Acetabular-vision hip developmental dysplasia’s (AV-DDH)
Description
Acetabualr-vision hip developmental dysplasia’s (AV-DDH) Description: AV-DDH dataset is a comprehensive collection of 2417 raw X-ray images and their corresponding annotated labels for diagnosing developmental dysplasia of the hip (DDH). The dataset has been meticulously curated to aid in developing and validating machine learning models for DDH classification and diagnosis using hip X-rays. Each image in this dataset is paired with relevant demographic data, including age, gender, and the Acetabular Index in degrees for both hips. The annotations for the images were performed by a team of eight expert doctors and further reviewed by four orthopedic surgeons with extensive experience. Additionally, the dataset includes: • Unlabeled Data: A file containing raw, unannotated X-ray images. • Labeled Data: A file containing augmented, annotated X-ray images. The labeled data includes essential diagnostic information, including the Acetabular Index angle measurements for both the right and left hips, which specialized medical professionals have verified. The labeled dataset is organized as follows: • Training Set: 16,205 images • Validation Set: 463 images • Test Set: 926 images Content: • Images: 4630 raw X-ray images in PNG format depicting hip X-rays from infants under three years of age. • Annotations: Each image is annotated with the locations of critical anatomical features relevant to DDH, stored in both JSON and YAML formats. • Demographics: Each record includes demographic data such as age, gender, and the Acetabular Index in degrees for both the left and right hips. • Angle Measurements: The dataset includes angle measurements for the Acetabular Index for both hips, verified and labeled by a team of medical professionals. The dataset was collected from the radiology department at the Jordan University Hospital and is enriched with diagnostic information confirmed by orthopedic specialists. The images were gathered following strict inclusion criteria and are free from duplicates, ensuring the integrity of the dataset. Use Case: The AV-DDH dataset is specifically designed to aid in the development of machine learning models, especially deep learning-based approaches, to automate the diagnosis and analysis of DDH using hip X-ray images. It provides a high-quality, reliable source of labeled data that can be used for training, testing, and validating AI models aimed at improving the early detection and diagnosis of DDH, which is crucial for timely treatment and preventing long-term complications such as osteoarthritis and hip replacement surgery. Data Format: • Images: Raw X-ray images in JPG format. • Annotations: Annotations in JSON and YAML formats. • Additional Information: Demographic data in an Excel sheet. Keywords: Developmental Dysplasia of the Hip (DDH), X-ray, Machine Learning, Medical Imaging, Deep Learning, Data Annotation, Acetabular Index, AI, Dataset, Hip Imaging, Radiology, Diagnosis.
Files
Steps to reproduce
After applying the inclusion criteria, a collection of about 4630 X-ray images that were captured using equipment from Philips, GE, and Canon, with an exposure dose of 1 to 2.5 μGy/m² was downloaded from Synapse Radiology PACS, and then uploaded to a first file named “un-labeled data” in Google Drive, each image was named by the patient's ID. Also, there was an Excel file with important patient data like the patient's gender, age, and right and left Acetabular Index in Degrees for both the right and left sides. Data was labeled by Roboflow by an expert team.