Open Data / Repository

The RSNA Intracranial Hemorrhage Detection dataset is one of the largest publicly available head CT datasets for intracranial hemorrhage detection. It contains over 25,000 CT examinations and approximately 874,000 annotated CT slices collected from multiple institutions. Expert neuroradiologists labeled each slice for the presence of intracranial hemorrhage and its five major subtypes: epidural, subdural, subarachnoid, intraparenchymal, and intraventricular hemorrhage. The dataset is widely used for developing and benchmarking AI and deep learning algorithms for automated brain hemorrhage detection and classification.

The EEG in Schizophrenia dataset is a public resting-state EEG dataset released by the Institute of Psychiatry and Neurology, Warsaw. It contains EEG recordings from 14 patients with paranoid schizophrenia and 14 healthy controls. Signals were acquired in the eyes-closed resting state using the international 10–20 EEG system with 19 channels at a 250 Hz sampling rate. The recordings are provided in EDF format under the CC0 license and are widely used for schizophrenia detection, functional connectivity analysis, brain network research, and machine learning applications.

OpenNeuro DS003944 is a public resting-state EEG dataset containing recordings from 82 participants, including individuals with First Episode Psychosis (FEP) and healthy controls. EEG signals were acquired using a 64-channel system with a sampling rate of 1000 Hz and are organized in the BIDS standard. The dataset also includes clinical assessments, cognitive measures, medication information, and demographic metadata, making it a valuable benchmark for schizophrenia detection, brain connectivity analysis, and machine learning research.

The RSNA Pediatric Bone Age Challenge Dataset is one of the world's most widely used public datasets for automated bone age assessment. Released for the 2017 RSNA AI Challenge, it contains 14,236 left-hand pediatric radiographs with expert-annotated skeletal age (in months) and patient sex. The dataset was curated from Children's Hospital Colorado and Lucile Packard Children's Hospital Stanford and divided into 12,611 training, 1,425 validation, and 200 test images. It is the benchmark dataset for developing and evaluating machine learning and deep learning models for skeletal maturity assessment and pediatric bone age estimation.

LIDC-IDRI is one of the most widely used public datasets for lung CT analysis and lung cancer research. It contains thoracic CT scans from 1,010 patients with pulmonary nodules independently annotated by four experienced thoracic radiologists. The dataset includes DICOM CT images, XML annotation files, lesion characteristics, nodule size measurements, and malignancy ratings. It serves as a benchmark for pulmonary nodule detection, segmentation, classification, radiomics, computer-aided diagnosis (CAD), and deep learning research.

CBIS-DDSM is a curated and standardized version of the Digital Database for Screening Mammography (DDSM), released by The Cancer Imaging Archive (TCIA). It contains mammography images in DICOM format with lesion ROI masks, segmentation annotations, bounding boxes, pathology labels, BI-RADS assessments, breast density, patient age, and other clinical metadata. The dataset includes 891 mass cases and 753 calcification cases from 1,566 patients and is widely used as a benchmark for breast cancer detection, lesion segmentation, classification, radiomics, computer-aided diagnosis (CAD), and deep learning research.

The Cancer Imaging Archive (TCIA) is one of the world's largest open-access repositories of de-identified cancer medical images, funded by the U.S. National Cancer Institute (NCI). It hosts hundreds of imaging collections covering multiple cancer types, including CT, MRI, PET, and digital pathology, along with associated clinical, genomic, segmentation, and radiomics data. TCIA is widely used for cancer imaging research, medical image analysis, radiomics, and artificial intelligence development.

The Human Connectome Project (HCP) is a landmark neuroimaging initiative designed to map the structural and functional connectivity of the human brain. It provides high-resolution Structural MRI, Diffusion MRI, resting-state and task fMRI, behavioral, cognitive, and demographic data, together with standardized preprocessing pipelines. HCP has become one of the most widely used resources for connectomics, neuroimaging, machine learning, and brain network analysis. Data are available to researchers after registration and agreement to the HCP Data Use Terms.

The Parkinson's Precision Medicine Initiative (PPMI) is a large, longitudinal, open-access study designed to identify biomarkers of Parkinson's disease onset and progression. The repository includes clinical assessments, MRI, DaTSCAN imaging, genetic and multi-omics data, biospecimen analyses (CSF, blood, urine), wearable sensor data, and longitudinal follow-up from more than 5,000 participants. Access is available to qualified researchers after registration and acceptance of a Data Use Agreement (DUA).

ADNI is a large longitudinal neuroimaging dataset for Alzheimer's disease research. It contains MRI, PET, DTI, cognitive assessments, genetic information, CSF and blood biomarkers, and clinical data collected from more than 2,000 participants. The dataset is widely used in medical image analysis, machine learning, and Alzheimer's disease progression studies. Access is free for researchers after registration and approval of a Data Use Agreement (DUA).

The Autism Brain Imaging Data Exchange (ABIDE ) represents the first ABIDE initiative. Started as a grass roots effort, ABIDE involved 17 international sites, sharing previously collected resting state functional magnetic resonance imaging (R-fMRI), anatomical and phenotypic datasets made available for data sharing with the broader scientific community. This effort yielded 1112 dataset, including 539 from individuals with ASD and 573 from typical controls (ages 7-64 years, median 14.7 years across groups).