Kidney Ultrasound Image Dataset (UKUID_REV01) A Clinical Ultrasound Dataset of Kidney Images for Machine Learning Research

Dataset 1: Kidney Ultrasound Image Dataset (UKUID_REV01) A Clinical Ultrasound Dataset of Kidney Images for Machine Learning Research The UKUID_REV01 dataset contains ultrasound images of human kidneys acquired from patients examined at Kosar Hospital, which is affiliated with Semnan University of Medical Sciences in Semnan, Iran. The data were collected as part of a clinical imaging study designed to support the development of artificial intelligence methods for renal image analysis. A total of 47 patients participated in the study, including 11 male and 36 female participants between 30 and 70 years of age. The weight range of the participants was between 60 kg and 100 kg. None of the participants had a prior history of kidney surgery, kidney transplantation, or congenital renal anomalies such as malrotation, horseshoe kidney, or ectopic kidney. All participants provided written informed consent before image acquisition. Ultrasound examinations were performed using a Vinno G8 Ultrasound System equipped with a 3.5 MHz convex probe. The imaging system was calibrated monthly to maintain measurement accuracy and image consistency. All scans were conducted by a single experienced sonographer with more than ten years of clinical experience. Imaging sessions were carried out in a controlled environment with a room temperature maintained at approximately 24 ± 2 °C. To improve imaging conditions, participants performed deep breathing exercises prior to scanning. The right kidney was scanned while the patient was in the supine position, whereas the left kidney was imaged in a left semi-lateral position. Each examination lasted approximately ten minutes, and multiple frames were captured during a sweeping motion along the mid-sagittal plane of the kidney until both poles were clearly visualized. Ultrasound beams were oriented perpendicular to the long axis of the kidney to improve visualization of the corticomedullary boundary. Standardized imaging parameters were applied across all examinations. The imaging depth was set to 15 cm with no zoom, and the gain was adjusted based on patient body habitus. Imaging used harmonic mode (CEVINNO HAR) with a transducer frequency of 4.5 MHz, AP100% power, DG46% gain, and a frame rate of approximately 25.3 Hz. The S1-8C curved array probe and ABD preset were employed to ensure consistent visualization of renal structures. Safety indices remained within recommended limits with a Mechanical Index (MI) of 1.4 and a Thermal Index (TIS) of 0.7. Across the dataset, the average kidney length was approximately 10.5 ± 1.2 cm and the average parenchymal thickness was 1.5 ± 0.3 cm. Images with severe bowel gas artifacts were excluded to maintain quality. Clinical indications for the scans included evaluation of flank pain, urinary symptoms, routine examinations, and physician-requested renal assessments. This dataset contains 183 ultrasound images without manual annotations. The images provide clear visualization of renal parenchyma, collecting systems, and surrounding anatomical structures. The dataset is intended to support research in medical image processing, ultrasound analysis, and machine learning approaches for kidney assessment. Keywords kidney ultrasoundrenal imagingmedical imaging datasetultrasound imagingartificial intelligence in healthcarekidney analysisrenal ultrasound datasetmedical image segmentationdiagnostic ultrasoundAI medical dataset README File (Dataset Description) Overview This repository contains ultrasound images of human kidneys collected for research on artificial intelligence and medical image analysis. The dataset was acquired from the Radiology Department of Kosar Hospital, affiliated with Semnan University of Medical Sciences. The dataset is designed to support the development and evaluation of machine learning models for renal ultrasound analysis. The dataset includes two subsets: an unlabeled image dataset and a smaller expert-annotated dataset intended for supervised learning and benchmarking tasks. Dataset Composition Dataset A – Unlabeled Kidney Ultrasound Images Number of images: 183 Format: Ultrasound grayscale images Source: Clinical renal ultrasound examinations Annotation: None These images capture kidney structures including renal parenchyma and collecting systems and can be used for unsupervised learning, preprocessing research, or algorithm development. Data Acquisition Ultrasound scans were performed using a Vinno G8 Ultrasound System equipped with a convex probe operating at 3.5 MHz. The system was calibrated monthly to maintain imaging accuracy. A total of 47 patients participated in the study: 11 males 36 females Age range: 30–70 years Weight range: 60–100 kg All scans were performed by a trained sonographer with over 10 years of clinical experience. Each examination lasted approximately 10 minutes, producing 6–8 frames per kidney under optimal abdominal conditions to reduce bowel gas artifacts. Imaging Parameters Standardized ultrasound parameters were used across all patients: Imaging depth: 15 cm Imaging mode: Harmonic imaging Transducer frequency: 4.5 MHz Frame rate: ~25.3 Hz Power level: AP100% Gain: DG46% Probe: S1-8C curved array Safety indices followed clinical guidelines: Mechanical Index (MI): 1.4 Thermal Index (TIS): 0.7 Patient Positioning To improve visualization: Right kidney scanned in supine position Left kidney scanned in left semi-lateral position A sweeping motion along the kidney’s mid-sagittal plane was used until both poles were visible. Inclusion Criteria Participants included adult patients undergoing renal ultrasound evaluation for clinical indications such as: Flank pain Urinary symptoms Routine renal evaluation Physician-requested kidney examination Exclusion Criteria The following conditions were excluded to maintain imaging consistency: Kidney surgery history Kidney transplantation Congenital anomalies (horseshoe kidney, ectopic kidney, malrotation) BMI greater than 35 Limitations Patients with high BMI were excluded due to ultrasound penetration limitations, which may restrict the dataset’s representation of obese populations. Research Applications This dataset can support research in: medical image segmentation ultrasound image processing AI-based kidney analysis diagnostic support systems deep learning in medical imaging Ethical Considerations All imaging procedures were conducted following written informed consent from participants. The dataset was anonymized before release to ensure patient privacy. Brief This dataset provides a collection of kidney ultrasound images intended to support research in artificial intelligence and medical image analysis. The data were collected at Kosar Hospital, affiliated with Semnan University of Medical Sciences, during routine renal ultrasound examinations. A total of 47 patients participated in the study, including 11 males and 36 females between the ages of 30 and 70 years. Ultrasound imaging was performed using a Vinno G8 Ultrasound System equipped with a convex probe operating at 3.5 MHz. All scans were conducted by an experienced sonographer under standardized imaging conditions to ensure consistent image quality. The dataset consists of two subsets designed for different research purposes. The first subset contains 183 ultrasound images without manual annotations, which can be used for image processing research, unsupervised learning, and algorithm development. The second subset includes 91 ultrasound images that were manually labeled and verified by a medical imaging expert, enabling supervised machine learning applications such as segmentation, classification, and detection of renal structures. Images were acquired using harmonic ultrasound imaging with standardized parameters to improve tissue contrast and reduce artifacts. All images were anonymized before publication to protect patient privacy. By providing both unlabeled and expert-annotated ultrasound data, this dataset aims to facilitate the development and benchmarking of artificial intelligence algorithms for kidney ultrasound interpretation and computer-assisted diagnostic systems. Ethical Considerations All imaging procedures were conducted following written informed consent from participants. The dataset was anonymized before release to ensure patient privacy.  Dataset Structure (Zenodo + GitHub) To make your dataset more useful and easier to cite, organize it in the following structure before uploading. Kidney-Ultrasound-Dataset/│├── README.md├── LICENSE├── dataset_description.pdf├── citation.txt│├── Dataset_A_Unlabeled/│   ├── image_001.png│   ├── image_002.png│   ├── ...│   └── image_183.png│├── Dataset_B_Labeled/│   ├── images/│   │   ├── img_001.png│   │   ├── img_002.png│   │   └── ...│   ││   └── labels/│       ├── label_001.png│       ├── label_002.png│       └── ...│└── metadata/    ├── patient_info.csv    └── acquisition_parameters.txt   4. Suggested License For medical datasets, the most common license is: Creative Commons Attribution 4.0 International (CC BY 4.0) Copyright © 2026 Ata Jahangir Moshayedi This dataset is distributed under the Creative Commons Attribution 4.0 International License (CC BY 4.0). Ethics Statement for Medical Dataset Release The kidney ultrasound images included in this dataset were collected during routine clinical examinations at Kosar Hospital, affiliated with Semnan University of Medical Sciences. All imaging procedures were performed following standard clinical protocols by qualified medical personnel using a Vinno G8 Ultrasound System. Participation in the study was voluntary, and written informed consent was obtained from all participants prior to image acquisition. The study followed the ethical principles outlined in the Declaration of Helsinki regarding research involving human subjects. Participants were informed that their anonymized imaging data could be used for scientific research and the development of artificial intelligence tools for medical imaging. To ensure patient privacy and confidentiality, all data were fully anonymized before being included in the dataset. Any personally identifiable information, including patient names, identification numbers, dates of birth, and other sensitive metadata, was removed. The released images contain no information that can be used to identify individual patients. The dataset is provided strictly for academic and research purposes, particularly for the development and evaluation of machine learning and artificial intelligence algorithms in medical imaging. Users of the dataset are expected to comply with applicable ethical guidelines, data protection regulations, and responsible research practices when using these data. The dataset does not include images from patients with prior kidney surgery, kidney transplantation, or congenital anatomical abnormalities, and individuals with a body mass index greater than 35 were excluded to ensure imaging quality. While these criteria helped standardize the dataset, they may limit the generalizability of models trained exclusively on this data. By accessing and using this dataset, researchers agree to use the data responsibly and acknowledge the original source in any resulting publications or derived works.