Validation of an artificial intelligence and scoring system for the detection of subclinical keratoconus in Asian eyes

Session Details

Session Title: Imaging

Session Date/Time: Tuesday 16/09/2014 | 08:00-10:30

Paper Time: 08:12

Venue: Capital Hall B

First Author: : C.Chan SINGAPORE

Co Author(s): :    M. Ang   A. Saad   D. Gatinel        

Abstract Details

Purpose:

To investigate the efficacy of an artificial intelligence and scoring system, the SCORE Analyzer (TechnoLas Perfect Vision, Munich, Germany), in the detection of subclinical keratoconus in Asian eyes.

Setting:

Singapore National Eye Centre, Singapore, Rothschild Ophthalmology Foundation, Paris, France

Methods:

Cornea topographies performed with the Orbscan IIz analyzer (Bausch + Lomb TechnoLas) were retrospectively collected with masking of patient identifiers and tested with the SCORE analyzer. Eyes tested were classified into two groups. The first group involved clinically and topographically normal eyes with contralateral keratoconus (forme fruste keratoconus (FFKC)). The second or control group involved normal pre-operative topographies of patients who had laser insitu keratomileusis (LASIK) performed at least 4 years prior with no keratectasia. Results were classified as true positive, true negative, false positive and false negative, from which sensitivity and specificity were calculated.

Results:

There were 130 Orbscans of 130 patients. Mean age was 29+/-0.4 years, with 116 Chinese(89%), 9 Malay (7%) and 5 Indian (4%) patients. There were 24 eyes in the FFKC group and 106 eyes in the control group. In the FFKC group, the SCORE was negative in 7 eyes (false negative) and 4 in the control group were scored as positive (false positive). Sensitivity was 70.83% (95% Confidence Interval (CI) 48.91% to 87.33%). Specificity was 96.23% (95% CI 90.61% to 98.94%). Positive Predictive Value was 80.95% (95% CI 58.08% to 94.44%) and Negative Predictive Value 93.58% (95% CI 87.21% to 97.37%).

Conclusions:

This artificial intelligence and scoring system shows good sensitivity and specificity in the detection of subclinical keratoconus in Asian eyes which were undetected by a conventional scanning slit and placido-disc cornea topography system alone.

Financial Interest:

One or more of the authors... receives consulting fees, retainer, or contract payments from a company producing, developing or supplying the product or procedure presented, One or more of the authors... travel has been funded, fully or partially, by a competing company, One or more of the authors... travel has been funded, fully or partially, by a company producing, developing or supplying the product or procedure presented, One or more of the authors... research is funded, fully or partially, by a company producing, developing or supplying the product or procedure presented, One or more of the authors... receives consulting fees, retainer, or contract payments from a competing company