Development of a time-dependent model of the human pupil with adaptive receptor weighting

点击翻译按钮获取中文摘要

Research projects concerning the human pupil diameter have intensified in recent years, achieving considerable progress, especially in the application of the pupil diameter as a biomarker in interdisciplinary fields of application. For example, it is possible to assess the human cognitive performance or to detect various physiological states such as fatigue, arousal, or deep depression via the temporal variation of the human pupil size. In lighting technology, the pupil diameter is considered a relevant topic, especially in the context of the ophthalmic optical apparatus and application-oriented indoor lighting. The size of the pupil diameter varies the retinal illuminance, which can indirectly affect the humans' brightness perception or visual acuity. The discovery of intrinsically photosensitive retinal ganglion cells (ipRGCs) has led to more intense research in the field of the afferent pupillary control path and the pupillary light response. From the results of neurophysiological fundamental research, it can be deduced that the photopically adapted pupillary light response is controlled by a combination of outer retinal photoreceptors and the ipRGCs. The proportion of different photoreceptor classes in controlling the pupillary light response is not only time-dependent, but also depends on various properties of light and experimental conditions such as presentation time, spectral composition of the light, intensity, and adaptation time. Despite outstanding research successes in the field of neurophysiological research on the pupillary light reflex, there is still a lack of a mathematical model that can predict the temporal variation of the absolute human pupil diameter in response of different light spectra. Existing pupil models mainly apply a V(𝜆)-weighted quantity as an independent parameter and do not integrate the influence of the ipRGCs, which can lead to significant prediction errors of the pupil diameter when using LED spectra. This project is concerned with the development of a new model for predicting light-induced pupil diameter, which through the use of artificial neural networks (Deep Learning) might for the first time be able to predict the temporal and spectral-dependent human pupil diameter. Our preliminary modeling attempts on test data sets have clarified that with the proposed methodology, the temporal pupil diameter can be reconstructed from photometric quantities with a mean absolute error (MAE) of less than 0.1 mm. In this continuation proposal, we plan to develop the deep learning-based approach further and, among other things, open up the possibility to integrate the latest findings on the neurophysiological mechanism of light-induced pupil behavior into a mathematical model.