ISFCT COVID subject-wise

A dataset of CT images (ISFCT Dataset) with labels indicating the subject-wise split to train and test deep learning algorithms in an unbiased manner.

2026-09-1557348977

Overview

Schema Version

https://atlas.rsna.org/schemas/2025-11/dataset.json

Name

ISFCT COVID subject-wise

Link

https://doi.org/10.3390/jimaging9080159

Indexing

Keywords: COVID-19, Deep Learning, Subject-wise data split, Slice-wise data split, Repeatability, CT images, Lung features

Content: CT, CH, RS

Author(s)

Parsarad Shiva

Saeedizadeh Narges

Soufi Ghazaleh Jamalipour

Shafieyoon Shamim

Hekmatnia Farzaneh

Zarei Andrew Parviz

Soleimany Samira

Yousefi Amir

Nazari Hengameh

Torabi Pegah

S. Milani Abbas

Madani Tonekaboni Seyed Ali

Rabbani Hossein

Hekmatnia Ali

Kafieh Rahele

Organization(s)

Isfahan University of Medical Sciences

ETH Zurich

Deakin University

St. George’s Hospital

University of British Columbia

Cyclica Inc.

Durham University

Sepahan Radiology

License

Text: Creative Commons Attribution (CC BY) license

URL: https://creativecommons.org/licenses/by/4.0/

Contact

raheleh.kafieh@durham.ac.uk

Funding

This work was supported in part by the Vice Chancellery for Research and Technology, Isfahan University of Medical Sciences, under Grant No. 198337.

Ethical review

The study was approved by the Isfahan University of Medical Sciences Institutional Review Board (IRB) (IR.MUI.RESEARCH.REC.1399.003) and adheres to the tenets of the Declaration of Helsinki. Written informed consent was obtained from all participants.

Comments

A new dataset of CT images (ISFCT Dataset) with labels indicating the subject-wise split to train and test deep learning algorithms in an unbiased manner. A key feature is the inclusion of more specific labels (eight characteristic lung features) rather than being limited to just COVID-19 and healthy labels.

Date

Published: 2023-08-08

References

[1] Parsarad S, et al.. "Biased Deep Learning Methods in Detection of COVID-19 Using CT Images: A Challenge Mounted by Subject-Wise-Split ISFCT Dataset". Journal of Imaging. 2023. doi:10.3390/jimaging9080159. PMID: 37623691.

Dataset

Motivation

To provide a dataset for COVID-19 detection from CT images that addresses the shortcomings of previous works, specifically the unreliable accuracy and lack of repeatability caused by slice-wise data splits. This dataset is designed with a subject-wise split to enable unbiased training and testing of deep learning models.

Partitioning scheme

The dataset is designed for subject-wise splitting, where all CT slices from a single subject are allocated to only one set (e.g., training or testing) to avoid data leakage and overestimation of model performance, which is a common issue with slice-wise splits.