2.1.

Instrumentation

The system, SpectraCure P18 with IDOSE^® (Interactive Dosimetry by Sequential Evaluation), is functionally similar to previously described instruments.^{20, 21} However, for this study, it was scaled up to accomodate up to 18 optical fibers for light delivery and monitoring. The system can operate in either of two modes: light-delivery mode or monitoring mode. All optical fibers are used both for light delivery and monitoring of optical properties. Switching between the two modes is accomplished by means of an internal optomechanical fiber switch. Eighteen individually controlled diode lasers with a laser power of $250\mathrm{mW}$ and operating at a wavelength of $652\mathrm{nm}$ are used as light sources. The output power from each bare-end fiber is calibrated to $150\mathrm{mW}$ prior to each treatment session. Bare-end fibers are used to ensure well-defined measurement points. In monitoring mode, the system alternates between three different monitoring light sources: a laser diode of the same type as used for treatment, a blue light-emitting diode (LED) centered at $410\mathrm{nm}$ , and a twin near-infrared (NIR) LED operating in the range of $750–850\mathrm{nm}$ . While in monitoring mode, for each cluster of seven fibers, one fiber is emitting light while the other six surrounding fibers are used to collect scattered light. The detecting fibers are then coupled, through the optical switch, to individual spectrometers that operate in the range of $640–880\mathrm{nm}$ . The monitoring cycle is completed when all 18 fibers have emitted light once, meaning that a full monitoring sequence results in $18\times 6$ measurements for each of the three monitoring light sources. With this setup, it is possible to monitor the optical properties of the tissue at the treatment wavelength, the fluorescence of the photosensitizer excited at 410 and $652\mathrm{nm}$ , and the oxygen saturation through NIR spectroscopy. The system also supports the use of up to six additional verification fibers, which are only used to monitor the light fluence at critical positions during treatment light delivery. The system is CE approved as a class IIb medical device for use with Foscan^® for prostate cancer.

2.2.

Dosimetry

A more detailed description of the dosimetry platform, IDOSE^®, is found in Refs. 22, 23. Briefly, a 3-D model of the prostate and surrounding tissues is generated from transrectal ultrasound (TRUS) images. The software calculates optimum fiber positions and presents the positions to the user. When the optical fibers are in place, a first monitoring sequence is performed. The optical properties are evaluated by assuming that the tissue is locally homogeneous around each subcluster of seven optical fibers (one emitter plus six receivers). A simple fit, based on the diffusion approximation, recovers the effective attenuation coefficient for each cluster. The effective attenuation coefficients are used to compute a dose plan that is presented to the user, who then has to approve the start of the treatment. At specific intervals, the light delivery is interrupted for monitoring sequences. In each monitoring sequence, the effective attenuation coefficients are evaluated and the dose plan is updated if significant changes occur to the optical properties that would perturb the delivered light dose. In the current dose planning algorithm, based on Cimmino’s method, the aim is to deliver a minimum threshold light dose to all parts of the target volume while attempting to minimize exposure to the surrounding organs outside the target tissue. The light propagation is determined by the optical properties; hence, the basis for the dosimetry protocol is the measurement of these properties. In this study, the photosensitizer fluorescence and oxygen saturation data were not used in the computation of the dose plan because no dosimetry model to incorporate this information is yet available. Data are instead collected to serve as the basis for future development. Relying on the light fluence dose only and not taking the photosensitizer nor oxygenation into account is a simplification to the PDT dosimetry problem. Despite this simplification, it is a relevant first approach because the light fluence dose distribution is the only parameter that can be controlled during treatment by adjusting the light delivery time.

2.3.

Threshold Light Dose

The threshold light dose was set to $5\mathrm{J}∕{\mathrm{cm}}^2$ (fluence). This was based on estimates of the lesion size of necrosis determined by MR imaging (MRI) in the Moore study,^{24, 25} and data of optical properties of prostate tissue taken from a previous study.²⁶ With a lesion radius of $7\mathrm{mm}$ and assuming homogeneous optical properties of ${\mu }_{\mathrm{a}}=0.6{\mathrm{cm}}^{-1}$ and ${\mu }_{\mathrm{s}}^{\prime }=10{\mathrm{cm}}^{-1}$ , this yields an estimated threshold dose for tissue necrosis of $\sim 8.5\mathrm{J}∕{\mathrm{cm}}^2$ . This estimate was further verified by comparing to the EMEA-approved recommended light dose for superficial illumination, $20\mathrm{J}∕{\mathrm{cm}}^2$ . Assuming optical properties ${\mu }_{\mathrm{a}}=1{\mathrm{cm}}^{-1}$ and ${\mu }_{\mathrm{s}}^{\prime }=10{\mathrm{cm}}^{-1}$ and a depth of necrosis of $5\mathrm{mm}$ , this yields a threshold dose of $6–10\mathrm{J}∕{\mathrm{cm}}^2$ (fluence), which is in the same range as estimated from the prostate study. The threshold dose for this clinical study was conservatively set to $5\mathrm{J}∕{\mathrm{cm}}^2$ as precaution to avoid damage to organs at risk because this was the first time IDOSE was used clinically.

2.4.

Tissue Phantom Experiments

The capability of the instrument to assess optical properties interstitially was tested within an optical phantom study preclinically. Eight different optical phantoms composed of varying concentrations of Intralipid^© (Fresenius Kabi, Uppsala, Sweden), ink (Pelican Fount, Feusisberg, Switzerland), and water were made to mimic different scattering and absorption properties. Absorption and scattering coefficients, chosen to correspond to a broad range of relevant prostate optical properties,^{16, 25, 26} were evaluated separately using a time-of-flight (TOF) spectroscopy instrument as described in Ref. 26. Briefly, this TOF instrument operates by sending a short laser pulse into the medium. The diffusely scattered light is detected, using time-correlated single-photon counting electronics, at some distance where the laser pulse is temporally dispersed. On the basis of the temporal dispersion, the absorption and scattering coefficients can be deduced. The optical properties are stated in Table 1 . The fibers were positioned, using a fiber grid, into the liquid phantoms, and a complete treatment sequence was executed for each phantom.

Table 1

Tabulated optical properties for the intralipid phantoms as evaluated using TOF spectroscopy.

2.5.

Clinical Study

The clinical study was performed at Malmö University Hospital, Sweden, and Karolinska University Hospital, Stockholm, Sweden. The inclusion criteria for patients into this clinical study included histologically proven, untreated, organ-confined prostate cancer (stage T1c, Gleason score <7, PSA $<10\mathrm{ng}∕\mathrm{ml}$ ). This nonrandomized clinical study, approved by the Swedish Medical Products Agency, is currently still recruiting patients, and each patient has a follow-up time of $12\text{months}$ . Patients were sensitized with Foscan (Biolitec Pharma) according to the European Medicines Agency (EMEA)-approved protocol for head and neck cancer: $0.15\mathrm{mg}∕\mathrm{kg}$ given intravenously, and the drug-light interval was four days. Patients were anesthetized either by general of epidural anaesthesia, placed in the lithotomy position, and TRUS images were acquired as transverse slices $5\mathrm{mm}$ apart aided by a mechanical stepper on which the ultrasound transducer was mounted (B-K Medical, Herlev, Denmark). Tissue contouring was performed by the urologist, and the system then computed the fiber positions following a three-dimensional rendering of the target volume. Eighteen-gauge needles were inserted transperineally under TRUS guidance, aided by a mechanical grid template with a coordinate system calibrated to the ultrasound scanner. The actual coordinate positions, which could differ slightly from those recommended by the system, were reported back and entered into the dosimetry software. Bare-end $600\text{-}\mu \mathrm{m}$ core diameter silica fibers (CeramOptec, Bonn, Germany) were inserted into the needles and extended $2\mathrm{mm}$ beyond the needle tip. A first monitoring sequence was run to establish optical properties and compute the first dose plan, then light delivery commenced. The light delivery was interrupted at intervals of 2, 4, 9, 14, 19, and $29\min$ for monitoring sequences. The dose plan was updated at each monitoring sequence. A constant light power of $150\mathrm{mW}$ was emitted from each fiber, and the illumination time for each fiber was varied to achieve a dose according to the dose plan.

The ability of online dosimetry to optimize the light dose plan was assessed using dose maps and dose-volume histograms (DVH), and by reviewing the measurements from the monitoring sequences. Clinical assessments were made using MRI, prostate-specific antigen (PSA), and biopsy. Baseline MRI was done approximately two weeks before PDT using a protocol similar to that reported in Ref. 24. Follow-ups were then done at two weeks, eight weeks, six months, and $12\text{months}$ after PDT. PSA levels were checked regularly throughout the study. Twelve-core biopsy was done six months after PDT.

3.1.

Tissue Phantom Experiments

The evaluated effective attenuation coefficient for each tissue phantom is displayed in Fig. 1 . It is seen that the steady state spatially resolved measurements performed within the IDOSE scheme underestimates the effective attenuation coefficient by $\sim 10\mathrm{\%}$ relative to the time-resolved evaluation. The evaluated optical properties were introduced into a real 3-D prostate organ model. This allowed the optimal treatment time and subsequently the light fluence dose to be calculated. By comparing the light dose calculated from spatially resolved measurements with the light dose calculated from time-resolved measurements, it was possible to determine the amount of treatment underestimation. Figure 1 shows the relative change, $\Delta \mathrm{TF}$ (in percent), of the treatment fraction due to underestimation of the effective attenuation coefficient. Treatment fraction represents the relative tissue volume of the total target volume that receives the threshold dose. Here, $\Delta \mathrm{TF}$ is defined by

Fig. 1

(a) Evaluated effective attenuation coefficient using IDOSE (white) and TOF (black) for different tissue phantoms. The mean of the evaluated ${\mu }_{\mathrm{eff}}$ for all 18 fibers are shown, and the error-bars indicate $\pm 1$ standard deviation. (b) The relative difference in treatment fraction, defined by 1, between optical properties evaluated with IDOSE or TOF.

3.2.

Clinical Study

Four patients, aged 70, 60, 56, and $67\text{years}$ underwent Foscan-mediated PDT. The prostate volumes were 40 (36), 36 (46), 43 (46), and 30 $\left(40\right){\mathrm{cm}}^3$ , respectively, as determined by TRUS with an ellipsoid approximation (numbers in parentheses represent volumes evaluated by the voxel-based IDOSE 3-D tissue model). For all four patients, all 18 optical fibers were used to deliver therapeutic light and acquire measurements used for light fluence dose planning. In addition, two to four verification fibers were inserted at the capsule border and close to the rectal wall.

3.2.1.

Dosimetry

The DVH, i.e., fraction of the tissue volume that receives a certain light fluence dose is presented in Figs. 2, 2, 2, 2 for all four patients. These DVHs are based on the calculated dose to the tissue after the treatment was finished. Prior to the start of the study, dose acceptance limits for the different tissue types were agreed on after consultation with the urologist. The acceptance limits are listed in Table 2 together with the dose metrics for all four patients. The dose metric values calculated for the sphincters differ significantly between the patients. This difference is due to smaller volumes of tissue compared to the other tissue types. Fig. 3 shows an example (patient 3) of calculated dose maps, which indicate the extent of the light dose distribution in the tissue. The dose maps represent calculated light fluence dose after the treatment, based on the 3-D geometry of the organs, the positions of the optical fibers, the measured optical properties at each fiber, and the known light energy delivered by each fiber. Figure 3 shows what parts of the different organs had received a light dose below or above the threshold dose, respectively. Figure 4 illustrates the predicted total light dose before and after each monitoring sequence during treatment progression as well as the final dose delivered for all four patients. The predicted dose plan before measurements is calculated based on average optical properties taken from Ref. 26. Additional light delivery parameters are listed in Table 3 .

Fig. 2

Dose-volume-histogram for (a) patient 1, (b) patient 2, (c) patient 3, and (d) patient 4. Three tissue types are included indicated by (○) prostate, (▽) urethra, and (◻) rectum.

Fig. 3

Dose maps for patient 3, at the depths (a) 10, (b) 15, (c) 20, and (d) $25\mathrm{mm}$ along the craniocaudal axis. Color bar shows light fluence dose in joules per centimeters squared. (Color online only.)

Fig. 4

The predicted total dose before and during treatment, as well as the final delivered dose for patient 1–4.

Table 2

Dose metrics for patients 1–4 compared to acceptance limits.

TissueType	Limit(%)	Patient
		1(%)	2(%)	3(%)	4(%)
Prostate	95	98.7	99.2	98.3	99.1
Urethra	$<90$	87.4	82.3	71.6	79.7
Rectum	$<80$	10.7	7.7	2.0	0.0
Upper sph.1	$<80$	22.0	54.9	8.4	0.0
Lower sph.1	$<50$	37.2	0.0	25.1	0.0

^aShort for sphincter.

Table 3

Light delivery parameters for patients 1–4.

Patient	1	2	3	4
Max. time/fiber (min)	12	29	15	14
Average time/fiber (min)	6	10	7	9
Min. time/fiber (min)	2	2	2	2
Total light dose (J)	986	1681	1155	1404
Average dose/fiber (J)	55	93	64	78
Dose per unit volume $(\mathrm{J}∕{\mathrm{cm}}^3)$	27	37	25	35

3.2.2.

Optical properties

The effective attenuation coefficient as calculated during treatment for each fiber is shown in Figs. 5, 5, 5, 5 for the four patients within the clinical trial. The error bars in Fig. 5 denote the standard deviation of all measurements performed during treatment, for each treatment fiber. The standard deviation seems to indicate that the optical properties did not vary markedly during treatment. The square markers denote evaluations that have an $r^2$ -value above 0.5.

Fig. 5

The effective attenuation coefficient as a function of fiber number for (a) patient 1, (b) patient 2, (c) patient 3, and (d) patient 4. The line indicates the mean of the effective attenuation coefficient for all monitoring sequences times where the error bars indicate $\pm 1$ standard deviation. The ◻-marked values mark evaluations with $r^2\text{-value}⩾0.5$ .

3.2.3.

Clinical assessment

There were no serious complications for any of the patients during the PDT treatment. The urinary catheters could be removed one to three days after treatment. The patients experienced urgency for about one week after the treatment. All patients, who previously had suffered from varying degree of voiding difficulty, experienced relief of urinary tract symptoms with a lowering of mean International Prostate Symptom Score. PSA values for all patients are shown in Fig. 6 . PSA levels for all four patients were stabilized at a level below or at the level pre-PDT.

Fig. 6

PSA values from patients 1–4. Immediately after PDT, PSA rose to $40–78.1\mathrm{ng}∕\mathrm{ml}$ . Patient 1 had a prostatectomy after $45\text{weeks}$ . Patient 4 was withdrawn from the study at $35\text{weeks}$ .

MRI scans from patient 2 are shown in Fig. 7 . Zones of decreased enhancement were visible in some cases two weeks post PDT, as illustrated in Fig. 7. No necrotic lesions were seen in the MRI images two months post PDT. The prostate volume was evaluated from MRI at baseline, two weeks and eight weeks post-PDT. A clear decrease (23–35%) of the prostate volume was seen in all four patients, indicating a significant treatment response of the tissue. The volume reduction can be seen in the scans in Fig. 7. No tumors were visible on MRI scans in any of the patients.

Fig. 7

(a) Baseline and (b) two weeks post-PDT contrast enhanced MRI scans from patient 2. The dark structure below the prostate is the rectum. The arrows mark zones of reduced enhancement.

Tissue biopsies taken at six months follow up revealed viable cancer cells in patients 1, 2, and 4. Samples from patient 3 were benign, indicating either a successful treatment response or that the biopsy failed to sample viable cancer cells. Tumor viability was indicated by positive immunohistochemistry for proliferation, bcl-2 expression, and focal p53 upregulation. However, in all biopsies from the four patients, there were areas of secondary changes in the benign epithelium, such as vacuolisation, fibrosis, and hyaline degeneration. These histological changes indicate a response to the treatment but seem to be focal and without involvement of confluent areas of necrosis.

4.1.

Tissue Phantom Experiments

An underestimation of the effective attenuation will cause an alteration of the treatment volume. As seen in Fig. 1, the underestimation of the effective attenuation renders a lower treatment fraction; hence, the prostate tissue receives a lower light fluence dose, at worst 7% for phantom G. Despite a decrease of the treatment fraction the threshold dose is still reached in 90% of the prostate tissue for all phantoms. This outcome is in analogy with the discussion by Johansson, ²² where it was concluded that the dosimetry algorithm adheres to a certain robustness against changes in the effective attenuation coefficient.

The effective attenuation coefficient is compared between IDOSE and a TOF spectroscopy system. It should be noted that after the validation experiments were performed, our group has reported on improvements on the evaluation scheme as well as a calibration method of time-resolved measurements.^{27, 28} Herein it was discussed that the former rationale, adopted within this paper, slightly overestimated ${\mu }_{\mathrm{a}}$ and ${\mu }_{\mathrm{s}}^{\prime }$ ; hence, we believe that the accuracy might be better than shown in Fig. 1.

4.2.

Clinical Study

4.2.2.

Optical properties

The optical properties shown in Fig. 5 are within the range of previously reported values for human prostate.^{16, 26, 27} Svensson ^{26, 27} reported on pronounced inter- and intrapatient variations of the optical properties. The effective attenuation coefficient assessed by IDOSE shows similar variations. These variations should be analysed with care because the time-resolved scheme employed by Svensson ^{26, 27} probes the tissue between two optical fibers, whereas the IDOSE scheme probe a substantially larger portion of the prostate. The assumption made within the spatially resolved evaluation protocol is that the region around each treatment fiber is homogeneous. Hence, due to the evaluation model, we would expect a more homogeneous appearance of the effective attenuation coefficients because it averages ${\mu }_{\mathrm{eff}}$ over a larger volume. Intrapatient variations due to abnormal prostate tissue features, such as calcifications, heterogeneous vascularization, and blood occlusions in front of fibers, will influence the measurements. Because the probed volume is large, such heterogeneites will appear as outliers in the measurements leading to poor linear fitting, which in turn can explain the low $r^2$ -values in Fig. 5. As discussed by Johansson, ²² and also seen in this paper, the dosimetry algorithm is still capable of delivering a predefined dose correctly when the effective attenuation coefficient is slightly altered. This means that the heterogeneous ${\mu }_{\mathrm{eff}}$ has a limited effect on the dose calculations. On the other hand, potential calcifications and blood occlusions close to a source fiber can attenuate the treatment light, effectively decreasing the output power.

In order to assess how large effect potential output power perturbations will cause on the treatment fraction a simple first approximation is considered. The perturbation is mimicked by only imposing a multiplicative factor, within the interval [0, 1], on the output power when performing the dose calculations. The optical properties and final treatment times for the patients are considered, where the dose calculations are repeated 100 times for each patient. At each simulation, the perturbations are randomly sampled from a uniform distribution of the interval [0, 1]. The treatment fraction for the prostate is seen together with the true treatment fraction, in Fig. 8 , which indicates that, on average, the treatment fraction is diminished by $<20\mathrm{\%}$ due to occlusions in front of the fiber. This analysis is based on the treatment fraction, which does not account for spatial variations within the target volume. This is significant because the local dose will be limited in situations when a fiber is subject to a high output power attenuation. In order to analyze the spatial dependencies on the fiber perturbations, a more rigorous approach is required.

Fig. 8

Treatment fraction (TF) for the patient data where (⋆) denote true treatment fraction and (◻) indicate treatment fraction when output power is perturbed. The error bars show $\pm 1$ standard deviation of 100 randomized samples of fiber-specific attenuation.