Purpose: To evaluate skin toxicity through daily changes in blood flow to the skin from optical-wavelength video processing.
Methods: A conventional digital camera was used to record daily videos for two weeks of skin damage caused by an incidental sunburn in an investigator self-monitoring feasibility study. The camera and investigator’s leg were positioned reproducibly to maintain imaging conditions. A fingertip pulse oximeter was utilized to provide a ground-truth heart rate. Videos were post-processed to identify cardiac signals using a binned pixel technique. For each frame, the original videos were down-blurred and progressively resampled to obtain a final 1D signal with improved signal-to-noise ratio (SNR). Heart rate was then identified by Fourier transforming this signal. The original 1920x1080 pixels video was downsampled to a coarse grid of 4x4 pixels for increased SNR. Each bin was Fourier transformed, and the frequency component amplitude corresponding to heartrate was presented in a false color image power map. Variations in the amplitudes of these power maps were evaluated.
Results: Subtle oscillations of skin tone during the cardiac cycle were able to be extracted from 10-second videos. The technique evaluated changes in the magnitude observed in binned regions of the power map for both undamaged and damaged skin, theoretically representing differences in blood moving into and out of areas of the skin. Regions of higher magnitude in the power map were correlated with skin damage. The first day, the average measure of amplitude was 37.6% higher in the damaged skin, whereas at the end of the two weeks it was only 8.6% higher.
Conclusion: This feasibility study evaluated this novel signal processing method’s capability to track changes in power map magnitude over time. These power map changes should be correlated with blood flow and may be usable as prognostic indicators for skin toxicity during radiotherapy.