Purchase this article with an account.
G. Sartani, Y. Lang, A. Pollack, R. Siegal, J. R. Ferencz, I. Yeshurun, J. Karp, T. Lifshitz, A. Loewenstein; Computational Detection of Visual Field Changes in Patients With Intermediate AMD. Invest. Ophthalmol. Vis. Sci. 2008;49(13):4219. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Early detection of CNV has the potential of significantly improving the final visual outcome of patients. The Home Macular Perimetry (HMP) device (Notal Vision Inc, Israel) is designed to deal with this problem by monitoring the visual field of intermediate AMD patients. The HMP runes a computerized test that is composed of hyperacuity (Vernier) stimuli. Each test is scored by a machine learning classifier for its indication of the existence of a visual field defect. A patient that receives the device is scored daily or weekly. We developed a statistical analysis that makes use of the sequential testing for the early detection of visual field changes that may be indicative of CNV development.The objective of the study is to design effective statistical analysis to detect visual field changes in patients at risk for developing CNV using sequential Home Macular Perimetry testing.
In the first few weeks, the device learns the patient's scoring statistics. During the learning time the patient is monitored by a Wilcoxon rank test to detect deterioration in the patient's (yet unknown) statistics. Once the statistics is evaluated, we employ the CUSUM change detection algorithm. CUSUM is used in two modes: (i) Conversion mode, where the algorithm detects deterioration of the patient's average score beyond the classifier predetermined threshold. (ii) Progression mode, where the algorithm detects any deterioration in the patient's test statistics.
The CUSUM's parameters are fixed to meet two demands: (i) The yearly false alarm rate is less than 5%. (ii) Each patient working in the conversion mode has a threshold-crossing detection time of less than a month. We use computer simulations over a wide range of possible patients' statistics to show that the algorithm maintains the required yearly false alarm rate without compromising the demand for early detection.
We designed an effective statistical scorer that utilizes a sequential HMP testing for early detection of visual field changes in intermediate AMD patients. Combined with the HMP accurate single test scoring, this algorithm makes may help in the early detection of CNV.
Clinical Trial: :
This PDF is available to Subscribers Only