holds:
- 1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
- 0 scripts, each with its path and the digest of its content;
- no match between paragraphs and code yet;
- neither the text of the paper nor the code itself.
Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.
Data
Datasets cited
- data.mendeley.com/
datasets/ , at Mendeley Data; found in “Data availability”rscbjbr9sj
Data availability statement
The paper has a data availability statement. Its license (CC BY-NC-ND) does not allow reproducing it here; in short, from what the harvester recognized in it:
- it points to a dataset: data.mendeley.com/
datasets/ rscbjbr9sj
Read it in the paper: doi.org/10.1038/s41598-026-51120-x.
Versions
The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.
Version 1, 28 September 2026: the first record
Recorded: type, language, journal, volume, issue, pages, dates, 5 authors, 12 keywords, 8 MeSH terms, 14 references.
Cite
This paper
Ayivi, W., Zhang, X., Ativi, W. X., Sam, F., & Aligayev, A. (2026). A compact and interpretable multi-source framework for heterogeneous medical image classification. Scientific reports, 16(1), 20414. https://
BibTeX
@article{ayivi2026compac
author = {Ayivi, Williams and Zhang, Xiaoling and Ativi, Wisdom Xornam and Sam, Francis and Aligayev, Amil},
title = {{A compact and interpretable multi-source framework for heterogeneous medical image classification}},
journal = {Scientific reports},
year = {2026},
month = may,
volume = {16},
number = {1},
pages = {20414},
publisher = {Nature Publishing Group},
issn = {2045-2322},
doi = {10.1038/
url = {https://
pmid = {42069855},
pmcid = {PMC13328668}
}
RIS
TY - JOUR
AU - Ayivi, Williams
AU - Zhang, Xiaoling
AU - Ativi, Wisdom Xornam
AU - Sam, Francis
AU - Aligayev, Amil
TI - A compact and interpretable multi-source framework for heterogeneous medical image classification
T2 - Scientific reports
J2 - Sci Rep
PY - 2026
DA - 2026/
VL - 16
IS - 1
SP - 20414
SN - 2045-2322
PB - Nature Publishing Group
DO - 10.1038/
UR - https://
LA - en
ER -
CSL-JSON
{
"id": "10.1038/
"type": "article-journal",
"title": "A compact and interpretable multi-source framework for heterogeneous medical image classification",
"container-title": "Scientific reports",
"author": [
{
"family": "Ayivi",
"given": "Williams"
},
{
"family": "Zhang",
"given": "Xiaoling"
},
{
"family": "Ativi",
"given": "Wisdom Xornam"
},
{
"family": "Sam",
"given": "Francis"
},
{
"family": "Aligayev",
"given": "Amil"
}
],
"container-title-short":
"volume": "16",
"issue": "1",
"page": "20414",
"DOI": "10.1038/
"PMID": "42069855",
"PMCID": "PMC13328668",
"ISSN": "2045-2322",
"publisher": "Nature Publishing Group",
"URL": "https://
"language": "en",
"issued": {
"date-parts": [
[
2026,
5,
2
]
]
}
}
The tracing map gets a citation of its own once an author has validated it and it has a DOI.
Similar papers
The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.
- [1] doi:10.3390/bioengineering10070823
- Multi-Stage Classification of Retinal OCT Using Multi-Scale Ensemble Deep Architecture.Journal: Bioengineering (Basel, Switzerland)In common: data.mendeley.com/datasets/rscbjbr9sj, other, methods / tools, 1 reference
- [2] doi:10.3390/e23121651
- Multilevel Deep Feature Generation Framework for Automated Detection of Retinal Abnormalities Using OCT Images.Journal: Entropy (Basel, Switzerland)In common: data.mendeley.com/datasets/rscbjbr9sj, other, methods / tools, 1 reference
- [3] doi:10.1038/s41598-025-32886-y [code]
- Language assisted learnable hyperdimensional computing framework for retinal disease classification.Journal: Scientific reportsIn common: data.mendeley.com/datasets/rscbjbr9sj, other, other condition, 1 reference
- [4] doi:10.1371/journal.pone.0304943
- Stitched vision transformer for age-related macular degeneration detection using retinal optical coherence tomography images.Journal: PloS oneIn common: data.mendeley.com/datasets/rscbjbr9sj, other, other condition, 1 reference
- [5] doi:10.3389/frai.2026.1849571
- NeuroPlast: a learnable activation function evaluated under knowledge distillation for medical image classification.Journal: Frontiers in artificial intelligenceIn common: data.mendeley.com/datasets/rscbjbr9sj, methods / tools, other condition, 1 reference
- [6] doi:10.3390/jimaging9100219
- Evaluating Retinal Disease Diagnosis with an Interpretable Lightweight CNN Model Resistant to Adversarial Attacks.Journal: Journal of imagingIn common: data.mendeley.com/datasets/rscbjbr9sj, 1 reference
- [7] doi:10.3390/jimaging9070140 [code]
- Conv-ViT: A Convolution and Vision Transformer-Based Hybrid Feature Extraction Method for Retinal Disease Detection.Journal: Journal of imagingIn common: data.mendeley.com/datasets/rscbjbr9sj, 1 reference
- [8] doi:10.1371/journal.pone.0307317
- A deep learning framework for the early detection of multi-retinal diseases.Journal: PloS oneIn common: data.mendeley.com/datasets/rscbjbr9sj, other, other condition
- [9] doi:10.1007/s11042-023-14831-1
- MediNet: transfer learning approach with MediNet medical visual database.Journal: Multimedia tools and applicationsIn common: data.mendeley.com/datasets/rscbjbr9sj, methods / tools, other condition
- [10] doi:10.1117/1.jbo.31.8.086006
- Noise-aware dynamic convolution for improved generalizability of retinal disease diagnosis using optical coherence tomography images.Journal: Journal of biomedical opticsIn common: data.mendeley.com/datasets/rscbjbr9sj, other
Contribute
The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.
Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.
Claim this paper
Correct its record
Say what each link of this record is, remove the ones that are not the paper's, add the ones that are missing. The correction becomes a new version of the record, in its Versions section.
Validate its tracing map
You validate the map as this page shows it: 1 repository of the authors' code, each at its verified commit and with its license, 0 scripts, and 0 matches between paragraphs and code (see the Code and Map sections). It then receives a DOI on Zenodo, with you (your ORCID iD) and OSCR as its creators; the code itself is not deposited.
The map's fingerprint: sha256:eb6743ce2ca3da0b…
Add the badge to the code's README
If you are an author or a maintainer of this paper's code, you can add this badge. It says what OSCR knows of the paper (its code, its validated tracing map) and links the code to this page. Paste one of these snippets under the README's title. Keep the two marks around it, so that the badge can be replaced in place and is never added twice. OSCR changes nothing in your repository.
Markdown
<!-- oscr-badge:start -->
[](https://openscicode.org/badge/doi/10.1038/s41598-026-51120-x)
<!-- oscr-badge:end -->reStructuredText
.. oscr-badge:start
.. image:: https://openscicode.org/badge/doi/10.1038/s41598-026-51120-x.svg
:target: https://openscicode.org/badge/doi/10.1038/s41598-026-51120-x
:alt: OSCR
.. oscr-badge:endHTML
<!-- oscr-badge:start -->
<a href="https://openscicode.org/badge/doi/10.1038/s41598-026-51120-x"><img src="https://openscicode.org/badge/doi/10.1038/s41598-026-51120-x.svg" alt="OSCR"></a>
<!-- oscr-badge:end -->On GitHub, open the README, choose Edit (the pencil), paste the snippet under the title, then "Commit changes".
If you choose to, you can have the badge kept up to date for you, with oscr badge --write from the command line or with the GitHub Action in your repository's workflows. Every other form and option (a card, AsciiDoc, LaTeX, shields.io) is on the badge page.
Request its removal
To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).
Discussion, reproductions, activity
Discussion: code errors and mismatches between the code and the paper about this paper and its code, from readers and its authors, beside the paper and its tracing map.
Report a mismatch between the code and the paper · Report a code error (signed in)
Reproductions: reports from readers who ran the authors' code and did not get the paper's result, what they ran, with which environment, commit and data.
Report a reproduction failure (signed in)
Versions of its code: the releases of its code tied to a version of this paper (the preprint, the accepted manuscript, the version of record…), each with the tracing map versioned with it.
Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.
