ASOPRSFall Scientific SymposiumSyllabus
196
F3
AutomatedPtosisMeasurements fromFacial Photographs
ZacharyBodnar, JohnHolds. Ophthalmology, St. LouisUniversity, St. Louis,MO, UnitedStates
Introduction:
Ptosis is objectively evaluated bymanual measurements of theMRD1 andMRD2, aswell as visual fields. However,
thesemethods are limited by operator dependence and variability, aswell as patient factors suchmovement and the cognitive ability
to participate in testing.
We have developed software that can analyze photographs of patients face and identify features including the corneal light reflex and
lidmargins. Using this softwarewe can automatically extract theMRD1 and theMRD2 from a single photograph of a patient’s face
(see Figure 1 for sample output).
Methods:
We obtained bothmanual measurements of theMRD1 andMRD2 of both normal and abnormal subjects. Photographs
of the subjects’ faces taken at the time of manual measurementswere analyzed using our software to extract automated
measurements of theMRD1 andMRD2. Patientswith surgically or pathologically altered eyelidmargins, heterotropia of 4 vertical
or 8 horizontal prism diopters in primary gaze, nystagmus or abnormal corneal light reflexeswere excluded.We used Bland-Altman
analysis to evaluate the agreement between the twomeasurement methods.
Results:
Bland-Altman analysis showed good agreement between the twomeasurement methodswithmean error of -0.06mm
(standard deviation 0.32) in automatedMRD1measurements as compared tomanual measurements and amean error of 0.22mm
(standard deviation 0.70) in automatedMRD2measurements as compared tomanual measurements.
Conclusions:
Our algorithm can extract accuratemeasurements of theMRD1 andMRD2 from patient photographs.This could
provide a rapid, highly reproduciblemethod for determining ptosismeasurements that could supplement manual measurements and
provide additional documentation for insurers. It could also be amore accurateway tomeasure ptosis in children or other patients
wheremanual measurements are difficult to obtain.
References:
Burmann,T. G., &Valiatti, F. B. (2008).Medida da distância
reflexomargem pormeio de processamento computadorizado de imagens
em usuários de lentes de contato rígidas, 71(1), 34-37. [Portuguese]
Han, S. J., Guo,Y., Granger-Donetti, B.,Vicci,V. R., &Alvarez,T. L. (2010).
Quantification of heterophoria and phoria adaptation using an automated
objective system compared to clinical methods. Ophthalmic &Physiological
Optics :The Journal of the BritishCollege of OphthalmicOpticians
(Optometrists), 30(1), 95-107. doi:10.1111/j.1475-1313.2009.00681.x
Hasebe, S., Ohtsuki, H.,Tadokoro,Y., Okano,M., & Furuse,T. (1995).
The reliability of a video-enhancedHirschberg test under clinical conditions.
InvestigativeOphthalmology &Visual Science, 36(13), 2678-85. Retrieved from
Miller, J.M.,Mellinger,M., Greivenkemp, J., & Simons, K. (1993).Videographic Hirschbergmeasurement of simulated strabismic
deviations. InvestigativeOphthalmology &Visual Science, 34(11), 3220-9. Retrieved from
/
pubmed/8407230Model, D., & Eizenman,M. (2011).An automatedHirschberg test for infants. IEEETransactions on Bio-Medical
Engineering, 58(1), 103-9. doi:10.1109/TBME.2010.2085000Schaeffel, F. (2002). Kappa andHirschberg RatioMeasuredwith an,
79(5), 329-334.Yamanobe, S.,Taira, S.,Morizono,T.,Yagi,T., & Kamio,T. (1990). Eyemovement analysis system using computerized
image recognition.Archives of Otolaryngology--Head&Neck Surgery, 116(3), 338-41. Retrieved from
/
pubmed/2306353
DetailedProgram
—Friday,October 17, 2014
POSTERS
