CRICVA Database
Description
The CRICVA dataset contains eye tracking data for a subset of 232 digitized Pap smear images of the CRIC Center (https://sites.google.com/view/centercric). The CRIC images were acquired with a Carl Zeiss microscope equipped with a Zeiss AxionCam MRc camera at 40x magnification. The images have 0.255 μm/pixel and resolution of 1392 x 1040 pixels (8-bit). The specimens were prepared via conventional Pap smears and contain cervical cells as well as other artifacts often collected as part of the exams. All cervical cell images were collected from SUS (Brazilian Universal Health System). ---- Detailed Information 1. Pap smear images The CRIC images were presented to cytopathologists at a resolution of 1280x1024 pixels, keeping the original aspect ratio by adding white pixel lines at the image bottom. Notice that the added white pixel lines have been removed from the attached images. 2. Eye tracking device We used an EyeLink 1000 system designed by SR Research Ltd., Mississauga, Canada, with a sampling rate of 1000 Hz on the right eye recording to register the visual attention of cytopathologists. 3. Subjects CRICVA has eye fixation maps from three cytopathologists, all of them with normal or corrected-to-normal vision via lens glasses. The cytopathologists have experience with cervical care and reading conventional Pap smear slides on clinical laboratory. 4. Methodology The attention data was collected by a task-driven experiment in which the expert interpreted each cervical cell image and marked the abnormal cells with mouse clicks. The participants had free time to analyze the images. ---- Database Organization CRICVA comprises samples for normal (Negative) and abnormal (ASC-US: Atypical Squamous Cells of Undermined Significance; ASC-H: Atypical Squamous Cells of High Significance; LSIL: Low-grade Squamous Intraepithelial Lesion; HSIL: High-grade Squamous Intraepithelial Lesion; and CA: Carcinoma) cases. Our dataset contains eight folders each being relative to a trial. In each folder, there are the cervical cell images (sub-folder: images), the eye fixations (sub-folder: fixation_locs), and the attention maps (sub-folder: fixation_maps). For each trial, we also provide a text file (labels_trial_xx.txt) with the following information: image_id, image_name, image_class For example, 1,011fda505d7e4af4b8cc57545343624d,ASC-US 2,02c7fb946ad5c5e5f9c1e1178c21fc92,ca More information can be found in the references (related links) below. ---- Special Thanks: Dr. José Soares de Andrade Júnior and Dr. Humberto de Andrade Carmona for authorizing the use of the eye tracking device at the Complex Systems Laboratory, Department of Physics, Universidade Federal do Ceará, Fortaleza, CE, Brazil. ---- Feedback on the dataset is welcome. The person to contact is Daniel Ferreira (daniels@ifce.edu.br)
Files
Steps to reproduce
1. Download the zip files from the CRICVA folder. 2. Unzip all folders within a container folder (i.e CRICVA) 3. Read the data appropriately. Notice that we use the same filename for each sample in all sub-folders. We arrange the eye fixations in .mat files (struct field name: fixLocs). You can use the loadmat function from scipy.io library (Python) to read these files.