CT Volume Samples for Lung Adenocarcinoma Classification

Published: 24-07-2020| Version 2 | DOI: 10.17632/r3tbsgtpzg.2
Yuanli Feng


Major content: This dataset consists of lung adenocarcinoma samples locally cropped from 3D CT images. Samples are discriminated by 3 pathological categories, respectively Atypical Adenomatous Hyperplasia (AAH), Adenocarcinoma In Situ (AIS), and Minimally Invasive Adenocarcinoma (MIA). This dataset totally contains 21 samples of AAH, 444 samples of AIS, and 585 samples of MIA. Usage: Samples are stored with .npy files, as can be loaded in Python by the package of NumPy, and introduced in deep learning on the task of lung adenocarcinoma classification. Each sample is a 3D volume with 128*128*128 voxel resolution, where the lung adenocarcinoma is located at the center of the volume, and the volume can be cropped with any scale (e.g. 64*64*64) to fit the input size of 3D convolutional neural networks.