TY - JOUR
T1 - Novel Fluorescein Angiography-Based Computer-Aided Algorithm for Assessment of Retinal Vessel Permeability
AU - Serlin, Yonatan
AU - Tal, Geva
AU - Chassidim, Yoash
AU - Parmet, Yisrael
AU - Tomkins, Oren
AU - Knyazer, Boris
AU - Friedman, Alon
AU - Levy, Jaime
N1 - Funding Information:
The authors have the following interests. This study was partially supported by BG Negev Technologies. There are no patents, products in development or marketed products to declare. This does not alter the authors’ adherence to all the PLOS ONE policies on sharing data and materials, as detailed online in the guide for authors.
PY - 2013/4/23
Y1 - 2013/4/23
N2 - Purpose: To present a novel method for quantitative assessment of retinal vessel permeability using a fluorescein angiography-based computer algorithm. Methods: Twenty-one subjects (13 with diabetic retinopathy, 8 healthy volunteers) underwent fluorescein angiography (FA). Image pre-processing included removal of non-retinal and noisy images and registration to achieve spatial and temporal pixel-based analysis. Permeability was assessed for each pixel by computing intensity kinetics normalized to arterial values. A linear curve was fitted and the slope value was assigned, color-coded and displayed. The initial FA studies and the computed permeability maps were interpreted in a masked and randomized manner by three experienced ophthalmologists for statistical validation of diagnosis accuracy and efficacy. Results: Permeability maps were successfully generated for all subjects. For healthy volunteers permeability values showed a normal distribution with a comparable range between subjects. Based on the mean cumulative histogram for the healthy population a threshold (99.5%) for pathological permeability was determined. Clear differences were found between patients and healthy subjects in the number and spatial distribution of pixels with pathological vascular leakage. The computed maps improved the discrimination between patients and healthy subjects, achieved sensitivity and specificity of 0.974 and 0.833 respectively, and significantly improved the consensus among raters for the localization of pathological regions. Conclusion: The new algorithm allows quantification of retinal vessel permeability and provides objective, more sensitive and accurate evaluation than the present subjective clinical diagnosis. Future studies with a larger patients' cohort and different retinal pathologies are awaited to further validate this new approach and its role in diagnosis and treatment follow-up. Successful evaluation of vasculature permeability may be used for the early diagnosis of brain microvascular pathology and potentially predict associated neurological sequelae. Finally, the algorithm could be implemented for intraoperative evaluation of micovascular integrity in other organs or during animal experiments.
AB - Purpose: To present a novel method for quantitative assessment of retinal vessel permeability using a fluorescein angiography-based computer algorithm. Methods: Twenty-one subjects (13 with diabetic retinopathy, 8 healthy volunteers) underwent fluorescein angiography (FA). Image pre-processing included removal of non-retinal and noisy images and registration to achieve spatial and temporal pixel-based analysis. Permeability was assessed for each pixel by computing intensity kinetics normalized to arterial values. A linear curve was fitted and the slope value was assigned, color-coded and displayed. The initial FA studies and the computed permeability maps were interpreted in a masked and randomized manner by three experienced ophthalmologists for statistical validation of diagnosis accuracy and efficacy. Results: Permeability maps were successfully generated for all subjects. For healthy volunteers permeability values showed a normal distribution with a comparable range between subjects. Based on the mean cumulative histogram for the healthy population a threshold (99.5%) for pathological permeability was determined. Clear differences were found between patients and healthy subjects in the number and spatial distribution of pixels with pathological vascular leakage. The computed maps improved the discrimination between patients and healthy subjects, achieved sensitivity and specificity of 0.974 and 0.833 respectively, and significantly improved the consensus among raters for the localization of pathological regions. Conclusion: The new algorithm allows quantification of retinal vessel permeability and provides objective, more sensitive and accurate evaluation than the present subjective clinical diagnosis. Future studies with a larger patients' cohort and different retinal pathologies are awaited to further validate this new approach and its role in diagnosis and treatment follow-up. Successful evaluation of vasculature permeability may be used for the early diagnosis of brain microvascular pathology and potentially predict associated neurological sequelae. Finally, the algorithm could be implemented for intraoperative evaluation of micovascular integrity in other organs or during animal experiments.
UR - http://www.scopus.com/inward/record.url?scp=84876548638&partnerID=8YFLogxK
U2 - 10.1371/journal.pone.0061599
DO - 10.1371/journal.pone.0061599
M3 - Article
C2 - 23626701
AN - SCOPUS:84876548638
VL - 8
JO - PLoS ONE
JF - PLoS ONE
SN - 1932-6203
IS - 4
M1 - e61599
ER -
