Purchase this article with an account.
Marja Pitkanen, Ossi Kaikkonen, Ari O Koskelainen; Retinal temperature determination ex vivo based on ERG photoresponses. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):473.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Heating of the retinal pigment epithelium (RPE) has been considered as a potential treatment for retinal diseases. At present there are no effective methods for measuring the temperature rise of retina and/or RPE. In this study, we calibrated an ex vivo retinal thermometer which is based on two temperature-dependent properties of ERG responses: 1) photoresponse kinetics is accelerated towards higher temperatures 2) warming increases relative photoreceptor sensitivity to long wavelengths (Stiles, W.S., 1948).
ERG responses to dim and strong flashes were recorded from six isolated mouse (C57BL/6J) retinas at random temperatures between 35.0 and 44.0 °C (training dataset, ~16 temperatures per retina). Dim flash stimuli (532 nm and 780 nm) were adjusted to produce b-wave amplitudes of 10 - 15 % of the maximum. Stronger stimulus (~200 R*/rod, 532 nm) was used to induce a response with a clear a-wave. For each temperature, four dim flash responses per wavelength were averaged while strong flash responses were used as such. The temperature-dependence was determined for several features extracted from the responses and the feature values at 37.0 °C were used as references. 3-fold cross validation of the training data was used in order to select the best features and a multivariable linear regression model was created based on them.
The selected features of dim flash responses (b-wave) were time-to-peak and the amplitude ratio of 780 nm and 532 nm responses. The selected features of strong flash responses (a-wave) were time to half of the peak and time to the inflection point of the leading edge. The model was tested with a separate dataset (three retinas, ~13 temperatures per retina) by comparing the temperatures predicted by the model with the actual measured temperatures. The mean absolute prediction error was 0.46 °C and the RMS prediction error (representing the standard deviation of the errors) was 0.61 °C.
Based on the results, the temperature can be determined with accuracy adequate for RPE heating applications. The next phase of this study is to calibrate the model for in vivo mouse ERG and to develop an algorithm for on-line use.
This PDF is available to Subscribers Only