2.1.

Instrument and Calibration

Our broadband reflectance spectrometer system⁵⁰ is based on recently developed optical instrumentation.⁵¹ The original design principle is due to Wilson, Farrell, and Patterson.⁵² A schematic diagram of the system is shown in Fig. 1. The system has four major parts: light source, fiber optic probe head, dispersion system (monochromator), and charge-coupled device (CCD) camera. We simultaneously collect tissue reflectance spectra at many source-detector separations.

Figure 1

Schematic diagram of a broadband reflectance spectrometer consisting of a halogen white light source, fiber optic probe, monochromator, and liquid nitrogen-cooled CCD camera.

Briefly, light from a 250-W quartz tungsten halogen lamp (Cuda Fiberoptics, Jacksonville, Florida) was coupled to the tissue surface through the source optical fiber. Diffuse light reflected back from the tissue surface was collected by a linear array of detector fibers. Light from the detector fibers was then coupled to the entrance slit of a monochromator (Acton Research, Acton, Massachusetts). The detector fibers were arranged in a vertical line within the entrance slit, with equal spacing. The fiber tips were imaged through the monochromator, and onto a liquid nitrogen-cooled CCD camera sensor (Roper Scientific, Trenton, New Jersey). The grating dispersed the fiber output light, creating a series of vertically spaced bright strips in the image plane of the monochromator (i.e., at the CCD). Each strip represented the spectrum R _tissue (ρ,λ) dispersed from a single detection fiber with source-detector separation distance ρ. The average spectrum was obtained by binning vertically over 27 to 35 pixels.

The nitrogen-cooled CCD camera contained 330 (vertical)×1100 (horizontal) pixels (24 μm) spread over an 8×26 mm area. The pixel noise was <1 electron/s at 173 K, and the CCD quantum efficiency was 35% at 700 nm. The spectral resolution was approximately 0.5 nm/pixel for our grating (300 groves/mm). We typically set the central wavelength to 620 or 700 nm, and we collected spectra in the 380 to 880 nm or 450 to 950 nm ranges, respectively. The integration time for measurements was approximately 100 ms.

We used two linear fiber optic probe heads for the IP-PDT clinical data. One probe consisted of a source fiber and six detection fibers, each with 600-μm core diameter. The source-detector separation distance ρ ranged from 1.3 to 7.8 mm; detection fibers were equally spaced. A second probe was similar, except the fiber core diameters were 400 μm and the ten detection fibers were spaced nonuniformly with source-detector separation ranging from 0.6 to 10 mm.

To account for the spectral features of the white light source and the optical throughput of our detection system, we took additional measurements immediately after each clinical study using a 6-in.-diam integrating sphere (LabSphere, Incorporated, North Sutton, New Hampshire). A single optical fiber, coupling light from the same light source, illuminated the integrating sphere through its 0.5-in.-diam opening; the fiber optic probe head was placed in the other opening of the integrating sphere to collect a calibration signal R _sphere (ρ,λ). An effective measured reflectance R _measured (ρ,λ) was then calculated as:

2.2.

Patients and Clinical PDT Treatments

All patients were enrolled in the phase 2 study of intraperitoneal PDT conducted at the Hospital of University of Pennsylvania. The primary endpoint of this study was to evaluate the effectiveness of debulking surgery and adjuvant intraperitoneal photodynamic therapy utilizing Photofrin II. Patients with ovarian cancer, sarcoma, or gastrointestinal malignancies were enrolled in the study, and the protocol was amended to include measurements of optical properties intraoperatively. All patients signed a study-specific informed consent. The protocol was approved by the University of Pennsylvania Institutional Review Board and the Clinical Trial Scientific Review and Monitoring committee of the University of Pennsylvania Cancer Center. The protocol was conducted under an investigator-sponsored investigational new drug application (IND) with the U.S. Food and Drug Administration.

Treatment details for intraperitoneal PDT are described elsewhere.^{14 15} Briefly, patients received Photofrin (2.5 mg/kg) administered intravenously (IV) approximately 48 h before surgery. Surgery consisted of debulking of all gross tumors to a maximal residual thickness of 5 mm or less. Six sterile photodiodes were placed in the right upper quadrant, left upper quadrant, right paracolic gutter, left paracolic gutter, and midline pelvis to monitor light delivery in situ. The small and large bowel, and mesentery were treated with 532-nm green light using a flat-cut fiber; dose was monitored with a mobile flat diode. The rest of the abdomen was then treated with 630 nm light. Prior to 630 nm light treatment, the peritoneal cavity was filled with an Intralipid solution (0.01%), which scattered light and maximized light delivery to the peritoneal cavity. Light was delivered to all surfaces of the abdominal cavity by using an optical fiber sheathed within a modified endotracheal tube (balloon cuff inflated and filled with dilute Intralipid).

The cw absorption probe was sterilized prior to surgery. Before and immediately after PDT treatment, the physician placed the contact probe head on the tissue surface and then optical property measurements were taken. To minimize patient time in the operation room, one measurement was taken from each organ of each patient (with the exception of some tumor tissues). Measurements reported here were taken from normal tissues including small bowel, large bowel, peritoneum, skin (adjacent to the surgical incision), and liver, as well as from tumor tissue.

2.3.

Theory and Analysis of Optical Data

Biological tissues such as these are often modeled as homogeneous turbid media with scatterers and absorbers. Photons diffuse in such turbid media and their transport are reasonably well described by a diffusion equation, i.e.,

In our measurement scheme, source and detector fibers reside on the same tissue surface, separated by distance ρ along that surface. The tissue sample is then modeled as a semi-infinite medium. The diffuse reflectance is R[ρ,μ_a(λ),μ_s ^′(λ)]. The solutions for R[ρ,μ_a(λ),μ_s ^′(λ)] from semi-infinite media with steady-state excitation are well known:^{51 53 54 55 56 57}

The fidelity of our measurements is improved substantially because we use many optical wavelengths. For analysis we employed a multiwavelength algorithm that simultaneously fits all reflectance spectra in the wavelength range of 600 to 800 nm using multiple source-detector separation distances. To extract μ_s ^′(λ) and μ_a(λ), we have made two additional assumptions about μ_s ^′(λ) and μ_a(λ) that stabilize the analysis even more: 1. μ_a(λ)=∑_ic_iε_i(λ), and 2. μ_s ^′(λ)=Aλ^−B. ^{30 50 58} Here, ε_i(λ) is the molar spectral absorbance (or the extinction coefficient) of the i’th chromophores, and c_i is equivalent to the molar concentration of the i’th chromophore. The major chromophores in the spectral range of 600 to 800 nm are oxy- (HbO ₂), deoxy-hemoglobin (Hb), water, and Photofrin. We obtained the extinction coefficient of oxy-, deoxy-hemoglobin, and water from the literature,⁵⁹ and that of Photofrin by direct measurement using an absorption spectrometer. The approximation μ_s ^′(λ)=Aλ^−B has been shown to follow from the Mie theory over our spectral range, and thus simulates tissue scattering reasonably well.^{30 31 60 61}

We directly reconstructed the concentrations c _HbO2, c _Hb , c _water , and c _Photofrin , and the parameters A and B using a nonlinearly constrained optimization method, FMINCON, implemented in MATLAB (The MathWorks, Incorporated, Natick, Massachusetts). We constrained the upper and lower boundaries for each extracted quantity i.e., 0⩽A⩽107, 0⩽B⩽3, 0⩽C _HbO2 , C _Hb ⩽ 500 μM, 0⩽c _Photofrin ⩽40 μM, and 0⩽c _water ⩽1 . The scheme minimizes χ²,

At each separation distance, there are 441 data points with spectral resolution 0.4545 nm/pixel. To save calculation time, we used half of the dataset (221 data points) with spectral resolution 0.9091 nm/pixel for processing. The number of separation distances employed determined the final number of data points used for fitting. We reoptimized the combination of separation distances for each sample for two reasons: signal-to-noise degraded at large separation distances, and tissue heterogeneities (on the surface and below the surface) sometimes altered fitting performance. The source detector separation distances used in our analysis were typically between 1 and 5.2 mm. We also empirically found that data from the ρ=0.6 mm separation in the 10 detector probe introduced a large error while fitting. Therefore, we excluded this detector from our analysis. The typical processing time for analysis of each clinical measurement is less than 3 s on a 700-MHz Pentium III processor.

Using the reconstructed parameters A, B, and c_i, we obtained c _Photofrin directly and calculated μ_s ^′(λ), μ_a(λ), attenuation coefficients [μ _eff (λ)= 3μ _s ^′μ _a ], penetration depth [δ(λ)=1/μ _eff (λ)], total hemoglobin concentration (c _HbO2+c _Hb ), and tissue blood oxygenation [c _HbO2/(c _HbO2+c _Hb )]. We report μ_s ^′, μ_a, μ _eff , and δ at 630 nm because it is the treatment wavelength for Photofrin. Optical properties at other wavelengths between 600 and 800 nm can be calculated from these data as well. Water content was not reported in this study because its value was relatively insensitive to the fitting performance of the multiwavelength algorithm. This is probably because of its low absorption in the 600 to 800 nm spectral range. Data from tissue phantoms were used for validating the instrument and algorithm, and are reported as mean, ± standard deviation (SD), ± standard errors of the mean (SEM, the SD divided by the square root of the number of observations or sample size n), and the coefficient of variation (CV, the standard deviation divided by the mean). Accuracy is defined here as the measured value divided by the true value.

2.4.

Statistical Analysis of Clinical Data

For the clinical study, measurements were collected on 12 patients. Duplicate measurements from the same tissue were averaged prior to analysis. However, issues related to patient care prevented us from taking measurements on all tissues at both before and after PDT time points, and from both tumor and normal tissues for all patients. To compare results from different normal tissues, only data from those patients receiving measurements in all tissues before or after PDT were analyzed. To compare results before and after PDT, ratios were formed for individual patients receiving measurements in a specific tissue both before and after PDT. Similarly, tumor-to-normal tissue ratios of each parameter of interest were formed for each individual and used in the analysis. Statistical analyses were carried out using the freeware R 1.7.0.⁶²

Because, to our knowledge, this is the first report of data of this type, we graphically present the data for individual patients, and we provide summaries of the data, i.e., medians and or means, together with either the SEM, standard error of the mean, or a 95% confidence interval (95% CI) on the mean, and number of patients n. The confidence intervals were based on the T distribution and the estimated SEMs; if a 95% confidence interval did not cover a hypothesized value, e.g., value 1.0 for a mean of a ratio, then it indicates that the means were significantly different from the hypothesized value. Similarly, if the p-value for a test was smaller than a type 1 error rate of 0.05, we declared the test to be statistically significant. All hypothesis tests were two-sided. Friedman’s test, a nonparametric analog of a repeated measures analysis of variance, was used to determine whether there were significant differences in each optical and physiological property between at least one pair of normal tissue types.⁶³ If significant differences were found, then a paired t-test was used to make pairwise comparisons among the tissues. By using a global test for each parameter of interest (Friedman’s test) followed by a pairwise test in cases where the global test achieves statistical significance, we reduced the number of tests used in the analyisis. Thus in Table 2, 46 pairwise tests out of 96 possibilities were constructed. By using this approach, we hoped to reduce the number of false positive tests.

To compare parameters before versus after PDT, we did not preceed our pairwise tests with a global test because of certain parameters (e.g., Photofrin concentration), the number of patients with measurements in all tissues was too small to carry out the test. Descriptive statistics (means of ratios and their associated 95% CIs) and hypothesis tests (paired T-test) for these ratios were based on log-transformed data. Because measured concentrations of Photofrin were zero in some tissues, differences instead of ratios were used to analyze these data, and the data were not log transformed. If the lower endpoint of a 95% CI was negative, it was truncated to zero if a negative value was meaningless.

2.5.

Validation of the Instrument and the Algorithm with Tissue Phantoms

We validated the stability and accuracy of our instrument and the global algorithm using tissue-simulating phantoms. Three types of phantom were used. One was a scattering phantom, Intralipid (IL) at various concentrations to validate μ_s ^′. A second phantom was a mixture of hemoglobin (human blood) and Intralipid to validate the physiological properties of total hemoglobin concentration (THC) and blood oxygen saturation (%S _t O ₂). Finally, Photofrin was added into the hemoglobin phantom to validate Photofrin concentration. A magnetic stirring rod was used during the phantom studies to insure solution uniformity.

A 30% Intralipid solution was diluted to 0.5, 1, 1.5, and 2%. The solution had μ_s ^′=5, 10, 15, and 20 cm⁻¹ at 830 nm, respectively, according to Mie theory approximations,⁶⁰ which predict that μ_s ^′ of 10% Intralipid is given by μ_s ^′=μ_s(1−g) [cm ⁻¹], with μ_s=(2.54×10⁹)(λ [nm]^−2.4) [cm ⁻¹] and g=1.1−(0.58×10⁻³)(λ [nm]). Then, the μ_s ^′ of X% Intralipid was calculated using the relation μ_s ^′X%IL ×V^X%IL =μ_s ^′10%IL ×V^10%IL for given volumes (V) of 10% Intralipid and X% Intralipid and μ_s ^′10%IL . The results from ten measurements at each Intralipid concentration show that the measurement coefficient of variance (CV) was typically less than 0.3% and the measurement accuracy was 135, 106, 105, and 101% for 0.5, 1, 1.5, and 2% IL, respectively. The reconstructed μ_s ^′ for the lower concentration solution (i.e., μ_s ^′⩽5 cm ⁻¹ at 830 nm) was clearly overestimated.

In the hemoglobin phantoms, we simultaneously monitored the oxygen partial pressure (pO ₂) by using a Clark-type electrode probe. After adding fresh human blood into the 1% Intralipid and waiting until a pO ₂ equivalent to air oxygen concentration (21% or 159.6 Torr) was achieved, the sample oxygenation was decreased by pumping nitrogen gas into the phantom solution continuously. Optical and partial pressure measurements were taken in intervals that varied from 30 s to 5 min depending on the speed of pO ₂ variation. The measured blood oxygen saturation (%S _t O ₂) is plotted versus pO ₂ in Fig. 2(a). The data were fit to a solid line in Fig. 2(a) using an optimal P-spline with 95% confidence intervals (dotted lines) shown.⁶⁴ The oxygen dissociation curve or Hill curve of human blood⁵¹ is also shown in the same figure for comparison (dashed line). Compared to this Hill curve, the fitted curve suggests %S _t O ₂ is slightly overestimated at low pO ₂, and slightly underestimated at high pO ₂. However, the difference between the smoothed fit and the theoretical values (Hill curve) is less than 5%.

Figure 2

Measured tissue blood oxygenation (%S _t O ₂), total hemoglobin concentration (THC), and Photofrin concentration (c _Photofrin ) in hemoglobin phantoms using broadband reflectance spectroscopy. (a) %S _t O ₂ was plotted versus electrode probe measured oxygen partial pressure (pO ₂). The solid line was fit to the data using an optimal P-spline. Dotted lines are approximately 95% confidence intervals on the mean fit. A human blood dissociation curve (dashed line) was plotted for comparison. (b) THC was measured at 50 (□) and 100 μM (○) and plotted versus time. (c) Individual measurements in the hemoglobin phantom with five different Photofrin concentrations were plotted versus the number of measurements. Ten measurements were taken at each Photofrin concentration. (d) The mean and standard deviation of measured Photofrin concentrations were plotted versus the true concentration. The measurements underestimated the true concentration at lower concentration (<5 μM), but are accurate otherwise.

The stability and accuracy of total hemoglobin concentration (THC) measurements were determined from similar hemoglobin phantoms having 0.8% IL and two different THC concentrations of 50 and 100 μM, respectively, as shown in Fig. 2(b). The mean values (±SD) from 50 to 60 repeated measures of THC were 44.8±1.3 and 111.5±3.5 μM with a corresponding coefficient of variation (CV) of 2.9% and 3.2% and an accuracy of 110.4% and 111.5%, respectively. The stability of %S _t O ₂, μ_s ^′, μ_a, and μ _eff was also validated in the hemoglobin phantom with THC 100 μM. The mean values (±SD) of %S _t O ₂ and μ_s ^′, μ_a, and μ _eff at 630 nm were 70.5%±1.2%, 12.1±0.07 cm⁻¹, 0.216±0.003 cm⁻¹, and 2.824±0.013 cm⁻¹ with corresponding CVs 1.7, 0.6, 1.3, and 0.6%, respectively.

Figures 2(c) and 2(d) show the results of measuring Photofrin concentration in the hemoglobin phantom (THC 50 μM, 0.8% IL). Photofrin was increased from 0 to 15 μM (corresponding to 7.5 mg/kg, assuming the molecular weight of Photofrin is 500) with increments of 3.7 μM. At each concentration, ten measurements were performed [Fig. 2(c)] and the measured mean values (±SD) are 1.6±0.9, 6.2±0.6, 10.6±0.9, and 15.6±1.4 μM, respectively, with corresponding coefficients of variation (CVs) of 56, 9.7, 8.5, and 9.0%, respectively [Fig. 2(d)]. Overall, the values of measured Photofrin concentration fluctuated more than other optical and physiological properties. Photofrin concentration was also underestimated at low true Photofrin concentration (=3.7 μM). However, the values of other optical (μ_s ^′, μ_a, and μ _eff ) and physiological properties (THC and %S _t O ₂) were measured to stay constant (within measurement stability) during the entire experiment (data not shown).

3.1.

Algorithm Performance in Clinical Data

The performance of the fitting between measured and calculated spectra is shown in Fig. 3. Figure 3(a) shows the normalized plot of measured and calculated diffuse reflectance spectra from large bowel, as an example, using the six-detector probe. The source-detector separation distances used in this case to extract the optical and physiological parameters were 1.3, 2.6, 3.9, and 5.2 mm. The reconstructed diffuse reflectance spectra were back calculated based on Eq. (3) by applying the algorithm outputs A=1499.5, B=0.8423, c _HbO2=122.42 μM, c _Hb =37.25 μM, and c _Photofrin =4.3 μM. The fitting error ξ (defined previously) is 7.1%. To compare the fitting performance in human tissues and tissue phantoms, Fig. 3(b) shows a similar plot from a hemoglobin phantom using a similar range of source-detector separation distances (1.2, 1.8, 2.4, 3, 4, 5, and 6 mm) of the ten-detector probe. The fitting error is 2.3% in tissue phantom, three times less than in the large bowel tissue sample. These differences could have arisen from a variety of factors that are difficult to know and/or control including tissue heterogeneity.

Figure 3

Measured (—) and calculated (----) reflectance spectra at ρ=1.3, 2.6, 3.9, and 5.2 mm in (a) a normal large bowel tissue sample and at (b) ρ=1.2, 1.8, 2.4, 3, 4, 5, and 6 mm hemoglobin phantom, after normalized to the spectrum at ρ=1.3 and 1.2 mm, respectively. The spectra from top to bottom denote the increasing values of ρ (as indicated in arrows). The calculated reflectance spectra were calculated based on the output values of the global algorithm that A=1499.5, B=−0.8423, c _HbO2=122 μM, c _Hb =37 μM, and c _Photofrin =10 μM for the large bowel tissue sample, and A=1500.5, B=−0.7801, c _HbO2=32 μM, c _Hb =21 μM, and c _Photofrin =15 μM for the hemoglobin phantom. The tissue sample shows a larger fitting error (ξ=7.1%), defined in the text, than tissue phantom (ξ=2.3%) as expected.

3.2.

Between and Within Patient Heterogeneity

Variation in μ_s ^′, μ_a at 630 nm, and THC, all measured before PDT in three intraperitoneal tissues, is shown both between individual patients and within patients (Fig. 4). Patients with complete data on all five sites where optical measurements were collected (C1 to C7) are distinguished from those with incomplete data on at least one site (I1 to I5). The plot shows the variability in %S _t O ₂ before PDT for six patients with complete data on all five sites (C1 to C6) and for six patients with incomplete data on at least one site (I1 to I6). For patients with complete data, the normal intraperitoneal tissues have %S _t O ₂ in the range of 32 to 100% and a THC in the range of 19 to 263 μM.

Figure 4

Boxplots showing median (horizontal midline) and mean (squares) in μ_s ^′(λ=630 nm), μ_a(λ=630 nm), %S _t O ₂, and THC for three intraperitoneal sites (small and large bowel and peritoneum) in individual patients. Vertical lines extend to largest and smallest values for each patient and the box bisects the distance between the median and either the smallest or largest datapoint. The C series indicates patients with complete data from all five sites (intraperitoneal skin and liver); the I series indicate patients with incomplete data in at least one site.

Within the intraperitoneal tissues before PDT, substantial heterogeneity was observed between individual patients in μ_s ^′, μ_a at 630 nm, %S _t O ₂, and THC. For example, median values of μ_s ^′ ranged from 6.2 cm⁻¹ for C4 to 22.9 cm⁻¹ for I2, and the median values of %S _t O ₂ ranged from 51.8% for C6 to 97.3% for C5. The range of measurements for μ_s ^′, μ_a at 630 nm, and THC within patients was highly variable as well. For example, values of μ_s ^′ for patient C4 ranged from 4.9 to 6.6 cm⁻¹, while μ_s ^′ for patient I2 ranged from 13.5 to 47.6 cm⁻¹.

3.3.

Differences among Normal Tissues

The means (±SEM) of optical properties (μ_s ^′, μ_a, and μ _eff at 630 nm) and physiologic parameters (%S _t O ₂, THC, and c _Photofrin ) and their corresponding sample sizes for five different normal tissues (small and large bowel, peritoneum, skin, and liver) before and after PDT are listed in Table 2. In the pre-PDT measurements, Friedman’s test for each optical property, μ_s ^′, μ_a, and μ _eff , showed evidence of statistically significant differences among at least one pair of the five tissues. Mean values for liver were substantially higher than for all other tissues, with skin and peritoneum somewhat higher than small and large bowel. In pairwise comparisons, differences between liver and other tissues were consistently significant. In addition, for μ _eff , means for small and large bowel were significantly smaller than for peritoneum.

Table 2

Qualitatively, the trends in optical properties were similar in post-PDT measurements, although measurements were collected on fewer subjects. Mean values for liver were consistently higher than for other tissues, with peritoneum and skin for μ_s ^′ and μ _eff tending to be higher than small and large bowel. Friedman’s test indicated that differences between at least one pair of tissues were statistically significant for μ_s ^′ and μ _eff . With the exception of μ_s ^′ in skin, pairwise differences in mean values for liver and all other tissues were consistently significant for μ_s ^′ and μ _eff .

For the comparison of physiological properties, liver was omitted from the analysis because we were unable to extract the oxygenation and THC information from liver tissues in our current analytical model. The diffusion model breaks down at high absorption (μ_a/μ_s ^′>0.1), wherein the ratio of absorption to scattering is large; the optical properties of liver fall in this range. For the comparison of Photofrin concentration among tissues, skin was also omitted for the same reason. For the other tissues, data were available for between five and seven patients before PDT, and for three patients after PDT. Friedman’s test did not find any significant differences in the distribution of physiological properties among tissues except for %S _t O ₂ before PDT. Pairwise tests indicated that means of %S _t O ₂ for large bowel were significantly higher than peritoneum and skin, and approached statistical significance for small bowel as well (P=0.065).

3.4.

Differences in Normal Tissue before Versus after PDT

Because of the large amount of variability among patients, differences in optical and physiologic parameters before versus after PDT were analyzed only for patients who had both measurements for the parameter of interest. Table 3 shows the mean of the individual ratios (95% CI) of optical properties, %S _t O ₂, and THC before versus after PDT. The concentration of Photofrin was sometimes measured as zero before or after PDT. Instead of calculating ratios, the mean difference in Photofrin concentration before versus after PDT was reported in Table 3. For the optical properties, and for %S _t O ₂, the mean of the individual ratios were qualitatively similar to 1.0 and no significant differences from 1.0 were found. In small bowel and peritoneum, mean THC ratios exceeded 1.0 before versus after PDT, and for small bowel, the ratio was significantly larger than 1.0 (P=0.013). For Photofrin, decreases in mean concentrations after-PDT compared to before-PDT ranged from 2.0 to 5.5 μM, and were significantly different from zero in the peritoneum (P=0.026).

Table 3

3.5.

Differences between Normal and Tumor Intraperitoneal Tissues

The optical and physiological properties of tumors from six patients, and their corresponding normal intraperitoneal tissues, are shown in Figs. 5 and 6, respectively. Table 4 shows the mean values of each parameter along with the mean of the individual tumor-to-normal tissue ratios for all parameters except c _Photofrin . For c _Photofrin , mean differences are shown. For the optical properties, μ_s ^′ at 630 nm from tumors tended to be similar to normal small and large bowels, but significantly smaller than peritoneum (P=0.024). In contrast, μ_a and μ _eff at 630 nm from tumors tended to be somewhat higher than from normal intraperitoneal tissues, but the ratios were not significantly different from 1.0. The mean %S _t O ₂ in tumors was 33% with values ranging from 11 to 44%. The mean %S _t O ₂ in tumors was significantly lower than those of the corresponding normal intraperitoneal tissues in the same patients for both small bowel and peritoneum (P=0.018 and P=0.004, respectively). The mean THC in tumor was 117 μM with values ranging from 61 to 224 μM, while mean Photofrin concentration was 7.4 μM with values ranging from 1.27 to 15.6 μM. Differences between THC and Photofrin in tumor and normal intraperitoneal tissues were not significant. However, for Photofrin, we noted the mean concentrations in tumor tended to be higher than for normal intraperitoneal tissues.

Figure 5

Boxplots showing median (horizontal line) and means (squares) of optical properties for patients with measurements in both tumor (T) and intraperitoneal tissues (S small bowel, L large bowel, P peritoneum). Vertical dashed lines extend to largest and smallest values for each tissue, and the box indicates the interquartile range, roughly the middle half of the data.

Figure 6

Boxplots showing median (horizontal line) and means (squares) of physiologic properties for patients with measurements in both tumor (T) and intraperitoneal tissues (S small bowel, L large bowel, P peritoneum). Vertical dashed lines extend to largest and smallest values for each tissue, and the box indicates the interquartile range, roughly the middle half of the data.

Table 4