2026-10-13T18:52:29Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33986374 2026-09-26T20:01:17Z category_26299 item_type_3 month_year_09_2026

NMPA-MedDevice: A Multimodal Regulatory Dataset of 52,000 Chinese Medical Devices for Risk Classification Research Yu Han (251741) Aaron Ceross (25114983) Jeroen H. M. Bergmann (25114986) Biomedical engineering not elsewhere classified Medical device informatics Regulatory science Unique Device Identification Risk classification Multimodal classification <p>This deposit replaces figshare item 31322272 (DOI 10.6084/m9.figshare.31322272), whose image-linked subset contained two defects found by the authors: 31 of the 39 Class I rows did not resolve to entries in the official NMPA UDI release, and images were shared between devices that are not the same product. The subset has been rebuilt rather than patched, and further problems in the earlier files were corrected (see codebook.pdf).</p> <p>The dataset covers the 66,472 records of the NMPA UDI full release of 1 July 2024; a cleaned corpus of 52,251 of them with a ninth-character risk label, the recommended route-derived label and a non-standard-identifier flag; a rebuilt image-linked subset of 1,013 devices (Class I/II/III = 50/495/468, route-derived) with precomputed BERT text and EfficientNet-B5 image embeddings and UDI-grouped five-fold cross-validation indices; an external temporal validation set of 300 later-registered devices; and records of every device attempted during sampling.</p> <p>This deposit contains no NMPA registry records. The NMPA website statement asks that its content not be republished in its original form and that content supplied by other units be republished only with their authorisation, so records are identified by their UDI-DI and rebuild_from_official.py retrieves their registry fields from an official NMPA full release (https://udi.nmpa.gov.cn/download.html). The July 2024 release is no longer offered by the NMPA; rebuilding from the release of 1 May 2026 retrieved 99.3% of the records, 92.0% with an unchanged description. The released labels, features and folds were computed on the July 2024 records and reproduce every table in the accompanying paper without the rebuild.</p> <p>No image files, and no links to them, are redistributed; images are represented by their EfficientNet-B5 embeddings, and images.csv records each image's source type, website, dimensions and match level. No two devices share an image.</p> <p>Licence: CC BY 4.0 applies to the data files and codebook in this deposit; the scripts are under the MIT licence (LICENSE in the code archive). Registry fields retrieved with rebuild_from_official.py come from the NMPA UDI database and remain subject to the NMPA's terms, and the images described in images.csv remain the copyright of their owners.</p> <p>Files: NMPA-MedDevice_data.zip (data), NMPA-MedDevice_code.zip (scripts; unzip both into one folder), and codebook.pdf (field definitions, usage and the differences from the superseded deposit; a plain-text copy is in the data archive).</p> 2026-09-26T20:01:17Z Dataset Dataset 10.6084/m9.figshare.33986374.v2 https://figshare.com/articles/dataset/NMPA-MedDevice_A_Multimodal_Regulatory_Dataset_of_52_000_Chinese_Medical_Devices_for_Risk_Classification_Research/33986374 CC BY 4.0