The Augmented IQ-OTHNCCD Lung Cancer Dataset

This repository contains the minimal reproducibility dataset supporting the research work titled:
"Contrastive Attention Learning Network Based Hybrid Model for Lung Cancer Detection in CT Images"

The study focuses on automated lung cancer classification from computed tomography (CT) images using deep learning and feature fusion techniques.

The original CT images were obtained from the publicly available IQ-OTHNCCD Lung Cancer Dataset. In accordance with repository and journal policies, raw CT images are not redistributed in this repository.

Instead, this repository provides the essential materials required to reproduce the findings reported in the manuscript, including:

• Patient-wise training and testing split indices
• Detailed preprocessing pipeline description
• Extracted deep learning feature representations
• Extracted radiomic feature representations
• Corresponding class labels

The patient-wise split strategy ensures the prevention of data leakage and supports reliable model evaluation. The materials provided here enable transparent and reproducible validation of the proposed methodology while respecting data ownership and ethical considerations.

Researchers can reproduce the results by downloading the original IQ-OTHNCCD dataset from its official source and applying the split indices and preprocessing steps described in this repository.

本仓库包含支撑下述研究工作的最小可复现数据集，该研究的标题为："基于对比注意力学习网络的混合模型用于CT图像肺癌检测"

本研究聚焦于利用深度学习与特征融合技术，实现计算机断层扫描（Computed Tomography，简称CT）图像的自动化肺癌分类。

本数据集的原始CT图像取自公开可用的IQ-OTHNCCD肺癌数据集。根据本仓库及期刊的相关政策，本仓库不重新分发原始CT图像。

为此，本仓库提供了复现论文中报道的研究结果所需的核心材料，具体包括：
• 按患者划分的训练与测试集索引
• 详细的预处理流程说明
• 提取的深度学习特征表征
• 提取的放射组学特征表征
• 对应的类别标签

按患者划分数据集的策略可有效避免数据泄露，保障模型评估的可靠性。本仓库提供的材料可在尊重数据所有权与伦理准则的前提下，实现所提方法的透明化与可复现验证。

研究人员可从官方渠道下载原始IQ-OTHNCCD数据集，并应用本仓库提供的数据集划分索引与预处理步骤，即可复现本研究的实验结果。