MétaCan
Menu
Back to cohort
Record W2999124253

Automatic pupil detection using off-axis iris images for alignment guidance in fundus cameras

2019· article· en· W2999124253 on OpenAlexaboutno aff
Poojan Dave, Andrew Wei, David S. Nolan, Simon Stock, Jing Guo, Angelina Covita, Michael Chen, Jochen Straub, Mary K Durbin, Niranchana Manivannan

Bibliographic record

VenueInvestigative Ophthalmology & Visual Science · 2019
Typearticle
Languageen
FieldMedicine
TopicGlaucoma and retinal disorders
Canadian institutionsnot available
Fundersnot available
KeywordsPupilArtificial intelligenceComputer scienceComputer visionIRIS (biosensor)Fundus (uterus)Minimum bounding boxDetectorBiometricsOpticsImage (mathematics)MedicinePhysicsOphthalmology
DOInot available

Abstract

fetched live from OpenAlex

Purpose : The purpose of this study is to create a pupil tracking algorithm to find the center of the pupil within 400 micrometers of ground truth (manual annotations) in non-mydriatic external eye images. Pupil detection is crucial for automation and alignment guidance, which can improve the quality of fundus image acquisitions. Methods : In the ultra-widefield fundus imaging system CLARUSTM 500 (ZEISS, Dublin, CA), two iris cameras provide an off-axis view of the patient’s eye along with the position of the pupil within the field of view. In this retrospective study, we used 654 external eye images (pixel size: 320x240) of non-mydriatic pupils (<3.5 mm pupil size) from 29 subjects. Manual annotations of pupil boundary and the center were marked by an expert grader. The dataset is divided into training (534 images from 18 subjects) and testing sets (120 images from 11 subjects). Fig 1 shows the flowchart of the proposed pupil detection algorithm. The algorithm consists of two blocks: 1) coarse region-of-interest (ROI) finder and 2) fine-tuned pupil detector. Coarse ROI finder consisted of a single-shot detector (SSD) with 7 convolutional neural networks (CNN). A bounding box with the highest confidence score is used as the starting point for the fine-tuned detection. The Shootingstar algorithm is an extension of the Starburst pupil detection algorithm. The Shootingstar implementation shoots rays at five positions (the center and on the four corners of the bounding box). The Euclidean distances between the pupil centers determined by the algorithm in the test set were compared with the manual annotations. Results : Fig 2 shows the results of correct and incorrect detections from the proposed algorithm. The algorithm achieved an accuracy of 91.5% in tracking pupils within 400 micrometers. Incorrect results are usually caused by the patient blinking or being in the middle of a blink. The execution time of the algorithm is 57.4 ± 3.8 ms using an Intel® Core™ i7-6920HQ CPU@2.90GHz. Conclusions : The proposed algorithm provides a reliable solution for pupil detection for alignment guidance for fundus image capture. The algorithm can detect up to 17 frames per second and would be suitable for real-time pupil tracking. This abstract was presented at the 2019 ARVO Imaging in the Eye Conference, held in Vancouver, Canada, April 26-27, 2019.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame machine prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.

metaresearch head score (Codex)0.001
metaresearch head score (Gemma)0.002
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Bench or experimental · Consensus signal: none
GenreCandidate signal: Methods · Consensus signal: Methods
Teacher disagreement score0.002
Threshold uncertainty score0.006

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0010.002
Meta-epidemiology (narrow)0.0010.000
Meta-epidemiology (broad)0.0010.000
Bibliometrics0.0010.000
Science and technology studies0.0000.000
Scholarly communication0.0010.001
Open science0.0010.001
Research integrity0.0010.000
Insufficient payload (model declined to judge)0.0020.001

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.036
GPT teacher head0.350
Teacher spread0.314 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designBench or experimental
Domainnot available
GenreMethods

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations0
Published2019
Admission routes1
Has abstractyes

Explore more

Same venueInvestigative Ophthalmology & Visual ScienceSame topicGlaucoma and retinal disordersFrench-language works237,207