1.

Introduction

Hydrocephalus is characterized by an increased intracranial cerebrospinal fluid volume that is independent of hydrostatic or barometric pressure.¹ This condition is prevalent and costly with some sources estimating that hydrocephalus costs the United States healthcare system nearly two billion dollars annually.^2,3 Hydrocephalus can be a result of several ailments including tumors, degenerative diseases, or trauma and may occur at any age and is also known to arise congenitally in pediatric patients (with one case of hydrocephalus for every 1000 births).^1,4 With a diverse range of associated symptoms, hydrocephalus reduces the quality of life for child-patients during diagnostic evaluation and possibly even after treatment.

Ventriculoperitoneal (VP) shunts are the gold standard in treating pediatric hydrocephalus.⁵ VP shunts, however, elicit challenges for patients including infection, mechanical, and functional issues.⁶ Additionally, nearly half of pediatric hydrocephalus patients reported headaches and generalized pain after VP shunt placements.⁷ Here we use the term shuntodynia to indicate patient-reported pain that is reported clinically in patients with an implanted VP shunt. Specifically, this pain can occur at the shunt valve site or more distally along the shunt distal tubing. Interestingly, shuntodynia has been reported with and without shunt malfunctions.^8,9 Though clinically recognized, the condition has not received significant attention since it is usually seldom life-threatening. The presence of shuntodynia is indicative of patient discomfort, and there is significant physical and psychosocial burden associated with implanted shunt devices.¹⁰

Optical sensing techniques, such as near infrared spectroscopy (NIRS), are well suited to explore questions pertaining to pain since they are noninvasive, painless and can be acquired during noxious stimuli.¹¹ The use of such methods for objective assessment of pain has widely been explored through measurement of hemodynamic responses in various brain regions via functional NIRS.^12–15 Optical spectroscopic methods have also been widely used to monitor and assess pain in exercise science and sports medicine.^16–19 Parallelly, a related diffuse optical technique called diffusion correlation spectroscopy has been used in pediatric subjects with hydrocephalus for rapid and noninvasive assessment of intracranial pressure.^20–24 However, to the best of our knowledge optical sensing has not been applied or explored in the realm of shuntodynia thus far. Additionally, most previous studies that have used optical sensing for pain studies have mostly focused on using these methods for detecting functional changes (such as blood oxygenation, volume, or tissue perfusion) from deep tissue layers. Again, to the best of our knowledge no previous diffuse optical studies have examined optically derived functional changes in skin tissues.

We have previously employed fiber-based diffuse reflectance spectroscopy (DRS) systems operating at short source–detector geometries to quantify functional properties in skin tissues in vivo.^25–27 We have also previously reported preliminary findings from our use of DRS in pediatric patients implanted with VP shunts for hydrocephalus with or without shuntodynia.²⁸ Here we extend our analysis to be statistically robust and examine if patient reported shuntodynia was detectable using optically derived, quantitative hemodynamic signatures obtained in skin-tissues adjacent to VP implants in subjects during a single clinical visit. In our investigation, we hypothesize that shuntodynia-related pain signals originate in or around the scalp-skin sensory system that reside above the skull where the shunt is implanted. Since this region is quite superficial (top 0.3 to 0.5 mm of the skin surface), we use optical probes with shallow-depth sensitivity.

2.

Methods

2.1.

Study Protocol and Clinical Data collection

All procedures for data collection were conducted with approval from the Institutional Review Board at Cincinnati Children’s Hospital Medical Center. Informed consent was obtained from the patient’s parent or legal guardian during routine clinic appointments. Invitation to participate in the study was offered to pediatric hydrocephalus patients that had a VP shunt at the discretion of the attending neurosurgeon. Patients with an inability to stay still for duration of DRS scans (which ranged between 5 and 15 min across subjects) or those that presented hypersensitivity to placement of optical probe were excluded. Of the 35 patients ($N=9$ with shuntodynia) that were recruited in the study, data from four subjects were missing (due to computer glitches), and one subject had VP shunts on both sides and was excluded. Thirty subjects ($N=8$ with shuntodynia) had complete datasets and were used for the analysis reported here. 16 subjects were female and the age range of subjects recruited for these studies ranged between 3 and 20 years. Skin tones ranged from white Caucasian–African–American and Fitzpatrick scores were not collected. However, in data analysis of the reflectance (as described below), melanin was used as an independent absorber by the inverse model to mitigate impact of skin tone on recovered hemoglobin variables. Subjects’ pain thresholds were assessed using the subjective but widely used Faces Pain Scale.²⁹

Optical data collection was performed by having the patients in the seated position. DRS scans were collected at seven different spatial sites for $N=25$ subjects (four sites circling the shunt labeled by O’clock positions, two at the scar sites proximal and distal to shunt, and one from a matched contralateral site on the opposite side to where the shunt was implanted) as a within-subjects control. The placement of the optical probes is indicated schematically by Fig. 1. For five subjects, data from the contralateral sites were not stored correctly and thus not available.

Fig. 1

Schematic depiction of the anatomical placement of the optical probes from a top view of the head. Placement of probes are shown as filled diamonds for the seven sites. Shunt sites are O0-O9 (represent O’clock positions) were approximately distributed around a circle the size of a US quarter (2.5 cm), with center atop the shunt. Sites S1 and S2 represent scar sites (S1 and S2 were located at edge of surgical scars) and were 3 to 5 cm away from closest O’clock sites (O0 and O6, respectively). $C$ is the contralateral matched location on the opposite side of the head relative to the shunt. The diameter of the fiber optical probe used was lesser than 1.5 mm.

3.

Results

3.1.

Optical Data

Figure 2 shows the averaged DRS data collected from all subjects enrolled in the study based on the reported pain status and tissue sites [Figs. 2(a)–2(c) for sites around the shunt, the scar tissue, and a matched contralateral region, respectively]. The mean diffuse reflectance observed at both the shunt and scar sites, the shuntodynia group to have higher reflectance [Figs. 2(a) and 2(b)], which was not seen at the contralateral sites. Additionally, the variance across subjects was higher in the shuntodynia group in comparison to the no-pain group.

Fig. 2

Averaged reflectance data across all patients in the pain (triangles) and no-pain (circles) groups obtained at the three different sites probed (a) DRS data at shunt sites, (b) DRS data at scar sites, and (c) DRS data on the contralateral side. Each figure shows mean DRS computed between patients in each group (symbols) and error bars are standard deviations.

Each measured DRS spectrum was fitted and filtered for goodness of fits as in Eq. (1). Figure 3 shows the calibrated DRS measurements from two representative subjects [Figs. 3(a), 3(c), and 3(e) for a subject with shuntodynia and Figs. 3(b), 3(d), and 3(f) for a subject without], for each site when fitted by the inverse MC model. Symbols indicate measured values while lines indicate MC fits and all data shown here had ${\chi }^2$ values between 0.9 and 0.92. Figures 3(c) and 3(d) show the extracted absorption coefficients for the subject in pain and no-pain group, respectively [Figs. 3(e) and 3(f) show the derived scattering coefficients]. These optical coefficients were obtained from the inverse MC model as a result of fitting the data shown in Figs. 3(a) and 3(b), respectively, as discussed briefly above.^26,30 As seen visually, measured data are well fitted by the inverse MC model and the extracted absorption spectra show well-defined characteristic features between 550 and 600 nm of oxygenated hemoglobin.

Fig. 3

Measured (symbols) and fitted (lines) DRS data from two representative subjects [(a) with shuntodynia and (b) without shuntodynia]. (c), (e) Derived absorption and scattering coefficients for each DRS data fitted in (a). (d), (f) Data for data shown in (b). In each figure, shunt, scar, and contralateral data are shown using triangles, circles, and asterisks, respectively.

3.2.

Hemodynamic Data

The absorption coefficients from inversely fitted DRS spectra were translated into the traditionally used optically derived functional endpoints of total hemoglobin ($[\mathrm{THb}]=[{\mathrm{HbO}}_2]+[\mathrm{dHb}]$) and vascular saturation percent (${\mathrm{SO}}_2=100\times [{\mathrm{HbO}}_2]/\mathrm{THb}]$), whereas the scattering coefficient was averaged across the spectrum to get the average scattering ($<{\mu }_s>$). Thus each DRS spectrum was reduced into three quantitative optical coefficients, which were analyzed statistically.

Figure 4 shows a comparison of the three optically derived functional variables when averaged across each tissue site, for each patient group. The mean oxygen saturation and total hemoglobin values appeared to be similar between the shunt-scar sites for the patients that did not present with shuntodynia but those with shuntodynia had elevated total hemoglobin and oxygen saturation in the scar sites relative to the shunt sites.

Fig. 4

Comparison of mean optically derived hemodynamic and scattering coefficients across the three tissue sites for each patient group (black bars: with shuntodynia and white bars: without shuntodynia). Bars represent mean value while error bars depict standard deviations. (a) Data for the vascular oxygen saturation, (b) data for total hemoglobin concentration, and (c) data for the mean wavelength averaged reduced scattering.

Table 1 presents a summary of the statistical analysis used to model these data obtained and to facilitate site-wise comparisons within each patient group. Table 1 lists modeled means and standard errors in optical variables at each site, for both patient groups. Tukey-adjusted $p$-values are shown for comparisons of each optical variable between any two tissue sites as indicated by Table 1. Comparisons were made within the pain and no-pain groups with the intent of discovering if optically derived variables were different across tissue sites, within individual patients in each group. Statistical analysis was not performed to establish a difference between the pain and no-pain groups at the same site. The principal goal of our study was to identify if optically derived hemodynamic metrics could signal the presence (or absence) of shuntodynia, within individual subjects.

Table 1

Tukey-adjusted, pairwise comparisons of optical variables within patient groups with and without shuntodynia. Means and standard errors for the groups are shown as organized by site and optical variable. p-values are obtained from adjusted Tukey comparisons to estimate if mean differences in relevant variable-pairs across different sites were zero or not. NS indicates p-values above 0.05.

Patient group	Optical variable	Mean ± Std values at			p-values for site–site differences
		Shunt	Scar	Contra–lateral	Shunt versus scar	Shunt versus contra	Scar versus contra
Shuntodynia	${\mathrm{SO}}_2$ (%)	17.9 ± 4.0	24.9 ± 4.3	35.4 ± 5.6	<0.05	<0.01	NS
	[THb] $\mu \mathrm{M}$	8.7±1.6	11.2 ± 1.7	7.0 ± 2.1	0.05	NS	<0.05
	${<\mu }_s>{\mathrm{cm}}^{-1}$	213 ± 16	220 ± 18	143 ± 23	NS	<0.01	.01
No shuntodynia	${\mathrm{SO}}_2$ (%)	33.0 ± 2.4	35.4 ± 2.7	32.3 ± 3.0	NS	NS	NS
	[THb] $\mu \mathrm{M}$	10.5 ± 1.0	11.0 ± 1.0	7.5 ± 1.2	NS	<0.01	<0.01
	$<{\mu }_s>{\mathrm{cm}}^{-1}$	222 ± 10	208 ± 11	205 ± 12	NS	NS	NS

Oxygen saturation proved to be the most useful metric in identifying pain within the group presenting with shuntodynia. For the pain group, oxygen saturation was significantly lower than both the scar and control sites. The extracted total hemoglobin coefficient also exhibited statistical significance within the shuntodynia group where this variable had elevated values at the scar site relative to the shunt sites. Finally, the control contralateral site showed significantly lowered scattering coefficient than all other sites for the patients experiencing pain and may be a statistical anomaly given the low number of contralateral scans available for this patient group.

Interestingly, in patient group without shuntodynia, oxygen saturation was not significantly different across any of the measured tissue sites. Likewise, the scattering coefficient also did not indicate any statistically significant differences between sites. However, total hemoglobin was elevated at both scar and shunt sites, which were located ipsilaterally on the patient’s heads as the implanted shunts. Though statistical comparisons were not performed between patient groups at the same site, Fig. 4 appears to indicate that the two patient groups could show differences at certain sites for some of these optical variables.

4.

Discussion

Fiber-based DRS was used in a preliminary clinical study to objectively measure the hemodynamic status of skin tissue in pediatric patients that had implanted VP shunts as a part of their treatments for hydrocephalus. When recruited subjects self-reported feeling pain around the VP shunt, they were placed in the shuntodynia group while the others were placed into the no-shuntodynia group. The optically acquired measurements were translated into three functional endpoints of vascular oxygen saturation, the total hemoglobin concentration, and the wavelength averaged tissue scattering. These optical endpoints were then analyzed using a two-way ANOVA with a mixed-effects model to statistically model whether the presence of shuntodynia could be detected in hemodynamic signatures of individual subjects by comparing across different tissue sites, per-subject in each group. For each subject, DRS measurements were acquired from seven different tissue sites as described in Fig. 1. To the best of our knowledge, this work marks a first attempt to use quantitative DRS to extract optically derived functional metrics from the skin tissue surrounding implanted VP shunts for the objective detection of shuntodynia.

The acquired DRS spectra were fitted between 450 and 700 nm and our inverse fitting operated with the assumptions that the medium has homogeneous optical properties and that its absorption was a linear combination of four independent absorbers (oxygenated hemoglobin, deoxygenated-hemoglobin, eumelanin, and pheomelanin) whose extinction coefficients were known in the spectral window.³¹ It is important to note that the DRS data as obtained here can be analyzed by generic spectral classifiers (such as PCA or spectral decomposition) as well.^35,36 However, the approach we use here seeks to quantitatively establish hemodynamic parameters using a physics-based light transport model and then explore if these hemodynamic markers could statistically be linked to the presence (or absence) of pain. As mentioned previously, our DRS data were obtained experimentally to be localized to the superficial layers of scalp tissue, rather than deeper tissue, such as skull or the brain using optical fibers separated by $600\mu \mathrm{m}$. Thus our extracted optical coefficients would represent scalp tissue properties more than the skull.

However, our experimental protocol only consisted of obtaining repeat scans at each tissue site for each subject and from these data, we observed that coefficient of variance in the DRS data across the spectrum of interest for repeated scans at a site was clearly lower than the coefficient of variance calculated between sites, which itself was comparable to the variance between subjects. This gives confidence that we had consistent data acquisition in our experimental protocol but does not establish reproducibility. The question of reproducibility would require a redesign of the experimental protocol and will be investigated in the future.

Selection of the random effects model used in the statistical analysis of hemodynamic endpoints incorporated any potential correlations in measurements—both within a site and across different sites in an individual. Therefore, both interpatient and intrapatient variabilities are fully considered in the statistical methods used to model extracted hemodynamic data and thus our analyses accounts for uncertainties and correlations present in the data within and between sites. The analysis shows that only for group of subjects that reported pain the oxygen saturations at the shunt sites was significantly lower relative to the scar or contralateral sites. Patients that did not report pain did not exhibit statistically significant differences in vascular oxygenation (see Table 1). For the total hemoglobin variable, we observed that in subjects that did not report shuntodynia, THb was significantly higher at the shunt and the scar sites than it was for the contralateral side. But for patients reporting shuntodynia, THb was only significantly higher at the scar sites than it was at the contralateral sites—the shunt site had lowered THb for this group, further lending credence to the idea of the vascular supply or function being adversely impacted for this group. In the no-shuntodynia group, there was greater variability across THb estimated at the shunt and scar sites. Overall, there were no discernable differences in the mean scattering for patients with versus without shuntodynia between shunt and scar sites. But we did observe a statistically significant difference for scattering measurements between the contralateral versus scar or shunt sites, in the pain group. Given the low number of scans available for contralateral measurements, this conclusion remains to be validated by future studies.

As described above, we relied on the goodness of fit metric [Eq. (1)] to reject poor-quality fits of measured data. This was done to only consider hemodynamic variables that were extracted from measurements that could be well fitted by inverse MC predictions. We hypothesized that these low-quality fits could be from experimental problems during collection from subjects—particularly with difficulties with keeping the optical probe in firm physical contact with the skin tissue for DRS acquisition. There could be information potentially retrieved from some of these discarded DRS scans here, perhaps using other heuristic approaches to assess hemodynamic variables but we elected to use a more conservative approach given the preliminary nature of the work.

Of the three optical variables derived from measured DRS data, oxygen saturation presented as the most promising metric for identifying shuntodynia. Not only was there a distinction between the pain and no pain groups for oxygen saturation, but pain patients exhibited a lower oxygen saturation around their shunt sites. We focused on using vascular signals detected superficially from scalp and skull tissue using DRS. This was motivated by a hypothesis that the experience of pain felt by patients emerged from the skin-sensory system below which VP shunts were implanted. In shuntodynia patients, the sensation of pain appears localized in the vicinity of the shunt, which we suspect is due to motion of the shunt under the skin, along with local pressure changes experienced by fluid flowing through the shunt. Given the sample size of our study along with the techniques used, we cannot conclude that it was the reduction of vascular oxygenation that causes shuntodynia, though previous reports using optical spectroscopy have also observed that subjects with reduced vascular oxygenation reported pain and have suggested that it could be caused by tissue hypoxia.³⁷ Another study has reported that patients suffering from complex regional pain syndrome type 1 have reduced oxygen vascular saturation in their affected hand when compared to their unaffected contralateral hand.¹⁷ This finding could also be supported by reported assertions of oxygen saturation being a potential biomarker of pain, whether directly related to increased sympathetic activity versus hypovolemia.³⁸

When the results of our study here are taken together with these previously published reports, it appears likely that the shunt placement together with its motion could be causing tissue inflammation and hypoxia in subjects with shuntodynia. Future explorations will consider use of other well-established noncontact optical imaging methods (such as spatial frequency domain methods) to measure vascular functions. Since shuntodynia was self-reported by the patient, our use of DRS here to measure pain was exploratory. However, we have succeeded in our intent, which was to first establish whether DRS was sensitive to presence of self-reported pain in VP shunt patients. In the future, we will consider if such methods could be adapted to explore fundamental pathways from where the painful stimuli emerge.

5.

Conclusion

Here we present preliminary findings for the first time that suggests there were differences in optically estimated hemodynamic signatures in scalp tissues of pediatric hydrocephalus patients that experienced shuntodynia post-VP shunt implantations. Our results showed that oxygen saturation held the greatest statistical value for identifying the presence of pain called shuntodynia. Though the use of DRS was direct and noninvasive, the fiber-based system required contact with skin tissue. As a practical issue in this study, the clinical condition naturally presented impediments to establishing good contact between the probe and skin tissues in subjects and took significant effort and costs in terms of clinical time and patient management for appropriate optical data collection. Thus having fully noncontact optical methods—such as spatial frequency domain imaging or hyperspectral imaging—for sensing tissue hemodynamics could vastly improve data collection ability, especially when working with patients experiencing pain. Additionally, it will also be valuable to integrate optical tissue perfusion measures, such as diffusion correlation spectroscopy, to build a more complete picture of subcutaneous hemodynamics in shuntodynia.