Discovering irregular pupil light responses to chromatic stimuli using waveform shapes of pupillograms
© Nakayama et al.; licensee Springer. 2014
Received: 27 December 2013
Accepted: 2 July 2014
Published: 7 October 2014
The waveforms of the pupillary light reflex (PLR) can be analyzed in a diagnostic test that allows for differentiation between disorders affecting photoreceptors and disorders affecting retinal ganglion cells, using various signal processing techniques. This procedure has been used on both healthy subjects and patients with age-related macular degeneration (AMD), as a simple diagnostic procedure is required for diagnosis.
The Fourier descriptor technique is used to extract the features of PLR waveform shapes of pupillograms and their amplitudes. To detect those patients affected by AMD using the extracted features, multidimensional scaling (MDS) and clustering techniques were used to emphasize stimuli and subject differences. The detection performance of AMD using the features and the MDS technique shows only a qualitative tendency, however. To evaluate the detection performance quantitatively, a set of combined features was created to evaluate characteristics of the PLR waveform shapes in detail. Classification performance was compared across three categories (AMD patients, aged, and healthy subjects) using the Random Forest method, and weighted values were optimized using variations of the classification error rates. The results show that the error rates for healthy pupils and AMD-affected pupils were low when the value of the coefficient for a combination of PLR amplitudes and features of waveforms was optimized as 1.5. However, the error rates for patients with age-affected eyes was not low.
A classification procedure for AMD patients has been developed using the features of PLR waveform shapes and their amplitudes. The results show that the error rates for healthy PLRs and AMD PLRs were low when the Random Forest method was used to produce the classification. The classification of pupils of patients with age-affected eyes should be carefully considered in order to produce optimum results.
The pupillary response has long been used for diagnostic procedures  and psycho-physiological studies -. The pupillary response, a reaction to light intensity, is well-known, and the pupillary light response controls the dilation and constriction of the pupil in response to changes in light intensity. The response is usually observed and recorded using pupillograms which consist of pupil diameter and time diagrams .
This response is often called the pupillary light reflex (PLR). It has also been used as an objective measure of retinal and optic nerve functions. In recent years, a growth in interest in the examination of PLR has been observed. It is the result of the discovery of a new type of retinal ganglion cells ,. Since the melanopsin-associated photoreceptive system (ipRGCs, intrinsically photosensitive retinal ganglion cells) in the human retina was discovered , various diagnostic procedures have been proposed and introduced.
In particular, the difference between the response behavior of this photoreceptive system to some types of chromatic stimuli and the behavior of the conventional rod-cone system has been compared  to determine the possibility of detecting the condition of the retina. In addition to receiving rod and cone inputs, the responses based on the newly discovered system have been studied as intrinsically photosensitive ,. Those cells are interchangeably referred to as ipRGC or melanopsin-mediated retinal ganglion cells (mRGC), and according to recent studies, they drive pupillary responses and circadian rhythms -. Also, this phenomenon has been applied to diagnostic procedures for patients with glaucoma  or retinas pigmentosa ,. Feature extraction and analysis of PLRs has made a significant contribution to the development of diagnostic procedures for these patients, as opthalmologically scientific evidence was discovered. Another well-known disease which is observed in aging patients is age-related macular degeneration (AMD) . This disease is related to the condition of the retina, and thus, PLRs are affected by the progress of the disease. Unfortunately, this disease spreads gradually. Therefore, detection and prediction procedures are required. A simple procedure which detects prominent symptoms, such as an easy at home test, is necessary.
In addition to some types of portable pupillometers currently available, a PC web camera and a smart phone with an additional lens can be used to observe pupillography. Some types of clinical consultations can be conducted anywhere using either of these pieces of equipment.
As mentioned above, PLRs can be applied to extract features and to detect irregular responses. This means that a record of temporal pupillary changes as pupillograms is a time series signal, requiring various types of signal processing techniques. This suggests that a simple signal processing technique may be used to extract some symptoms of the disease. For example, the Fourier descriptor technique is often used to indicate the shapes of waveforms -, and the possibility of using this method should be examined.
A procedure for extracting features of PLR waveforms of pupillograms is created, and the features used to compare these waveforms are analyzed to detect irregular responses.
To determine the possibility of detection in diseased eyes, several classifying techniques were applied to features of PLRs, and the performance of the techniques is discussed.
The study of ipRGC mentioned in related research work currently being conducted is concerned mainly with, on the one hand, the study of the ipRGC structure and functions and on the other hand, with the diagnostic use of their specific activity. A brief review of literature regarding this research is presented below.
The ipRGCs are atypical retinal photoreceptors distinct from classical rod and cone photoreceptors . They express the photopigment melanopsin and are intrinsically photosensitive, since they showed sluggish melanopsin-mediated responses. They can also act as conventional RGCs by receiving synaptic rod/cone input via bipolar cells. This integrated information is then transmitted to numerous discrete brain regions involved in both non-image and image-forming vision . The ipRGCs ultimately modulate a multiplicity of behaviors including circadian photoentrainment, PLR, activity masking, sleep/arousal, anxiety, light aversion, and even make a significant contribution to visual functions. Recently, it has been discovered that ipRGCs consist of several subtypes that are morphologically and physiologically distinct, which contribute differentially to the abovementioned non-image and image-forming functions ,. Detailed study of the ipRGC’s different types and their behaviour with varying attributes of light are being still analyzed. In particular, intensive research involving the relationship between melanopsin activity and PLR reflex is now being conducted ,. These studies have direct applications in clinical conditions for the diagnosis of retinal degeneration and sleep disturbances under clinical conditions ,.
The application of ipRGCs to the study of retinal degeneration is focused mainly on the use of pupil responses to chromatic light as a clinical marker, to allow differentiation between disorders affecting rod/cone photoreceptors (the outer retina) and those affecting retinal ganglion cells (the inner retina). As used above, the terms ‘inner retina’ and ‘outer retina’ are the consequence of the anatomical distribution of ipRGC photoreceptors (the inner nuclear layer) and rod/cone photoreceptors (the outer retinal layer) in the retina . Where a disease affects multiple retinal layers, the pupillary light reflex could be a useful tool in determining the contributions of the inner and outer retina to the disease process.
Kankipati et al. focused on the post illumination pupil response (PIPR) analysis of glaucoma patients, and compared them with normal subjects ,. They used a 10-s light stimulus (retinal irradiance, 13 log quanta/ cm2/s; light wavelengths, 470 nm (blue), and 623 nm (red)) and recorded pupillary response for 50 s after light cessation. They found that normal subjects displayed a significant PIPR for blue light (but not for red light), which is consistent with the proposed melanopsin-mediated response. When glaucoma patients are compared to patients using age-matching controls, there was a significant decrease in melanopsin-mediated PIPR. They concluded that PIPR has the potential for use as a clinical tool for evaluating patients with glaucoma. Feigl et al.  also tested PIPR to analyze whether glaucoma alters the function of ipRGC. They use a 10-s light stimuli with 488 nm and 610 nm light wavelengths and retinal irradiance of 14.2 log photons cm2/s. They have found that patients with advanced glaucoma have a dysfunctional ipRGC-mediated PIPR. It has been confirmed that PIPR may be a clinical indicator of progressive changes in glaucoma.
Kardon et al. focused on the percentage of pupil contraction in transient and sustained pupil responses in patients with retinas pigmentosa ,. They used a stimulus paradigm using red and blue light as a continuous Ganzfeld stimulus which produced a 13-s stepwise increase in intensity over a 2 log-unit range (low (1 cd/m2), medium (10 cd/m2), and high (100 cd/m2)). They have found that pupil responses to red and blue light stimuli which are weighed to favor cone or rod input are significantly reduced in patients with retinas pigmentosa. Their preliminary results suggest that pupil response to a low-intensity blue light, to a high-intensity red light, and to a high-intensity blue light may be reasonable markers of rod activity, cone-driven responses, and direct, intrinsic activation of mRGC, respectively.
AMD is another concern regarding retinal disease. This disease causes impairment of both the inner and outer retinal layers, depending on the stage of the disease, and pupil chromatic response measurements may allow the monitoring of the progression of the disease or facilitate the determination of different stages of the disease . In particular, the function of ipRGC concerns the PLR while AMD influences the waveforms of PLR .
Brozou et al. tested outer retinal contributions to pupil responses in patients with AMD . The study showed that AMD significantly affects the pupil’s response to light stimulus (20 msec duration and 24.6 cd/m2 intensity), when compared to normal subjects. Feigl and Zele present a new experimental paradigm for the first time that allows the differentiation of inner and outer retinal contributions to the pupil response in AMD . They used a 11.9-s duration 0.5 Hz sine wave stimulus with 464 and 635 nm light wavelengths and retinal irradiance of 15.1 log photon cm−2s−1. Additionally, they introduced a new metric called ‘a phase amplitude percentage’ (PAP) that reflects inner and outer interactions. PAP is determined from the average long-wavelength and short-wavelength peak-to-through phase amplitudes. PAP approaches zero for retinal irradiances below a certain melanopsin threshold, as PLR is driven by rods and cones. PAP is non-zero for retinal irradiances above a certain melanopsin threshold, as PLR is predominantly driven by cones with ipRGC contributions. They also showed that application of this paradigm to AMD patients can provide information that ipRGCs are altered. This result can be a very important step toward using PLR to determine the inner and outer retinal dysfunction of AMD patients. Future work will be focused on quantifying retinal inputs to the pupil response in order to determine the different stages of AMD or to monitor the disease’s progression.
Our work presents an attempt to use the pupil response to chromatic rectangular light pulses, which are routinely measured in clinical studies, as a simple and fast tool (indicator) for early detection of AMD. AMD (both the dry and wet forms) is the one of the most common irreversible causes of severe loss of vision. It usually affects older adults and results in a loss of vision because of damage to the central part of the retina, known as the macula. The gradual disappearance of the retina pigment epithelium (RPE) in the dry form results in patches of chorioretinal atrophy lacking any visual function. In the wet form, the damage is due to the escape of subretinal fluid/intraretinal fluid, blood, or destruction of photoreceptors and RPE by fibrous or fibrovascular tissue. Since AMD destroys photoreceptors which are initial receptors for PLR, it may reduce the PLR. The type or localization of the damaged photoreceptors may then be indicated using a pupil response to chromatic light. AMD is painless and, consequently, the lack of early warning signals in existing retinal pathology necessitates an indicator which will allow for earlier and easier detection (prediction) of this type of disorder. It is believed that the pupil reaction to chromatic light could be such an indicator.
Six healthy young subjects (20 to 21 years old) and six elderly AMD-affected patients (59 to 86 years old) participated. Each elderly patient had a diseased eye and a normal eye, according to a medical doctor.
Diseased eyes with choroidal neovascularization (CNV) in the macular region are often affected by new blood vessels, which bleed and form dense macular scars . Also, CNV is a major cause of visual loss due to AMD .
Type of disease
This study followed the tenets of the Declaration of Helsinki regarding research involving human subjects; informed consent was obtained from all subjects. The study protocol was approved at the Kitasato University School of Medicine Institutional Ethics Committee. This test does not determine the risk, however.
3.2 Experimental procedure
Pupil responses were measured using a PLR observation procedure and without any procedures to stimulate mydriasis, to determine the level of functionality of the melanopsin-associated photoreceptors ,.
Pupil responses were recorded using Iriscorder Dual equipment (Hamamatsu Photonics, Hamamatsu, Japan) at a sampling rate of 30 Hz. This equipment is designed for observing PLRs with ipRGCs activation in accordance with the measurements taken in previous studies , and consists of a measurement controller and a pair of goggles which have an infrared camera and a LED light source for each eye of the goggles. Therefore, the stimulus light presentation was binocular, and the recordings were also binocular. In the experiment, a long wavelength (635 ± 5 nm) red light and a short wavelength (470 ± 7 nm) blue light were used at two different light intensities (10 and 100 cd/m2). The light stimulation conditions were adjusted in accordance with the method of Kawasaki and Kardon , while the measurement validity was also confirmed ,. PLRs for light stimuli were simply recorded as single trial measure. The light pulses were projected within the housing of a pair of goggles. Subjects were asked to not blink for 20 s while their pupil diameter was recorded. The observation period consisted of a 10-s light pulse which caused a restriction of the pupil size, followed by 10 s without a light pulse during which restoration of the pupil size was allowed to occur. These measurements were taken in a dark room with constant lighting conditions. A dark adaptation period of 5 min was allowed prior to the taking of measurements.
In Figure 1, there is little difference in PLR waveform shapes between the left and right eyes. Pupil sizes are sustained during both blue light pulses, and restoration of pupil sizes can be observed after both red light pulses. These phenomena confirm pupil behavior using blue and red light pulses observed in previous studies. However, these phenomena influenced diseased eyes, as shown in Figure 2.
Mean pupil diameters for the stimuli
During all observations for pupil reactions, repeated measures were not taken and so the test-retest reliability of the paradigm has not been evaluated.
3.3 Fourier descriptors
3.4 Prediction procedure for AMD disease
Processing and prediction can be conducted for a few subjects without repeated measurements. Of course, the validity may not be sufficient, but it can be improved step by step when data from additional subjects is gathered. In this way, all of the techniques employed in this procedure are simple.
4.1 Feature definition
The features of observed PLRs are presented using the procedure specified in the ‘Experimental procedure’ section. The actual calculations were conducted using MATLAB (Mathworks, Inc., Natick, MA, USA). To extract features of pupil constriction in transient and sustained pupil reactions, pupil responses mainly during the first 10 s are analyzed because they are melanopsin-based reactions. Every individual observation of the pupillogram for a stimulus was set to a signal x(n) in Equation 1, then FD vectors were extracted such as in Equation 2.
As both Figures 1 and 2 show, the waveforms of PLRs in response to stimuli are different between subjects. To compare the shapes of waveforms quantitatively, the metrics of similarity and dissimilarity should be defined using waveform feature vectors which are noted above. This is a very popular approach for pattern recognition, such as categorization and discrimination of waveforms . Here, the Euclidean distance (or Minkowski’s power metric) can be defined as the Euclidean norm between two feature vectors. This is the dissimilarity metric. The distances between stimuli conditions mentioned above for a set of PLRs of an eye (left eye) are summarized in the following matrix (Ed_L) as a triangular matrix.
These components indicate dissimilarity between the two waveform shapes when their amplitude factors are excluded.
5Comparing PLRs between eyes
5.1 Features of two eyes
There are many cases where one of the eyes is diseased and the other is not. As Table 1 shows visual acuities of both diseased (D) and normal (N) eyes, where all patients have a level of visual acuity in one eye that is comparable to healthy people and a diseased eye with poor acuity. Regarding the progress of AMD, the disease seldom progresses simultaneously in both eyes. This means that there is a difference in retinal condition between the left and right eyes of most patients with AMD. Since the retinal condition is different, the PLR waveforms are also different, as shown in Figures 1 and 2 while both healthy eyes respond similarly in Figure 1. This phenomenon suggests that quantitative differences in PLRs between two eyes may provide some information regarding symptoms of AMD. Therefore, a procedure of feature extraction from PLRs and the creation of a distance matrix in the above section can be applied to other cases with two eyes.
In this notation, D means diseased eye and N means normal eye for each experimental stimulus condition, such as r10.
Regarding the procedure and feature vectors, a distance matrix across two eyes for one subject can be created as follows:
To maintain the presentation of distance matrix for two eyes, both the left and right eyes (L/R) are allotted instead of diseased and normal eyes (D/N) for healthy subjects since they have two healthy eyes.
The distances between PLR waveforms for subjects with normal eyes are relatively shorter than the distances between diseased eyes and distances between one normal and one diseased eye. A matrix can be created for every subject based on this procedure.
According to the matrix of one patient, the distances representing the dissimilarity between the conditions for diseased eyes are longer than the distances for one normal eye and one diseased eye and the distances for two normal eyes. The distances for subjects with two healthy normal eyes are the shortest. Distance matrices were created for all subjects.
5.2 Configurations using MDS
As a result, MDS produces a low-dimensional projection of the data which can present the paired distances between data points. As the low-dimensional projection provides a way to configure the data, the illustration can be used to visually understand the relationship between the data.
Three-dimensional information of MDS for stimuli
This procedure can illustrate some of the differences between healthy subjects and patient subjects, and the clustering results show the tendency toward deviations. Therefore, the subject categories (healthy or patient) are ambiguous. Regarding the specific purpose of this work, more quantitative classification may be required as a result.
In the next section, a possible procedure for this will be introduced and discussed.
6Classification using a combination of amplitudes and features of waveform shapes of pupillograms
6.1 Feature combinations
In the previous section, the features of waveform shapes were extracted as amplitudes and Fourier descriptors. There were significant differences in the mean amplitudes between the eye conditions when the features of waveform shapes were used to separate the subjects. Though they can indicate some of the conditions affecting the eyes, they are not used collaboratively. Since they are recognized as different scales of features, some consideration is required in order to combine them. Also, a more robust classification technique should be considered.
Here, w is a coefficient used as a weighted value to create a balance between the features of waveform shapes and standardized FD0 (FD0′). To optimize the coefficient w, the performance of this classification is evaluated in response to variations of w in the section which follows.
6.2 Classification procedure
In this paper, the number of subjects is small, and the number of trials for taking measurements is limited because the experimental stimuli influences the response. The Random Forest method , which uses an ensemble learning procedure to analyze the classification of a small sample, is used frequently. Also, the Random Forest method can show contributions of features and can be used to conduct cross validation calculations. Since the structure of the feature data set is not clear, the results of the Random Forest method may provide useful information to improve the detection procedure. The statistical package R and the ‘RandomForest’ package , were used for this analysis.
Here, ∞ means a case using only FD0′.
Classification of the modified set of features into two classes: healthy and normal (HN) and AMD (D), or into three classes: fealthy (H), AMD patient (D), and patient normal age-affected eyes (N) was conducted using the RF technique.
According to the preliminary analysis, the performance was low when the feature set of Equation 1 was used. Next, Euclidean distances between the four conditions were analyzed in the same way as in the previous assessment .
Here, a Euclidean distance matrix is shown as EdHealthy L for the example of the healthy subject’s left eye, and the distance feature is noted as FE. The FE vector consists of distance components without zero distance such as diagonal components in Ed matrix.
6.3 Two-class performance
Contingency table for two classes
AMD (D, n = 6)
Healthy and normal (HN, n = 18)
6.4 Three-class performance
Contingency table for three classes
AMD (D, n = 6)
Normal (N, n = 6)
Healthy (H, n = 12)
6.5 Features contributing to estimations
According to the classification results in the previous section, pupils can be classified accurately using features of PLR waveforms, except for the performance of age-affected normal eyes.
These results coincide with the visual differences in PLRs between healthy and AMD-affected patients.
6.6 Clustered PLRs
As the figure shows, the upper cluster displays a group of healthy subjects except for one patient with age-affected eyes. A sub-cluster consists of both eyes of most subjects. The lower clusters consist of patients with age-affected eyes in the upper part and patients with AMD-affected eyes in the lower part. Some healthy subjects and some patients with AMD-affected eyes have been incorrectly classified into these groups. These occurrences have been explained in the above section where performance was classified and evaluated.
To improve the performance of the test, responses should be ophthalmologically diagnosed. For the classification procedure, many different data mining tools have been developed, so a more effective procedure should be devised. The characteristics of features of pupillary response waveform shapes of pupillograms which result from using the Random Forest method will be a subject of our further study.
This paper proposes a possible procedure for detecting AMD-affected eyes and age-affected eyes using features of PLR waveforms. The Fourier descriptor technique was applied to extract the features of PLR waveform shapes of pupillograms and their amplitudes.
To detect affected patients using the extracted features, MDS and clustering techniques were applied to emphasize certain stimuli and subject differences, since there were restrictions on gathering patient data. The performance is shown as a qualitative tendency.
To quantitatively evaluate detection performance, an appropriate classifier such as the Random Forest method was introduced. For this classification, a combination of features was used in order to obtain detailed features of waveform shapes of pupillograms. This was a balanced combination of the two features, and was controlled using weighted values. As a result of classification analysis using the Random Forest method, the performance of the three categories (patients, aged, and healthy subjects) was compared after the weighted values were optimized using variations of the classification error rate.
The results show that the error rates for healthy pupils and AMD-affected pupils were low when the value of the coefficient for a combination of PLR amplitudes and feature of waveforms was optimized as 1.5. However, the error rates for patients with age-affected eyes was not low.
The pupils of patients with age-affected eyes were influenced by various factors, so it may not be easy to classify healthy and AMD-affected patients in some cases. Additional feature processing, such as processing of the pupil diameter , and having a sufficient number of patients for statistical analysis  may be required.
Also, the evaluation of additional diagnostic procedures will be a subject of our further study in the future.
- Ishikawa S, Ukai K: Pupillary escape and visual fatigue phenomena in optic nerve disease. Adv. Diagn. Vis. Opt 1983, 41: 187-195. 10.1007/978-3-540-38823-4_24View ArticleGoogle Scholar
- Hess EH, Polt JM: Pupil size in relation to mental activity during simple problem solving. Science 1964,143(3611):1190-1192. 10.1126/science.143.3611.1190View ArticleGoogle Scholar
- Beatty J: Task-evoked pupillary response, processing load, and the structure of processing resources. Psychol. Bull 1982,91(2):276-292. 10.1037/0033-2909.91.2.276View ArticleGoogle Scholar
- Lüdtke H, Wilhelm B, Adler M, Schaeffel F, Wilhelm H: Mathematical procedures in data recording and processing of pupillary fatigue waves. Vis. Res 1998, 38: 2889-2896. 10.1016/S0042-6989(98)00081-9View ArticleGoogle Scholar
- H Wilhelm, B Wilhelm, H Lüdtke, in Pupillography: Principles, Methods and Applications, Volume 18 of Clinical Pharmacology, ed by J Kuhlmann, M Böttcher. Pupillography - priciples and applications in basic and clinical research (W. Zuckschwerdt VerlagMunchen Germany, 1999), pp. 1–12.Google Scholar
- Hattar S, Liao HW, Takao M: Melanopsin-containing retinal ganglion cells: architecture, projections, and intrinsic photosensitivity. Science 2002, 295: 1065-1070. 10.1126/science.1069609View ArticleGoogle Scholar
- Dacey DM, Liao HW, Peterson BB, Robinson FR, Smith VC, Pokorny J, Yau KW, Gamlin PD: Melanopsin-expressing ganglion cells in primate retina signal color and irradiance and project to the LGN. Nature 2005, 433: 749-754. 10.1038/nature03387View ArticleGoogle Scholar
- Gamlin PD, McDougal DH, Pokorny J: Human and macaque pupil responses driven by melanopisn-containing retinal ganglion cells. Vis. Res 2007, 47: 946-954. 10.1016/j.visres.2006.12.015View ArticleGoogle Scholar
- Young R, Kimura E: Pupillary correlates of light-evoked melanopsin activity in humans. Vis. Res 2008, 48: 862-871. 10.1016/j.visres.2007.12.016View ArticleGoogle Scholar
- Kawasaki A, Kardon RH: Intrinsically photosensitive retinal ganglion cells. J. Neuroophthalmol 2007, 27: 195-204. 10.1097/WNO.0b013e31814b1df9View ArticleGoogle Scholar
- Feigl B, Mattes D, Thomas R, Zele AJ: Intrinsically photosensitive (Melanopsin) retinal ganglion cell function in glaucoma. Invest. Opthalmol. Vis. Sci 2011,52(7):4362-4367. 10.1167/iovs.10-7069View ArticleGoogle Scholar
- Kardon RH, Anderson SC, Damarjian TG, Grace EM, Stone E, Kawasaki A: Chromatic pupil responses -preferential activation of the melanopsin-mediated versus outer photoreceptor-mediated pupil light reflex. Ophthalmology 2009,116(8):1564-1573. 10.1016/j.ophtha.2009.02.007View ArticleGoogle Scholar
- Kardon RH, Anderson SC, Damarjian TG, Grace EM, Stone E, Kawasaki A: Chromatic pupillometry in patients with retinitis pigmentosa. Ophthalmology 2011,118(2):376-381. 10.1016/j.ophtha.2010.06.033View ArticleGoogle Scholar
- Photodynamic therapy of subfoval choroidal neovascularization in age-related macular degeneration with verteprofin Arch. Ophthalmol 1999, 117: 1329-1346. 10.1001/archopht.117.10.1329Google Scholar
- Zahn CT, Roskies RZ: Fourier descriptors for plane closed curves. IEEE Trans. Comput 1971,C-21(3):269-281. 10.1109/TC.1972.5008949MathSciNetView ArticleGoogle Scholar
- Pinkowski B: Robust Fourier descriptions for characterizing amplitude-modulated waveform shapes. J. Acoust. Soc. Am 1994,95(6):3419-3423. 10.1121/1.409961View ArticleGoogle Scholar
- D Zhang, G Lu, in Proceedings of the 5th Asian Conference on Computer Vision. A comparative study on shape retrieval using Fourier Descriptors with different shape signatures (Springer-VerlagBerlin, Germany, 2002), pp. 646–651.Google Scholar
- Schimidt TM, Dacey D, Lucas R, Hatter S, Matynia A: Melanopsin-positive intrisically photosensitive retinal ganglion cells: from form to function. J. Neurosci 2011,31(45):4-16101.Google Scholar
- Chen SK, Hatter S: Intrinsically photosensitive retinal ganglion cells: many subtypes, diverse functions. Trends Neurosci 2011,34(11):572-580. 10.1016/j.tins.2011.07.001View ArticleGoogle Scholar
- Hatori M, Panda S: The emerging roles of melanopsin in behavioral adaptation to light. Trends Mol. Med 2010,16(10):435-446. 10.1016/j.molmed.2010.07.005View ArticleGoogle Scholar
- Park JC, Moura AL, Rhee DW, Kardon RH, Hood DC: Toward a clinical protocol for assessing rod, cone, and melanopsin contributions to the human pupil response. Invest. Ophthalmol. Vis. Sci 2011,52(9):6624-6635. 10.1167/iovs.11-7586View ArticleGoogle Scholar
- Feigl B, Zele AJ: Melanopsin-expressing intrinsically photosensitive retinal ganglion cells in retinal disease. Ophthalmol. Vis. Sci 2014,91(8):894-903. 10.1097/OPX.0000000000000284Google Scholar
- Morgia CL, Ross-Cisneros FN, Hannibal J, Montagna P, Sadun AA: Melanopsin-expressing retinal ganglion cells: implications for human diseases. Vis. Res 2011, 51: 296-302. 10.1016/j.visres.2010.07.023View ArticleGoogle Scholar
- Markwell EL, Feigl B, Zele AJ: Intrinsically photosensitive melanopsin retinal ganglion cell contributions to the pupillary light reflex and circadian rhythm. Clin. Exp. Optom 2010,93(3):137-149. 10.1111/j.1444-0938.2010.00479.xView ArticleGoogle Scholar
- Kankipati L, Girkin CA, Gamlin PD: Post-illumination pupil response in subjects without ocular disease. Vis. Neurophysiol 2010,51(5):2764-2769.Google Scholar
- Kankipati L, Girkin CA, Gamlin PD: The post-illumination pupil response is reduced in Glaucoma patients. Vis. Neurophysiol 2011,52(5):2287-2292.Google Scholar
- Feigl B: Age-related maculopathy - Linking aetiology and pathophysiological changes to the ischaemia hypothesis. Prog. Retin. Eye. Res 2009, 28: 63-86. 10.1016/j.preteyeres.2008.11.004View ArticleGoogle Scholar
- Brozou CG, Fotiou D, Androudi S, Theodoridou E, Giantselidis C, Alexandridis A, Brazitikos P: Pupillometric characteristics in patients with choroidal neovascularisation due to age-related macular degeneration. Eur. J. Ophthalmol 2009,19(2):254-262.Google Scholar
- Olsen TW, Feng X, Rath PP, Steuer ER: Fluorescein angiographic lesion type frequency in neovascular age-related macular degeneration. Ophthalmology 2004,111(2):250-255. 10.1016/j.ophtha.2003.05.030View ArticleGoogle Scholar
- Ishikawa H, Onodera A, Asakawa K, Nakadomari S, Shimizu K: Effects of selective-wavelength block filters on pupillary light reflex under red and blue light stimuli. Jpn. J. Ophthalmol 2012, 56: 181-186. 10.1007/s10384-011-0116-1View ArticleGoogle Scholar
- Narita A, Shrai K, Kubota N, Tkayama R, Tkahashi Y, Onuki T, Numakura C, Kato M, Hamada Y, Sakai N, Ohno A, M S Matsushita Asami, Hayashi A, Kumada T, Fujii T, Horino A, Inoue T, Kuki I, Asakawa K, Ishikawa H, Ohno K, Nishimura Y, Tamasaki A, Maegaki Y: Abnormal pupillary light reflex with chromatic pupillometry in Gacher disease. Ann. Clin. Transl. Neurol 2014,1(2):135-140. 10.1002/acn3.33View ArticleGoogle Scholar
- I Morishita, H Kobatake, Signal Processing (In Japanese) (The Society of Instrument and Control Engineers, Tokyo, Japan, 1982).Google Scholar
- DGR Stork, O Duda, Hart P E, Pattern Classification, 2nd edn (John Wiley & Sons Inc.,2001). [Japanese translation by M. Onoue, New Technology Communications Co., Ltd., Tokyo, Japan].Google Scholar
- Y Takane, in handbook of statistics 26 – Psychometrics. ed by C Rao, S Sinharay. Applications of Multidimensional Scaling in Psychometrics (North-Holland, Amsterdam, Netherlands, 2007), pp. 359–400.Google Scholar
- M Nakayama, W Nowak, H Ishikawa, K Asakawa, in Proceedings of BIOSIGNALS2010. An assessment procedure involving waveform shapes for pupil light Reflex (INSTICCPortugal, 2010), pp. 322–326.Google Scholar
- Leeuw J, Mair P: Multidimensional scaling using majorization: SMACOF in R. J. Stat. Softw 2009,31(3):1-30.Google Scholar
- Package smacof. . Accessed date (07/03/2011)., [http://cran.r-project.org/web/packages/smacof/smacof.pdf]
- J Mingzhe, M Murakami, in Proceedings of the Institute of Statistical Mathematics, Volume 55. Authorship identification using random forests (ISMTokyo, Japan, 2007), pp. 255–268.Google Scholar
- J Mingzhe, Toukeiteki Tekisuto Kaiseki 15. ESTRELA1 (No. 182), 44–49 (2009).Google Scholar
- Package randomForest. . accessed date (14/03/2012)., [http://www.stat.berkeley.edu/~breiman/RandomForests/]
- M Nakayama, W Nowak, H Ishikawa, K Asakawa, Y Ichibe, in IEICE Technical Report. Waveform Shapes for Pupil Light Responses and Their Analyzing Procedure (IEICETokyo, Japan, 2011), pp. 1–6.Google Scholar
- Ren P, Barreto A, Gao Y, Adjouadi M: Affective assessment by digital processing of the pupil diameter. IEEE Trans. Affect Comput 2013, 4: 2-14. 10.1109/T-AFFC.2012.25View ArticleGoogle Scholar
- Ren P, Barreto A, Huang J, Ortega FR, Adjoudadi M: Off-line and on-line stress detection through processing of the pupil diameter signal. Ann. Biomed. Eng 2014, 42: 162-176. 10.1007/s10439-013-0880-9View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License(http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.