Tumor Segmentation

The BraTS 2015 dataset is a dataset for brain tumor image segmentation. It consists of 220 high grade gliomas (HGG) and 54 low grade gliomas (LGG) MRIs. The four MRI modalities are T1, T1c, T2, and T2FLAIR. Segmented “ground truth” is provide about four intra-tumoral classes, viz. edema, enhancing tumor, non-enhancing tumor, and necrosis. Source: Brain MRI Tumor Segmentation with Adversarial Networks Image Source: https://sites.google.com/site/braintumorsegmentation/home/brats2015

LiTS17 is a liver tumor segmentation benchmark. The data and segmentations are provided by various clinical sites around the world. The training data set contains 130 CT scans and the test data set 70 CT scans. Image Source: https://arxiv.org/pdf/1707.07734.pdf

Head and Neck Tumor Segmentation

Introduced by Da et al. in DigestPath: a Benchmark Dataset with Challenge Review for the Pathological Detection and Segmentation of Digestive-System ### Grand-Challenge Page ### 1. Signet ring cell dataset Signet ring cell carcinoma is a type of rare adenocarcinoma with poor prognosis. Early detection of such cells leads to huge improvement of patients' survival rate. However, there is no existing public dataset with annotations for studying the problem of signet ring cell detection. This dataset has positive samples and negative samples. Training positive samples contain 77 images from 20 WSIs, with cell bounding boxes written in xml. Training negative samples contain 378 images from 79 WSIs.These negative WSIs have no signet ring, but could contain other kinds of tumor cells. Each signet ring cell is labeled by experienced pathologists with a rectangle bounding box tightly surrounding the cell. Each image is of size 2000X2000. The training images are from 2 organs, including gastric mucosa and intestine. Because of the difficulty of manual annotation, there exist some signet ring cells who are missed by pathologists. In other words, this dataset is a noisy dataset with its positive images not fully annotated. All whole slide images were stained by hematoxylin and eosin and scanned at X40. ### 2. Colonoscopy tissue segment dataset Colonoscopy pathology examination can find cells of early-stage colon tumor from small tissue slices. Pathologists need to daily examine hundreds of tissue slices, which is a time consuming and exhausting work. Here we propose a challenge task on automatic colonoscopy tissue segmentation and screening, aiming at automatic lesion segmentation and classification of the whole tissue (benign vs. malignant). This dataset has positive samples and negative samples. Training positive samples contain 250 images of tissue from 93 WSIs, with pixel-level annotation in jpg format, where 0 means background and 255 for foreground (malignant lesion). You could simply get binary mask by a threshold 128. Training negative samples contain 410 images of tissue from 231 WSI. This negative images have no annotation because they don't have any malignant lesion. The average size of all images are of 5000x5000 pixels, some of them are extremely huge. We will also provide another 152 patients' 212 tissues as the testing set, in which 90 images from 65 patients contain lesion. All whole slide images were stained by hematoxylin and eosin and scanned at X20. Sign the DATABASE USE AGREEMENT first and download the dataset at the homepage! Da Q, Huang X, Li Z, et al. DigestPath: a Benchmark Dataset with Challenge Review for the Pathological Detection and Segmentation of Digestive-System[J]. Medical Image Analysis, 2022: 102485. （https://doi.org/10.1016/j.media.2022.102485）

BRATS 2016 is a brain tumor segmentation dataset. It shares the same training set as BRATS 2015, which consists of 220 HHG and 54 LGG. Its testing dataset consists of 191 cases with unknown grades. Image Source: https://sites.google.com/site/braintumorsegmentation/home/brats_2016

BRATS 2014 is a brain tumor segmentation dataset. Source: Learning Fixed Points in Generative Adversarial Networks: From Image-to-Image Translation to Disease Detection and Localization Image Source: http://people.csail.mit.edu/menze/papers/proceedings_miccai_brats_2014.pdf

A whole-body FDG-PET/CT dataset with manually annotated tumor lesions (FDG-PET-CT-Lesions) 1,014 studies (900 patients)

CSAW-S is a dataset of mammography images which includes expert annotations of tumors and non-expert annotations of breast anatomy and artifacts in the image. Source: CSAW-S

The ULS23 test set contains 725 lesions from 284 patients of the Radboudumc and JBZ hospitals in the Netherlands. It is intended to be used to measure the performance of 3D universal lesion segmentation models for Computed Tomography (CT). To prepare the data, radiological reports from both participating institutions where searched using NLP tools identifying patients with measurable target lesions, indicating that these lesions were clinically relevant. A random sample of patients was selected, 56.3% of which were male and with diverse scanner manufacturers. The lesions were annotated in 3D by expert radiologists with over 10 years of experience in reading oncological scans. ULS23 is an open benchmark, and we invite ongoing submissions to advance the development of future ULS models.

https://drive.google.com/file/d/1X_JTfD8Ch-IxmG5VHtKk_xGZT336Fl1Q/view?usp=drive_link

Highlights • Publicly available dataset with 82 H&E stained images of frozen sections. • Images are acquired on 19 patients with metastatic colon cancer in a liver. • Originally stained and two stain-normalized sets of images included • Pixel wise ground truths provided by seven domain experts.

This is the dataset released along with the publication: CUTS: A Deep Learning and Topological Framework for Multigranular Unsupervised Medical Image Segmentation [ArXiv] [MICCAI] [GitHub]

A dataset of abdominal CT studies in NifTi format from the open-source medical data repository Medical Decathlon was utilized. To expedite the partitioning process, the MONAILabel plugin of the MONAI framework within the 3D Slicer program was employed. A radiologist with 15 years of experience conducted a validation process, wherein the boundaries of the colon markup were verified on each slice. The existing colorectal cancer markings in the dataset remained unaltered. Validation by a radiologist reduced the size of the validated dataset to 122 studies. In this case, the 122 studies were categorized into three subsets based on the quality of the data: The "good" subset comprises 100 studies, while the "bad" subset contains 17 cropped studies (in which the entire colon is not visible on the image). The "bad" subset comprises five studies. Two of these studies were of poor quality and could not identify the entire colon. Two further studies involved colon stomas following surgery, while one study involved a hernia. Four studies were excluded from the set due to broken images, with axial plane images duplicated or the slice order mixed

Unique radiogenomic dataset from a Non-Small Cell Lung Cancer (NSCLC) cohort of 211 subjects. The dataset comprises Computed Tomography (CT), Positron Emission Tomography (PET)/CT images, semantic annotations of the tumors as observed on the medical images using a controlled vocabulary, segmentation maps of tumors in the CT scans, and quantitative values obtained from the PET/CT scans. Imaging data are also paired with gene mutation, RNA sequencing data from samples of surgically excised tumor tissue, and clinical data, including survival outcomes.

The ULS23 training dataset contains 38,693 diverse lesions from chest-abdomen-pelvis CT examinations. For the challenge, we introduced two novel 3D annotated datasets targeting lesions in the pancreas and bones, which are traditionally challenging to segment. Additionally, we aggregate 10 publicly available datasets with a lesion segmentation component into a single, easily accessible data repository. The dataset consists of fully 3D-segmented (5.633) and partially 2D-annotated lesions (33.060). Imaging data is provided to as volumes-of-interest (VOI's) of 256x, 256y, 128z voxels in the original scan spacing. The VOI's were sampled such that there always is a lesion voxel in the middle of the volume, simulating the lesion being identified by a radiologist or lesion detection model. This voxel in the center of the volume was selected randomly from within the lesion mask. Within each VOI, there is only one annotated lesion. Any masks representing nearby lesions not connected to the central lesion were removed.