ABSTRACT
BACKGROUND: The purpose of the study was to investigate the computer aided screening method using Purkinje image I and IV reflection patterns for the detection of inapparent eye misalignment and to compare this to an orthoptic examination. MATERIALS AND METHODS: 590 subjects up to 72 months of age with inapparent eye misalignment were recruited from the orthoptic outpatient department and externally. The computer aided screening consisted of taking a series of still video pictures with Purkinje reflection patterns. These were evaluated in an examiner independent way to reach a recommendation whether the child needed an ophthalmological referral or not. As gold standard, an orthoptic examination was performed. For analysis, the data were split by age groups. The orthoptic results were tested for certainty and repeatability. RESULTS: The computer aided examination had the highest sensitivity of 0.82 in the age group up to 2.5 years of age, and a specificity of 0.90. With an estimated prevalence for microtropia of 0.01, the extrapolated positive predictive value was 0.08, and the negative predictive value was 0.998. In the age group up to 2.5 years of age, the percentage of orthoptic examinations without clear result (neither non-referral, nor strabismic) was 22.4%, and 6.1% in the screening examination. Among the cases which were examined repeatedly, some were classified as "strabismic" in the beginning, and as "non-referral" in the end in the orthoptic examinations. CONCLUSIONS: The examiner independent, computer aided screening method is a cost effective option for the screening for inapparent eye misalignment, especially in the age group up to 2.5 years of age. If an orthoptic examination was carried out for screening, one should expect a higher rate of false positives, which entails more costly ophthalmological checks. Future studies should assess the validity of the single ortoptic examination as the gold standard in this age group.
Subject(s)
Orthoptics/methods , Strabismus/diagnosis , Amblyopia/diagnosis , Child , Child, Preschool , Diagnosis, Computer-Assisted , Humans , Infant , Mass Screening/methods , Orthoptics/standards , Pilot ProjectsABSTRACT
BACKGROUND AND PURPOSE: In Germany, 750,000 children are born per year who should be screened for developmental visual defects in the age range 24-48 months. However, the established pediatric screening program is not sufficient to prevent amblyopia. The purpose of this study was to examine the cost-effectiveness of alternatives for amblyopia and microtropia screening. METHODS: Three options were compared: (1) an orthoptic screening carried out in the field, for instance in kindergartens, (2) an examiner-independent objective apparatus-based screening, and (3) a complete ophthalmological and strabismological examination carried out in a practice. The costs of screening, follow-up examinations and of the treatment were modelled for prevalences of 1% (microtropia) and 5% (amblyopia). The benefit due to treatment was calculated as the result of an avoided whole-person impairment of 3% and 1%. The income related, increased tax and health care payments were used to cover the costs. RESULTS AND CONCLUSIONS: In options (1) and (2) there were favorable cost-effective ratios. The practice-based option 3 was economically less promising. The higher the prevalence was, the higher the resulting cost-effectiveness.
Subject(s)
Amblyopia/prevention & control , Vision Screening/economics , Amblyopia/economics , Child , Child, Preschool , Cost-Benefit Analysis , Female , Follow-Up Studies , Germany , Humans , Infant , Male , Patient Care Team/economics , Sensitivity and Specificity , Strabismus/economics , Strabismus/prevention & controlABSTRACT
This study introduces an improved computing scheme for determining eye rotation from Purkinje images I and IV. The original computing scheme systematically underestimated eye rotation. Paraxial raytracing calculations revealed that this error resulted from failure to account for the fact that Purkinje images I and IV fall at different distances behind the cornea. The error could be overcome with a correction factor derived from paraxial raytracing calculations. A series of experiments were carried out to test the validity of this correlation factor, involving exact raytracing calculations as well as measurements on physical model eyes and human eyes. The influence on the correction factor of ocular surface asphericity, accommodation, age and ocular component variations were examined. The new method was also compared to Hirschberg's technique, which makes use of Purkinje image I alone, as a means of screening for strabismus.
Subject(s)
Eye Movements , Optics and Photonics , Rotation , Strabismus/diagnosis , Child , Cornea/anatomy & histology , Humans , Infant , Mathematics , Models, Anatomic , Ophthalmology/methodsABSTRACT
PURPOSE: To develop a fast and reliable image processing algorithm for the extraction of Purkinje reflection data for measurement of ocular alignment with standard PC hardware. METHODS: Purkinje images I and IV are specular reflections of light sources, the positions of which are proportional to eye rotation. Fast histogram-based thresholding and gradient techniques were used to localize Purkinje images I and IV, the pupil centre and the limbus centre in images obtained with near infrared light. RESULTS AND CONCLUSIONS: Execution time was 500 ms per eye. A set of 540 images was evaluated of which 94% were automatically analysed.
