2.1.

Passive Esophageal Dosimetry System

A second design for an esophageal dosimetry system was constructed, to “passively” measure the porphyrin fluorescence (i.e., proportional to active concentration) during PDT light treatment. The dosimeter has capability for six input fibers from six different sites of the esophagus. In this design, the fluorescent signals excited by the treatment laser at a 635-nm wavelength and would be acquired by the fiber tips which are attached to the treatment balloon. The optical signals that contain porphyrin fluorescence signal (fluorescence channel measurement) and the treatment laser signal (reference channel measurement) are separated by a splitter and then measured by two different photomultiplier tubes (PMTs) (Fig. 2 ). These signals from different sites are recorded sequentially in a rapid cycling manner. Relative to the PpIX fluorescence spectrum excited by a 635-nm laser, two filters (620- to $650\text{-}\mathrm{nm}$ bandpass filter for the reference channel and a 690-nm high-pass filter for the fluorescence channel) are placed before the PMTs to reduce crosstalk between the two optical channel signals and eliminate the noise.

Fig. 2

Dosimetry system structure is outlined in (a), including a brief illustration of the fiber excitation at $635\mathrm{nm}$ and the way to split the signal of the detection fibers into 690 long-pass and 620 to 650 bandpass, enabling simultaneous detection and rationing. (b) and (c) The key hardware components of the dosimetry system, including the 6-to-1 optical switch (b), PMT detector [lower-left in (b)], electronic control board [lower-middle in (b)], and the DAQ (data acquisition) card (lower-right). PpIX has a fluorescence peak around $700\mathrm{nm}$ , and the peaks at 660 and 680 are thought to be photoproducts whose signal, and so emission, is collected through a 690-nm long-pass filter. The detection part includes two PMTs for the optical signals of excitation light and PpIX fluorescence.

Both the PpIX fluorescence signal and the PDT treatment laser signal were acquired by diffused end tip fibers. The diffused end tip is $1\mathrm{cm}$ long with a diameter of $200\mathrm{\mu }\mathrm{m}$ . Two different kinds of probes were used for signal acquisition. For large animal studies (not reported here), a balloon probe was used [Fig. 2a] with detection fibers attached to the balloon membrane at up to six different positions. After the probe is inserted into the esophagus and reaches the treatment area, the balloon probe is inflated and the detection fibers are closely attached to the esophagus; thus, those fibers would acquire both signal intensity and signal spatial information. For small-animal studies, a single detection fiber was attached right beside the PDT irradiation fiber.

The detection fibers were lined to a 6-to-1 fiber coupler. Right before the coupler, six shutter sets were introduced. Each shutter consisted of a mechanical block, two collimators, and an aluminum holder. The shutter was controlled by LabVIEW to open/block the optical path between two collimators. Each of the six input fibers of the coupler were $200\mathrm{\mu }\mathrm{m}$ and the output fiber was $800\mathrm{\mu }\mathrm{m}$ , which is lined to the optical collimator [Fig. 2c]. By sequentially opening and closing each shutter, the signal from six different detection fibers are sent to the PMT via this single optical pathway.

The optical signal from the detection fiber contained both PDT treatment laser information (around $635\mathrm{nm}$ ) and PpIX fluorescence information. The photoproduct of PpIX (most likely the result of PpIX photobleaching) yields fluorescence with a peak around $670\mathrm{nm}$ of 10-nm bandwidth, and thus it can be challenging to separate this from the 700-nm peak of PpIX. As illustrated in Fig. 2b, the photons from the detection fiber hit a dichroic mirror and are divided into two difference channels. The reference channel had a 620- to 650-nm bandpass filter right before the PMT. This reference channel was required to measure the PDT treatment laser (635-nm) intensity. The fluorescence channel used a 690-nm long-pass filter right before the PMT, to block both the excitation laser and the photoproduct fluorescence.

Two separate photomultipliers (Model HC120, Hamamatsu Inc.) were used to measure the optical signals. The PMT gains were fixed at 0.6 and $0.8\mathrm{V}$ for reference and fluorescence channels, respectively, as these provided a measurable range from 0 to $1.0\mathrm{V}$ . The current signals coming from PMTs where amplified and acquired as a voltage using a standard 12-bit data acquisition card (National Instruments DAQ 2064E) which was addressed through LabVIEW.

2.2.

Passive Dosimetry System Calibration in Liquid Phantoms

As the dosimeter system was constructed, a series of calibration studies in liquid phantoms were designed to verify the consistency of the dosimeter, and to investigate the correlation between the PDT treatment laser output power and reference channel signal, and also to investigate the correlation between PpIX concentration and fluorescence channel signals.

Liquid phantoms were made up of different concentrations of Intralipid, ink, and PpIX. A mixture of 1% Intralipid approximately equals the reduced scattering coefficient of ${\mu }_s^{\prime }=1.1{\mathrm{mm}}^{-1}$ , with 1% of the 100:1 diluted ink solution is approximately equal to an absorption coefficient of ${\mu }_a=0.025{\mathrm{mm}}^{-1}$ at $635\mathrm{nm}$ . To simulate the measurement in esophageal tissue, phantoms were made to approximately the same optical properties expected in tissue, with values of ${\mu }_s^{\prime }=1.1{\mathrm{mm}}^{-1}$ and ${\mu }_a=0.04{\mathrm{mm}}^{-1}$ . Liquid phantoms were held in 15-ml plastic tubes with a total volume of $10\mathrm{ml}$ .

2.3.

Animal Experiments

The Dartmouth College Animal Care and Use Committee (IACUC) approved all animal work in this study. Sprague-Dawley rats were purchased from Charles River Laboratories (Wilmington, Massachusetts), at 6 weeks old and housed in cages under standard laboratory conditions (room temperature, $22\pm 2°\mathrm{C}$ ; relative humidity, $55\%\pm 5\%$ ; and 12-h light/dark cycle).

The animals were a subset of a larger study with normal Sprague-Dawley rats, with a total of 28 used here. Animals were randomized into four groups and ALA was administrated at different doses [intraperitoneal (ip) at $50\mathrm{mg}∕\mathrm{kg}$ ]. After a 2-h incubation time, the rat esophagus was irradiated with a 633-nm laser. The irradiation was delivered into the lower esophagus through a $1\mathrm{cm}$ long diffusing tip fiber. A total of $20\mathrm{J}∕\mathrm{cm}$ linear irradiance was delivered to each rat with different fluence rates (25 or $50\mathrm{mW}∕\mathrm{cm}$ ) and different laser irradiation methods (continuous irradiation or fractionated irradiation with a 1-min laser on/off interval). During the treatment, rats were anaesthetized with ketamine $(90\mathrm{mg}∕\mathrm{kg})$ and xylazine $(10\mathrm{mg}∕\mathrm{kg})$ . The passive fluorescence dosimeter was used during the PDT treatment; both local laser fluence rate and PpIX fluorescence were monitored. The initial values and final values are used in this report.

Four PDT treatment groups were studied:

The passive esophagus dosimeter was used in the esophagus beside the treatment fiber during the PDT light delivery. This dosimeter could monitor both irradiation light fluence rate and PpIX fluorescence. The system is illustrated in Fig. 2, however, only one detection sensor was used in the rat esophagus study. After rats were anaesthetized, both the PDT irradiation fiber and dosimeter detection fiber were inserted down to rat lower esophagus together. During PDT treatment, measurements were taken every $10\mathrm{s}$ . The treatment laser fluence rate was measured through a 620- to 650-nm bandpass filter, and the PpIX fluorescence was measured through a 690-nm longer pass filter. The dual PMT detection of this dosimeter was used to provide simultaneous acquisition of the signals, and correction for laser intensity fluctuations.

2.4.

Clinical Active Photosensitizer Dosimetry

While the passive dosimeter has been successfully used in animals, it is not yet approved for human use, and therefore it is not possible to obtain continuous measurement of the fluorescence during treatment. The clinical dosimetry system in vivo was developed and demonstrated in previous studies. ^{11, 15, 17, 18, 19} Briefly, the system uses blue light ( $405\mathrm{nm}$ wavelength) excitation and very small optical fibers ( $100\mathrm{\mu }\mathrm{m}$ core diameter) to limit the penetration and distance traveled by light that is detected. The light that is captured from the tissue must have originated within a few scattering distances (typically 100 to $300\mathrm{\mu }\mathrm{m}$ ) of the fibers surface for significant probability of detection. This enables the signal to be linearly proportional to the fluorophore concentration and not significantly affected by variations in tissue optical properties.¹¹ The original design was proposed for surface measurements and the working end, which took measurements from the tissue, was designed to be $6\mathrm{mm}$ across. This end consisted of a set of seven fiber bundles, each having seven individual fibers, where the middle fiber delivered the excitation light and the surrounding ring of six fibers collected the fluorescence. Each set of fibers was separated from each other one by more than a millimeter to enable minimal crosstalk between them. This dosimeter system has been used here for esophageal measurements using a smaller single fiber, consisting of only a single set of seven fibers, with the inner fiber carrying the excitation and the outer six fibers picking up the emission.

Sampling of the fluorescence uses approximately $1\mathrm{mW}$ of total power, and lock-in detection with the light cycled on and off at $200\mathrm{Hz}$ . The signal is calibrated to be linearly proportional to typical in vivo concentrations of PpIX, and normal acquisition requires about $0.5\mathrm{s}$ . Due to the light sensitivity of this instrument, it cannot be used during laser therapy, without additional filtering optics.

This dosimetry system was produced through NCI sponsorship and three such dosimeters are in use at the biomedical laboratories of Massachusetts General Hospital (clinical and preclinical use), Cleveland Clinic (clinical and preclinical use), and Dartmouth Hitchcock Medical Center (preclinical use only). Due to the fact that the treatment beam must be turned off during acquisition, there are fluorescence data only for before and after therapy for clinical studies.

2.5.

Clinical PDT Treatment and Fluorescence Dosimetry

Human studies were carried out at the Massachusetts General Hospital under an Institutional Review Board (IRB)-approved protocol for Barrett’s esophagus PDT. Six subjects were recruited into this arm of the study and baseline fluorescence measurements of the buccal mucosa were taken using the active fluorescence dosimeter. Patients were then given a solution of ALA in orange juice to take orally ( $30\mathrm{mg}∕\mathrm{kg}$ ALA by body weight). Following a 4-h incubation period, patients received topical anesthesia to numb the pharynx as well as intravenous sedation. Pre-PDT irradiation measurements of PpIX levels were taken of the Barrett’s tissue and surrounding normal sites using the fluorescence dosimeter. Postirradiation measurements were taken of the treated sites with the Aurora to measure PpIX levels. Biopsy samples of the diseased and normal tissue were also obtained at the pre- and posttreatment times.

A 5-mm diffusing optical fiber nested in a cylindrical balloon dilator (Fig. 1) was then used to deliver 635-nm light from a KTP/YAG XP 800 dye laser (Laserscope, San Jose, California) to the esophageal lumen. The 5-mm radially diffusing optical fiber was placed within an 18-mm-diam clear cylindrical balloon dilator, so that the fiber was at the center of the esophageal lumen, providing even distribution of the light to the esophageal lumen. A total fluence of $150\mathrm{J}∕{\mathrm{cm}}^2$ , at a tissue surface irradiance of $150\mathrm{mW}∕{\mathrm{cm}}^2$ , was delivered either continuously or as a fractionated dose (60-s light/dark intervals).

The fluorescence dosimeter was used to measure PpIX fluorescence levels in the patient tissue. A 405-nm diode laser provides the excitation light to a fiber optic bundle composed of one excitation fiber surrounded by six fluorescence emission collection fibers with a total diameter of $0.1\mathrm{in.}$ at the distal tip.

2.6.

Statistical Analysis

In all analyses of fluorescence changes examined here, a one-tailed $t$ test was used, with two-sample unequal variance. The differences in fluorescence pre- and posttreatment were mainly examined for differences, to determine if photobleaching was significant.

3.1.

Passive Dosimeter Calibration and Correction for Crosstalk

A total of three sets of liquid phantoms were made with ${\mu }_s^{\prime }=1.1{\mathrm{mm}}^{-1}$ and ${\mu }_a=0.04{\mathrm{mm}}^{-1}$ , but no PpIX was added in phantoms. The detection fiber was inserted into the center of the liquid. Laser outputs were set at 0, 14, 21, 32, 67, 128, and $165\mathrm{mW}∕\mathrm{cm}$ , and measurements were taken at each of these levels.

As illustrated in Fig. 3a , the reference channel PMT output signals (voltage) had linear correlation with laser outputs. Since 25, 50, and $100\mathrm{mW}∕\mathrm{cm}$ are the most used laser powers for ALA-PpIX PDT treatment in the esophagus, our phantom study demonstrated a reliable laser dosimeter in our system. The reading from the reference channel PMT could be used to calculate the PDT light dose.

Fig. 3

(a) Correlation of the reference channel signal to the laser output, as measured in a liquid phantom with linear irradiances of 0 to $165\mathrm{mW}∕\mathrm{cm}$ ; (b) crosstalk from the reference channel to the fluorescence channel shown from a set of phantoms (fixed transport scattering ${\mu }_s^{\prime }=1.1{\mathrm{mm}}^{-1}$ ) with different absorption coefficient values, ${\mu }_a=0.0125$ , 0.025, 0.04, and $0.05{\mathrm{mm}}^{-1}$ ; (c) fluorescence and reference channel PMT readings (red crosses=reference, blue dots=fluorescence) for increasing concentrations of PpIX in a phantom; and (d) correlation of fluorescence channel signals to PpIX concentrations, after crosstalk adjustment. The detection precision was estimated to be $0.1\mathrm{\mu }\mathrm{g}∕\mathrm{ml}$ , and the linear range was up to $4\mathrm{\mu }\mathrm{g}∕\mathrm{ml}$ .(Color online only.)

From the light dosimeter calibration phantom study, it was also noticed that even without any PpIX inside the phantoms, there still were some readings from the fluorescence channel PMT. This was concluded to be crosstalk from the irradiation laser, assuming the 690 longer pass filter could not block all photons in the range of 620 to $690\mathrm{nm}$ . To investigate the crosstalk between reference channel and fluorescence channel, a new set of phantom studies was designed.

A total of four sets of phantoms were made. Considering that there could be some optical variations in individual targets, absorption coefficient ${\mu }_a$ was different in each phantom ( ${\mu }_a=0.0125$ , 0.025, 0.04, and $0.05{\mathrm{mm}}^{-1}$ ). Reduced scattering coefficient ${\mu }_s^{\prime }$ was still set at $1.1{\mathrm{mm}}^{-1}$ for all phantoms and no PpIX was added into the phantoms. Laser outputs were set at 32, 67, and $128\mathrm{mW}∕\mathrm{cm}$ , and measurements were taken at each of these settings.

As seen in Fig. 3b, the results showed that the crosstalk from the reference channel was linearly increasing with the laser output, yet the crosstalk was also very consistent as the optical properties varied within the pertinent range of possible values. These data were fitted into a linear function $y=kx+b$ , where the reference channel $\left(x\right)$ to fluorescence channel $\left(y\right)$ crosstalk functions were $y=0.108x+0.068$ $({\mu }_a=0.0125{\mathrm{mm}}^{-1})$ , $y=0.110x+0.062$ $({\mu }_a=0.025{\mathrm{mm}}^{-1})$ , $y=0.112x+0.060$ $({\mu }_a=0.04{\mathrm{mm}}^{-1})$ , and $y=0.118x+0.059$ $({\mu }_a=0.05{\mathrm{mm}}^{-1})$ , respectively. The slope rate $k$ had 3.8% variation, while $b$ has 6% variation, indicating that these calibration lines could work for a broad range of tissue optical properties. With this consistent linear crosstalk function, readings from the fluorescence channel could be adjusted according to reference channel readings and the crosstalk could be effectively corrected for.

To investigate the correlation between detected fluorescence signal and the actual PpIX concentration, eight sets of liquid phantoms were made with same optical properties ( ${\mu }_s^{\prime }=1.1{\mathrm{mm}}^{-1}$ and ${\mu }_a=0.04{\mathrm{mm}}^{-1}$ ). A total of eight different concentrations of PpIX were added into phantoms (0, 0.1, 0.2, 0.3, 0.5, 1.0, 2.0, and $3.0\mathrm{\mu }\mathrm{g}∕\mathrm{ml}$ , respectively). Irradiation laser output was set at $50\mathrm{mW}∕\mathrm{cm}$ and six measurements were taken for each phantom.

The readings are illustrated in Fig. 3c, with the fluorescence showing positive correlation with PpIX concentration, while the reference channel readings did not. Reference channel readings varied $\pm 4\%$ and decreased a little while the PpIX concentration (fluorescence intensity) increasing significantly. This minor decrease of reference channel signal was reasonable with the assumption that the PpIX would affect the phantom absorption coefficient somewhat. From the consistency of the reference channel readings, it was assumed that the crosstalk from the fluorescence channel to the reference channel was a minor factor. Only crosstalk from reference channel to fluorescence channel must be addressed.

Using the measured crosstalk function, $y=0.112x+0.060$ , to adjust the fluorescence channel readings for this minor correlation, the corrected results are shown in Fig. 3d. A linear relationship between the fluorescence channel signal (after crosstalk adjustment) and the PpIX concentrations was observed. It was encouraging that this passive dosimetry system could detect down to a $0.1\text{-}\mathrm{\mu }\mathrm{g}∕\mathrm{ml}$ difference in the PpIX concentrations, while the linear range included from 0 to $4.0\mathrm{\mu }\mathrm{g}∕\mathrm{ml}$ , which was also the in vivo PpIX concentration range expected in tissue.

3.2.

Animal Data

The full fluorescence kinetic change was examined in detail in a previous study, but here the focus is on the initial versus final fluorescence intensity values. There was a change in signal level from all animals studied in this work, and the average values for each treatment group are shown in Fig. 4a . This illustrates the fluorescence signal in esophagus, for a fixed ALA ip administration dose of $50\mathrm{mg}∕\mathrm{kg}$ , with a 2-h incubation time. The fluorescence intensity changes were $15\pm 2$ , $10\pm 3$ , $25\pm 3$ , and $21\pm 4$ , respectively, for the 50-mW/cm continuous group, the 25-mW continuous group, the 50-mW/cm fractionated group, and the 25-mW/cm fractionated group. Interestingly, the lower fluence rates did not cause less photobleaching than the higher ones, based on a one-tailed $t$ test analysis of the average data. However, the light fractionation groups both had significant increases in the photobleaching, with $p$ values of 0.02 and 0.016 for the 50- and the 25-mW/cm groups, respectively.

Fig. 4

(a) PpIX fluorescence as measured from rat esophagus under varying treatment conditions before and after PDT, with error bars showing the standard error in the groups, and (b) the calculated percentage change in the esophagus fluorescence.

3.3.

Clinical Data

Relative to post-ALA, pre-PDT fluorescence measurements, a significant decrease in PpIX fluorescence levels was observed in Barrett’s tissue that received either continuous illumination $(49\%\pm 5\%)$ versus the fractionated illumination $(74\%\pm 12\%)$ (Fig. 5 ). The fractionated illumination caused significantly more photobleaching than continuous illumination ( $p$ value $<0.001$ ). PpIX fluorescence levels did not change in the oral cavity and normal squamous control sites, as might be expected, since these areas were not irradiated with light. However, measurement of these tissues provides a simple internal control measurement for each subject. Biopsies of the Barrett’s tissue taken $4\mathrm{h}$ after administration of ALA, just before treatment with light, showed a heterogeneous distribution of PpIX fluorescence in the metaplastic tissue as confirmed by histology in a parallel tissue section (Fig. 6 ).

Fig. 5

PpIX fluorescence as measured from Barrett’s tissue and two nonilluminated control sites of oral cavity and normal squamous tissue, with standard error values in each group shown as error bars. (a) Data from continuous irradiation (b) and fractionated irradiation with 1-min on/off fractions. (c) Significant photobleaching was observed in the Barrett’s tissue for both continuous illumination $(P<0.001)$ and fractionated illumination $(P<0.001)$ with the 635-nm light ( $150\mathrm{J}∕{\mathrm{cm}}^2$ at $150\mathrm{mW}∕{\mathrm{cm}}^2$ ). The Barrett’s site treated with fractionated illumination photobleached significantly more than the site treated with continuous illumination following PDT $(P<0.001)$ .

Fig. 6

PpIX fluorescence levels shown as measured from Barrett’s tissue and two nonilluminated control sites of oral cavity and normal squamous tissue. Significant photobleaching was observed in the Barrett’s tissue following either (a) continuous illumination ( $P$ value $<0.001$ , $n=4$ ) or (b) fractionated illumination ( $P<0.001$ , $n=3$ ) with 635-nm light ( $150\mathrm{J}∕{\mathrm{cm}}^2$ at $150\mathrm{mW}∕{\mathrm{cm}}^2$ ). The Barrett’s site treated with fractionated illumination photobleached significantly more than the site treated with continuous illumination following PDT ( $P$ value $<0.001$ ).