# Halfway to 3D: Ensembling 2.5D and 3D Models for Robust COVID-19 CT Diagnosis

arXiv: https://arxiv.org/abs/2603.14832 · HTML: https://mlsanity.com/paper/2603.14832/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 11 of its 17 references that are in the map.
- First posted on arXiv 16 Mar 2026.

## Builds on (most cited first, 11 of 11)

- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [mixup: Beyond Empirical Risk Minimization](https://mlsanity.com/paper/1710.09412/) (2017)
- [Supervised Contrastive Learning](https://mlsanity.com/paper/2004.11362/) (2020)
- [Out-of-Distribution Generalization via Risk Extrapolation (REx)](https://mlsanity.com/paper/2003.00688/) (2020)
- [Can Spatiotemporal 3D CNNs Retrace the History of 2D CNNs and ImageNet?](https://mlsanity.com/paper/1711.09577/) (2017)
- [Deep Transparent Prediction through Latent Representation Analysis](https://mlsanity.com/paper/2009.07044/) (2020)
- [MIA-COV19D: COVID-19 Detection through 3-D Chest CT Image Analysis](https://mlsanity.com/paper/2106.07524/) (2021)
- [A Deep Neural Architecture for Harmonizing 3-D Input Data Analysis and Decision Making in Medical Imaging](https://mlsanity.com/paper/2303.00175/) (2023)
- [COVID-19 Computer-aided Diagnosis through AI-assisted CT Imaging Analysis: Deploying a Medical AI System](https://mlsanity.com/paper/2403.06242/) (2024)
- [SAM2CLIP2SAM: Vision Language Model for Segmentation of 3D CT Scans for Covid-19 Detection](https://mlsanity.com/paper/2407.15728/) (2024)
- [Multi-Source COVID-19 Detection via Variance Risk Extrapolation](https://mlsanity.com/paper/2506.23208/) (2025)

## Abstract (arXiv)

We propose a deep learning framework for COVID-19 detection and disease classification from chest CT scans that integrates both 2.5D and 3D representations to capture complementary slice-level and volumetric information. The 2.5D branch processes multi-view CT slices (axial, coronal, sagittal) using a DINOv3 vision transformer to extract robust visual features, while the 3D branch employs a ResNet-18 architecture to model volumetric context and is pretrained with Variance Risk Extrapolation (VREx) followed by supervised contrastive learning to improve cross-source robustness. Predictions from both branches are combined through logit-level ensemble inference. Experiments on the PHAROS-AIF-MIH benchmark demonstrate the effectiveness of the proposed approach: for binary COVID-19 detection, the ensemble achieves 94.48% accuracy and a 0.9426 Macro F1-score, outperforming both individual models, while for multi-class disease classification the 2.5D DINOv3 model achieves the best performance with 79.35% accuracy and a 0.7497 Macro F1-score. These results highlight the benefit of combining pretrained slice-based representations with volumetric modeling for robust multi-source medical imaging analysis. Code is available at https://github.com/HySonLab/PHAROS-AIF-MIH