A dataset for splenomegaly and its related findings in CT imaging

TL;DR

This paper introduces an open-access dataset of CT scans for studying splenomegaly and its related conditions, aiming to support AI-based diagnostics and education.

Contribution

The paper presents a new, publicly available dataset of CT scans and patient data for splenomegaly and associated diseases.

Findings

The dataset includes 248 de-identified CT scans from adult patients with and without splenomegaly.

It covers a range of diseases such as liver cirrhosis, lymphoma, and thalassemia.

The dataset is freely accessible via Zenodo to support AI model development and medical education.

Abstract

Splenomegaly, defined as an abnormal enlargement of the spleen, is a critical radiological finding associated with a spectrum of serious health conditions, particularly liver disorders and hematologic malignancies. While Computed Tomography (CT) scans are commonly utilized in clinical settings to detect and assess the severity of splenomegaly, recent studies have demonstrated its potential to reveal new insights into the underlying causes of splenomegaly, such as liver cirrhosis and lymphoma, when integrated with machine and deep learning models. This enables patients to receive timely and appropriate treatment, thereby reducing the risk of complications. However, research in this area remains in its early stages, primarily due to the limited availability of real-world datasets required to develop and train robust Artificial Intelligence (AI)-based models.

This article introduces a diverse dataset specifically curated for the study of splenomegaly and its associated findings. Collected at King Abdullah University Hospital in northern Jordan, the dataset includes 248 de-identified CT scans and patient data from 248 adult subjects (42% female), with a mean age of 48.2 years (ranging from 18 to 91 years). It encompasses cases both with and without splenomegaly, covering a range of diseases such as liver cirrhosis, lymphoma, leukemia, myeloproliferative neoplasms, and thalassemia. This work presents an open-access dataset dedicated to splenomegaly, aiming to address the lack of publicly available resources in this area. Furthermore, it supports the development of AI-driven diagnostic models for detecting splenomegaly, evaluating its severity, and identifying its potential causes. Additionally, it serves as an educational resource for medical students studying splenomegaly. The dataset is freely accessible via the Zenodo data repository, providing a valuable foundation for further research and advancements in diagnostic applications.

Figures