Radiological scans: DICOM or NIfTI Radiology image annotations: NRRD or NIfTI for segmentation and JSON for localization (e.g., Bounding boxes) Pathological scans: TIFF or DICOM or iSyntax or NDPI Pathology image annotations: GeoJSON Clinical data (including molecular data and radiology/pathology reports): EXCEL

We check the consistency of image quality, imaging conditions, diagnosis names, radiology/pathology findings, clinical data and annotation content. When necessary, we exclude inappropriate cases, such as those with noise or missing data, and standardize the data format. Annotations are performed by specialists such as radiologists, pathologists, orthopedic surgeons, or radiologic technologists with expertise in medical imaging, depending on the disease and modality. Depending on the difficulty and requirements, primary annotation may be performed by technologists and then reviewed by specialist physicians. Annotation types vary by dataset and may include segmentation, bounding boxes, classification labels, diagnosis labels, and findings information.

Yes. We provide datasets after removing or processing information that could identify individuals, such as patient IDs, names, dates of birth, examination dates, and other identifiers contained in clinical data, metadata, and image data. If personal information is embedded in the images, or if the images contain facial features or other information that may lead to individual identification, we perform masking or other appropriate processing as necessary while minimizing the impact on analysis.

Yes. The data are obtained and provided in a form suitable for secondary use, based on appropriate procedures at each medical institution. For each dataset, we confirm the necessary acquisition conditions, such as ethics review, opt-out procedures, informed consent, and internal approval at the medical institution, before providing the data.

Yes. The datasets can be used for R&D purposes, including medical AI, AI-driven drug discovery, medical devices, and clinical research. They can also be used in collaboration with external contractors, for regulatory submissions, academic publications, and marketing materials. We can also provide datasets to companies outside Japan.

Pricing varies based on data volume, rarity, and whether it includes clinical data, molecular data, or annotation. Please contact us for further details.

We issue an invoice for payment via bank transfer through Wise or SMBC Direct.

For custom datasets, you can specify requirements such as disease, modality, body part, number of cases, imaging conditions, manufacturer, slice thickness, clinical data, molecular data, annotation type, and exclusion criteria. In addition to extracting cases from existing datasets, we can also discuss additional annotations or the construction of new datasets based on your requirements.

Yes, we offer samples comprising 2-5 cases to aid in evaluating our scans, DICOM headers, clinical information, and annotation methodology against your specific needs.

For existing datasets, delivery is typically possible within approximately 1–2 weeks. For custom datasets or datasets requiring additional annotation, the delivery timeline depends on the number of cases, disease, modality, annotation requirements, and availability of clinical data. As a general guideline, delivery typically takes approximately 1–2 months. In principle, data are provided as password-protected files through secure cloud storage such as Box.

Yes, we provide medical feedback, customized AI development, and regulatory support for entering the Japanese medical device market. Our team comprises medical AI and MLOps engineers, software developers, diagnostic radiologists, radiation oncologists, pathologists, and dataset managers. Notably, our CEO, Changhee Han (Kallis), is a leading young medical AI researcher in Japan with over 2,000 first-author citations.