CRIC Cervix Collection

The CRIC Cervix Collection is a searchable image database — currently with 400 images (1,376 × 1,020 pixels) curated from conventional Pap smears, with manual classification of 11,534 cells —  that makes digital cell image collections available for reproducible research and FAIR machine learning, with the potential to advance current efforts in training and testing machine learning algorithms for the automation of tasks as part of the cytopathological analysis in the routine work of laboratories.

This free and open-source digital platform consists of real images of conventional cervical-vaginal cytology, obtained from the Pap smear, which were classified by specialists according to the Bethesda System (standardized nomenclature). The CRIC Cervix Collection is the largest available database of cervical cells obtained from conventional smears, which stands out in terms of both the total number of cells (11,534 identified and manually classified) and the number of cells per class of lesion, thus providing greater variety from pre-neoplastic and neoplastic lesions to conventional cervico-vaginal cytology.

The classifications were performed and certified by cytopathologists from the Laboratory of Cytology of the Department of Clinical Analysis, School of Pharmacy, and the digital platform was implemented by experts from the Department of Computer Science, both at the Federal University of Ouro Preto (UFOP) in Brazil. Other contributors have come from three main institutions: Federal University of Ceara, University of California, Berkeley, and Lawrence Berkeley National Laboratory.

The database has been used by BIDS Research Affiliate Daniela Ushizima and colleagues from the Center for Recognition and Inspection of Cells (CRIC) – in Brazil (UFOP, UFC, IFCE and UFPI) and abroad – to develop machine and deep learning methodologies to detect, identify and classify lesions among cervical cells, such as in this other recent science article by the CRIC team. Enabling predictive modeling using biomedical data and computer-aided diagnosis are some of the goals of CRIC. Watch this video for a quick journey into the CRIC way of looking at cancer cells.

CRIC searchable image database as a public platform for conventional pap smear cytology data 
June 10, 2021  |   Nature Scientific Data

A Hierarchical Feature-Based Methodology to Perform Cervical Cancer Classification
April 30, 2021  |   MDPI Applied Sciences

Video (1:40 mp4)CRIC Searchable Image Database 

BIDS Affiliates

Daniela Ushizima

Computational Research Division, CAMERA, LBNL
Research Affiliate