Weather-informed Light–tissue Model–Based Dose Planning for Indoor Daylight Photodynamic Therapy

Abstract

Daylight activation for photodynamic therapy (PDT) of skin lesions is now widely adopted in many countries as a less painful and equally effective treatment mechanism, as compared to red or blue light activation. However, seasonal daylight availability and transient weather conditions complicate light dose estimations. A method is presented for dose planning without placing a large burden on clinical staff, by limiting spectral measurements to a one-time site assessment, and then using automatically acquired weather reports to track transient conditions. The site assessment tools are used to identify appropriate treatment locations for the annual and daily variations in sunlight exposure for clinical center planning. The spectral information collected from the site assessment can then be integrated with real-time daily electronic weather data. It was shown that a directly measured light exposure has strong correlation (R²: 0.87) with both satellite cloud coverage data and UV index, suggesting that the automated weather indexes can be surrogates for daylight PDT optical dose. These updated inputs can be used in a dose-planning treatment model to estimate photodynamic dose at depth in tissue. A simple standardized method for estimating light dose during daylight-PDT could help improve intersite reproducibility while minimizing treatment times.

INTRODUCTION

Using daylight as an activation mechanism for photodynamic therapy (PDT) of skin has been investigated over the past decade, and is now widely accepted in several countries, as a less painful and equally effective treatment mechanism when compared to conventional red or blue light activation (1-3). However, seasonal daylight availability and transient weather conditions complicate light dose estimations, especially in northern latitudes (4-6). Clinically, appropriate treatment months are identified based on latitude and season, and patients are treated for approximately 2 h where appropriate sunlight is expected (7,8). Yet, the changes in solar irradiance due to the time of day and transient weather conditions confound reproducibility. In the current work, a method involving a one-time site assessment is proposed combined with programmatically acquired weather data to provide real-time estimates of light fluence rates and photodynamic dose at depth in tissue.

Daylight PDT and similar low-fluence rate activation methods have been reported to be less painful than conventional PDT. Additionally, the ability to treat multiple patients simultaneously has economic benefits in certain healthcare systems (9). While lights used in conventional treatments are regulated medical devices that have well-characterized narrowband spectra and fluence rates, daylight is broad spectrum and the fluence rate changes on a continuous basis. Despite these well-known fluctuations, daylight PDT is an approved treatment in many countries without explicit guidance on dealing with daily weather prediction. While reasonable approaches to delivering daylight PDT are always implemented, a more formalized approach to estimating light potential seems warranted.

With the variable nature of sunlight versus cloud, it is natural to question how much light is required to deliver an appropriate photodynamic dose. An international consensus by Wiegell et al. provides guidance on the minimum effective fluence needed to activate PpIX and provides a table of appropriate treatment months based on latitude (10). A model developed by O’Mahoney et al. provides a method for estimating the PpIX-weighted light fluence based on a simple low-cost lux meter measurement (6). While weather conditions have been reported in past studies, the recorded parameters are not standardized. Many clinical teams have patients sit outside for a fixed 2–2.5 h period, but it is still unclear whether clearance could be achieved with less time, or conversely whether more time would be beneficial (11). Additionally, intersite comparisons and reproducibility become an issue when the continuous irradiance rate is not considered. The following sections describe methods to improve dose planning without placing a large burden on clinical staff, by limiting spectral measurements to a one-time site assessment, and then using automatically acquired weather reports to track transient conditions during daylight treatments. These methods can be applied to both indoor and outdoor daylight treatments, where indoor treatments provide the added benefit of reliable climate control and reduce the need to apply sunscreen due to the UV-blocking nature of many windows; however, the spectral changes introduced by this barrier need to be characterized for proper light dose estimation.

MATERIALS AND METHODS

A solar site analysis was performed using a Solar Pathfinder (The SolarPathfinder Company, Linden, TN), which is a basic tool commonly used in the photovoltaic industry to site solar panels for maximal annual exposure. The tool consists of a grid showing solar time (vertical lines) which are intersected by months (horizontal lines) as shown in Fig. 1A. The grid is specific to a range of latitudes and printed as white lines on black paper. The base is oriented such that the paper grid fits in the plastic base at a specific orientation, and the whole system is aligned using a built-in compass. A clear plastic dome with open sides is placed on top of the base (Fig. 1A inset, B). A white wax pencil is used to mark the paper grid to denote the boundaries of objects blocking the reflected sky (Fig. 1B, green outline). The grid is then removed from the base, and the times and months indicated by markings are entered into a spreadsheet, which is visualized in Fig. 1C.

Figure 1.

The SolarPathfinder device showing compass alignment and nearby spectroradiometer (A), where the clear plastic dome (inset) reflects the sky. The outline of obstructions is traced (light green–shaded region, B), which can then be converted to a binary grid of sunny and shaded regions (C). Then, as a comparison, the same procedure was conducted in an open untried area outside of the building (C inset) where morning sunlight was available, but tall trees to the west blocked the afternoon sun.

A field spectroradiometer (SS-110; Apogee Instruments, Logan, UT) was used to collect spectral data at 5 min time intervals over the course of multiple weeks. Calibration of the device was performed by the manufacturer using National Institute of Standards and Technology (NIST) traceable light sources. The device was leveled and placed on a window sill in close proximity to where the site assessment was performed and where patients are to be treated (Fig. 1A, upper left). The total irradiance was calculated by integrating spectral measurements between 350 and 800 nm.

A weighting based on the absorption spectrum of PpIX (12) was used to find the effective irradiance. The PpIX absorption spectra were first normalized and then multiplied by the spectroradiometer measurements. This provides a metric of the PpIX-weighted effective irradiance, which, when combined with treatment time, provides the effective light dose.

Using a Python script to interface with an application programing interface (API), hourly weather data were programmatically accessed from weatherbit.io for an airport approximately 5 km from the treatment site. These data provide formatted key–value pairs of various weather parameters. The main values considered are a model-based solar irradiance estimate assuming clear skies, percentage of cloud coverage based on satellite imagery, and UV index. Temperature and humidity values are also provided; however, they only need to be considered for outdoor treatments.

RESULTS

The solar site analysis was performed in the waiting area of the Dermatology clinic at Dartmouth Hitchcock Medical Center in Lebanon, NH (latitude 43.6°N). This area is on the 2^nd floor and has large southwest-facing windows. The analysis shows there are between 1 and 5 h of direct sunlight each afternoon for this location (Fig. 1C). During the months of mid-April through mid-August, there is a decrease in noontime sunlight due to the building awning. Mid-October through mid-January have less than 2 h of direct sunlight due to the building orientation. A site assessment was also conducted in a nearby picnic area where the sky was unobstructed in the morning, but tall trees just to the west blocked afternoon sunlight (Fig. 1C inset).

As an example of the variability observed in the total irradiance, measurements collected over a 5 day period are provided in Fig. 2, where the blue-shaded regions are in the morning when the building blocks direct sunlight, and the red-shaded regions are the direct afternoon sunlight. The first day shows how cloudy conditions can impact the irradiance. A clear day and mostly clear day are shown on the 174^th and 175^th days of the year, respectively, while day 176 is mostly overcast. Even on the clear day, the total irradiance can fluctuate over a 2 h window and the peak irradiance is only observed over a short period of time. The horizontal lines during the night represent the base-line indoor lighting, which is turned off just after midnight most nights.

Figure 2.

Solar variation from total irradiance measurements is shown, as taken at 5 minute intervals over approximately 5 days, showing high irradiances only in the afternoon hours (pink-shaded area) and at least an order of magnitude less in the mornings (blue-shaded areas).

While Fig. 2 shows the total irradiance at 5 min time resolution, Fig. 3 provides representative examples of the spectral distribution at 14:00 on days in different seasons and with different weather conditions. Fig. 3A provides the spectral characteristics measured indoors on a clear day in mid-June, whereas Fig. 3B is on an overcast day the same week. The overcast day reports just 2% of the total irradiance for the same time two days before. To compare seasons, Fig. 3C provides spectral measurements for a clear day in mid-October, where the overall total irradiance is slightly less than June, but still sufficient for treatment. However, a partly cloudy day in October has approximately 10% as much light. The indoor CFL lights can be observed in Fig. 3B (spectral peaks at approximately 440, 560, and 610 nm) and provide a significant portion of the irradiance due to the overcast conditions at that time.

Figure 3.

Spectroradiometer measurements taken at 14:00 on a clear day in June (A), a cloudy day the same week (B), a clear day in October (C), and a partly cloudy day in October (D). The weather report is shown in the upper left of each graph, and the total irradiance and PpIX-weighted irradiance are given in the upper right.

Using the spectroradiometer measurements with our previously published light–tissue model (13), an estimate of the photodynamic dose at depths can be determined. This was performed for daylight spectra collected both outdoors and indoors. The spectral measurements are shown in Fig. 4A, where for comparison purposes the intensities were uniformly scaled such that the indoor irradiance is 75% of the corresponding total outdoor irradiance, which is similar to reports by others (14). The actual reduction is dependent on the transmission of the glass or acrylic barrier, which is concisely summarized by O’Mahoney et al. for many common materials (15).

Figure 4.

Scaled daylight spectra measurements of outdoor and indoor daylight (A) and their PpIX-weighted counterparts (B). Snapshots of photodynamic dose after 30 minute incubation for various treatment times (outdoor: solid; indoor: dotted) (C) and their corresponding maximal depth of effective photodynamic-dose (PDD) (D).

Most investigations of daylight PDT report light dose as the effective fluence, which is the fluence weighted by the PpIX absorption spectrum (4,6,14,16-18). So, even though the total irradiance of indoor daylight is 75% of its outdoor counterpart, the PpIX-weighted effective irradiance indoors is 60% of the outdoor complement because more UV light is blocked by the window (Fig. 4B). While the PpIX effective irradiance aims to account for the spectral characteristics of broad-spectrum activation, when these irradiance values are used to estimate photodynamic dose (PDD) at depth in tissue after a 30 min incubation period, where a combination of fluence rate, PpIX production rates, photobleaching, and time is considered, the overall depth of activation is largely similar (Fig. 4C-D).

To determine whether specific weather-based metrics could be used as surrogate for spectroradiometer measurements, time-correlated measurements were compared with corresponding weather data. Spectroradiometer measurements were averaged for over a 10 min window around each new hour for times that were previously determined to be in the “Sunny” region of Fig. 1C. These irradiance values were then compared to the product of the cloud percentage and the modeled ideal sunlight (Fig. 5A), or the UV index (Fig. 5B), reported by the weather station at the local airport. When only the modeled sunlight is assumed, the correlation is poor (R²: 0.49, not shown), but both the cloud-corrected sunlight and UV index show a strong correlation (R²: 0.87).

Figure 5.

Comparison of indoor PpIX-weighted spectroradiometer measurements (x-axis) and cloud coverage (A) and UV index (B), where the blue-shaded region provides a 95% confidence interval and the dashed lines provide the 95% prediction limits.

DISCUSSION

While weather data have been incorporated in previous studies of daylight PDT (4,11,14), the reporting mechanism is not standardized. Wiegell et al. asked patients to record weather conditions based on a 1–5 scale and retrieved maximal UV index during the treatment from an external source. This study further found UV index could not be used as a predictor of daylight PDT light dose. However, these results may be limited by the reporting mechanism, whereas our more specific time-correlated method indicates a correlation. A recent review by Philipp-Dormston et al. found there have been no clinical studies showing an impact of average daylight light dose on daylight-PDT efficacy, once a minimal threshold is met (19). While our model indicates there is a link between total light fluence and photodynamic dose, it also indicates treatment could be sufficient in well under 2 h on many sunny days, and even some cloudy days in the summer.

Before implementing a daylight PDT protocol, a site assessment should be performed ideally, to verify potential light exposure. Seasonal variations in the solar path at the treatment location are important to understand and various sites may have better seasonal or daily value. While the current study proposes an indoor daylight PDT protocol so the building orientation is the driving factor, outdoor treatments could have similar obstructions and can be vetted in the same manner. The seasonal solar availability at specific sites can be used for long-term clinical scheduling. For example, in our case the sun was generally available only in the afternoon so patients should only be scheduled during this period; however, an outdoor site near the building was identified that provides daylight from the early morning through mid-afternoon. Just as clinical procedures are scheduled in specific rooms, daylight-PDT locations could be scheduled based on solar availability of the site.

While seasonal solar availability is one important aspect of a site assessment, another is the spectral characteristics of the location. The spectral characteristics of daylight behind a glass window or outside are different, especially in the UV region. Spectroradiometer measurements taken during a site assessment can be used as an input to our model-based dose-planning tool. Using this tool with the input spectra, an estimate of photodynamic dose at depth in tissue can be obtained for different incubation and treatment periods.

Since the primary spectral changes introduced by a clear barrier are a reduction in UV–blue light, the PpIX-weighted effective irradiance will be reduced to a greater degree than the total irradiance changes. However, since these spectral changes are UV-weighted, their impact at depth in tissue will be minimal and changes in the photodynamic dose will primarily be localized to the first few hundred microns of tissue. It is also interesting to note that as the treatment time increases, there is more photobleaching at superficial tissue layers, so the overall photodynamic dose at specific treatment times is reduced near the surface. Even though the indoor effective irradiance is reported to be 60% of the outdoor counterpart, the PDT dose at depth remains equivalent beyond the first 200 μm of tissue. Additionally, within the 15 min of treatment after a 30 min incubation, there is sufficient photodynamic dose to depths of approximately 1 mm for the given irradiance; however, to treat the next mm, the treatment time needs to be lengthened by 10x.

While indoor daylight PDT requires less vigilance in applying sunscreen, careful consideration should be taken when choosing a sunscreen for outdoor treatments (6). The chosen sunscreen should not block wavelengths that are predominantly absorbed by PpIX. If sunscreens are applied, this model would need to be adjusted to account for higher light scattering in the superficial layers of the multilayer skin model. In some cases, a glass gazebo has been used to block UV and reduce the need for applying sunscreen effectively providing indoor daylight (14). With treatment locations that are identified to have full-day sunlight, the spectral composition of UV light will change throughout the day (20), which should be considered when performing the one-time site assessment.

The light dose given during daylight PDT is often reported to be 2–2.5 h in a range of weather conditions, seasons, and latitudes. While others have placed a wrist-based dosimeters on patients (21-23), the ideal dosimeter would be placed on the treatment site, yet current technology and large treatment fields have made this impractical. It may be feasible to place a spectroradiometer near the patient(s) being treated to provide better estimates of light dose, which can then be correlated with clearance; however, this is still slightly cumbersome and impractical.

As a middle-ground solution for improved dose planning, we propose a one-time site assessment. After this site assessment, weather data can be used as a surrogate for continuous spectroradiometer measurements. While seasonal and in some cases weekly or daily variations have been presented previously by others as a way to determine whether sufficient sunlight is available, we have shown transient weather conditions will impact the light availability (Figs. 2, 3). However, by collecting weather data at a higher time resolution that more closely aligns with the treatment period, a more accurate light dose estimate can be obtained. Using either satellite data for cloud coverage or UV index provides a strong correlation with spectroradiometer measurements (Fig. 5). This method can be used for both indoor and outdoor daylight PDT protocols.

Weather data can be acquired programmatically and are generally available from many locations such as airports. This automated process could easily be incorporated into an application on a mobile device, so real-time dose estimates could be obtained with minimal effort. Furthermore, the weather data can be used to estimate the light fluence rate based on the spectral data from the one-time site assessment. This can be used with the lookup tables generated by the dose-planning model, so real-time estimates of photodynamic dose at depth can be monitored. As weather forecasts become more accurate, the same method could transition from dose estimation to a dose-planning mechanism.

CONCLUSIONS

Measuring light dose during daylight PDT is difficult due to a number of reasons, including the length of treatment, broad spectral characteristics, seasonal changes, site latitude, and continuously variable weather conditions. It is unrealistic to expect the clinical team to monitor these factors continuously during treatments, and as a result, many treatment periods are set to 2 h. Yet, for the sake of reproducibility, quantifying the light dose is important to better understand the depth of treatment and potential clearance. We propose methods to improve dose planning without placing a large burden on clinical staff, by limiting spectral measurements to a one-time site assessment. Furthermore, automatically acquired weather data can be used to routinely account for transient conditions during daylight treatments. Using these data in a previously described model, a lookup table can be generated to propose minimal treatment times based on the desired depth of treatment. A simple standardized method for estimating light dose during daylight-PDT could help improve intersite reproducibility while minimizing treatment times and optimizing clearance rate.