2.1.

Patient Selection

Women with an abnormal Pap smear, routinely referred to the clinic for colposcopic examination, were asked to participate in this study. Pregnant or lactating women were excluded. A total of 30 volunteers enrolled in this study. Their ages varied between 19 and 45 years with a mean age of 30 years. We excluded six volunteers with a history of previous surgery that had caused unknown changes to the epithelium. Another eight volunteers were colposcopically unsuspect for CIN and consequently no biopsies could be taken. Finally, 19 biopsies from 16 volunteers were available for our analysis.

2.2.

Application of ALA

The ALA was purchased from Medac GmbH (Hamburg, Germany) in crystallized form. On the day of use it was dissolved in a concentration of 1 (wt/wt) in sterile saline solution with the pH adjusted to 5.5 using sodium hydrogen carbonate.²¹ Sterile gauzes drenched with 10 mL of this solution were applied to the cervix. The anticipated application time was 60 min, but in practice it ranged from 60 to 120 min.

2.3.

DR Fluorescence Imaging Prototype

The DR fluorescence camera we used is a general purpose prototype developed by our group. In practice it resembles an operating microscope. Preliminary clinical experiments were performed for neurosurgery and dermatology. The device uses two violet wavelengths for excitation of the cervix, 405 and 435 nm at intensities of about 100 and 200 μW/cm², respectively. These are not the optimum wavelengths for DR imaging of PpIX, but it is a compromise between the contrast and signal to noise ratio. Excitation light is generated with a 200 W Hg–Xe lamp, two line filters, and a chopper that alternates the two excitation wavelengths. Fluorescence of the cervix is detected with an image intensified camera (Philips IP 800) through a red and a green filter (550 nm, bandwidth 42 nm and 675 nm, bandwidth 110 nm, respectively). In this manner four different fluorescence images are acquired. These are processed into DR images by ratioing the two red and green images and subsequently ratioing the two ratio images. A detailed description of this imaging device and the theory behind it can be found elsewhere.^{22 23 24} For focusing and positioning a simple white light (monochrome) imaging mode is available.

2.4.

Procedure

At the time of colposcopy the moist gauze containing the ALA is removed. Then a regular colposcopic procedure is performed, including acetic acid staining and photographic documentation of the cervix. In case the decision is made to take a biopsy, the location at which the biopsy will be taken is marked on the photograph (a Polaroid) that was taken to document the colposcopy procedure. Subsequently, the colposcope is removed and the fluorescence imaging device is positioned and focused. Positioning the device took 1–5 minutes, acquiring the fluorescence images took approximately 60 s. The resulting DR image was not available to the clinician during the procedure. Next, the fluorescence imaging device is removed and the colposcope is used to take the biopsy. The biopsy is then subjected to standard histopathological evaluation.

2.5.

Data Analysis

There were two blind steps in the procedure to avoid bias. First, as mentioned above, the clinician has not seen any fluorescence image before the biopsy is taken. Second, the results of the histopathologic evaluation of the biopsies were not available to us at the time of data processing. Defining the location of the biopsy site in the DR image was not simple since it involved matching two different images taken with different devices (i.e., the colposcope and the fluorescence imaging device). Without these two blind steps possible bias could have occurred.

3.1.

General

Although the prototype functioned properly in the technical sense, its use in this particular clinical field proved to be very cumbersome. Positioning and focusing the device was especially difficult and time consuming. A typical example of a set of images is given in Figure 1(a). The speculum, visible on the right side of the white light image, often caused severe artifacts in the fluorescence image that were not easy to avoid. These artifacts were related to reflections of excitation and fluorescence light on the stainless steel surface that were not completely blocked by detection filters. In addition, the wall of the vagina (that part not covered by the speculum) was out of focus and usually fluoresced intensely. In the fluorescence image this area is painted black, so only the fluorescence on the cervix remains visible.

Figure 1

White light image and DR fluorescence images for two cases in which clinically no CIN was detected and consequently no biopsy was taken. (a) Normal cervix, fluorescence intensity is low, although some speckles are present. (b) Inflamed cervix. The DR values are close to 1, indicating relatively low PpIX levels.

DR fluorescence images were taken of 16 patients, on whom a total of 19 biopsies were obtained and resulted in various different stages of CIN (Table 1).

Table 1

3.2.

DR Fluorescence Images of No CIN

Figure 1(a) shows the white light image and the corresponding DR image of the cervix of a patient who was colposcopically diagnosed as having no CIN. Although some speckles are visible, DR values close to 1 are found throughout the cervix, suggesting very minimal PpIX production. A similar situation is depicted in Figure 1(b), where the main difference is that this cervix is severely inflamed. Nevertheless, the DR image shows a uniform low value, suggesting low PpIX levels throughout the cervix.

3.3.

Localization of CIN With DR Fluorescence Imaging

In four cases a region with clear boundaries could be recognized in the DR image. In all four cases this region matched the location where the biopsy was taken. The four lesions that showed localized fluorescence were all classified as CIN III. An example of this is shown in Figure 2(a). In the other 15 cases larger areas with unclear boundaries showed elevated DR values. An example of a more diffuse fluorescence image is given in Figure 2(b). In these cases the location with the highest DR never coincided with the location where a biopsy was taken.

Figure 2

White light images and corresponding fluorescence images of two cases histopathologically confirmed as CIN III. The locations where the biopsies were taken are indicated by a square, and the ellipse indicates the position of the portio in the fluorescence image. In (a) a clearly localized spot is visible in the fluorescence image and it corresponds to the location where the biopsy was taken. In (b) a more diffuse increase in the DR is observed.

3.4.

Staging of CIN With DR Fluorescence Imaging

Figure 3 shows the DR value averaged over a 16×16 pixel square located at the location at which the biopsy was taken versus the histopathological stage. Although some outliers are clearly present, there appears to be a significant correlation between the DR value and the histopathology [Spearman rank correlation, r=0.881, p<0.001 (confidence interval 0.7044–0.9552)].

Figure 3

DR value calculated at the location where the biopsy was taken vs the histopathological staging performed on the biopsy. Data from 19 biopsies.

4.1.

Localization

Fluorescence images were obtained of a total of 16 cervices. In 13 of these we have histopathological confirmation of intraepithelial neoplasia. In 4 (of these 13) a clearly localized lesion was visible in the fluorescence image at a location coinciding with the (previously determined) location of the biopsy. The biopsies from these sites were all classified as CIN III. In the other nine cervices the fluorescence images showed multiple hotspots of varying intensity or more diffuse elevation of the fluorescence DR. This suggests a more multifocal nature of the disease in these cases.

4.2.

Staging

The strong correlation between the DR value and the histopathological classification shown in Figure 3 is quite striking. This might suggest that there is a relation between the amount of PpIX that accumulates and the stage of the disease. However, we believe that a different mechanism is behind this phenomenon. In the derivation of Eq. (1) a semi-infinite homogeneous medium with a homogeneous PpIX distribution is assumed. The clinical situation is obviously a lot more complex. In the case of topically applied ALA on a mucosal surface a PpIX containing layer may have a thickness of up to a few hundred micrometers of the order of, or less than, the penetration depth of the violet excitation light. Furthermore, an epithelial tumor originates from the boundary between the stromal and epithelial layers and then progresses towards the top surface, into the monitored tissue volume. This will increase the average porphyrin concentration in this volume. As a consequence, we must seriously doubt the validity of the assumption of a semi-infinite homogeneous distribution of PpIX.

To gain more insight into the dependence of the DR on the layered clinical situation, we performed Monte Carlo simulations for which we defined a two-layered structure that represented the epithelial layer and an underlying stromal layer. Both layers were assumed to autofluoresce, and only the epithelial layer was assumed to produce PpIX fluorescence. Fluorescence emerging from the surface was calculated for both excitation and detection wavelengths and a DR value was calculated. By performing different simulations with various layer thicknesses we could assess the dependence of the DR on the layer thickness. The input parameters for the calculations were obtained from the literature and from personal measurements. The technique and the methods are described more in detail in the Appendix.

The results of the simulations are depicted in Figure 4. In comparing these results to Figure 3 it is highly likely to link the stage of the disease to the thickness of the tumor. However, the representation of tumor progression by a perfectly smooth walled cube is far from a realistic situation. Moreover, the optical properties of cervical tissue are not known and the values used in the Monte Carlo simulation were, at best, a guess. However, the values and the geometrical assumption used for the Monte Carlo simulation were used to investigate a concept that may give insight into the correlation between the DR and the stage of the disease. The validity of this concept does not depend on the exact values of the input parameters as long as the relative relations are reasonable.

Figure 4

Interrelationships of the DR value, CIN grade, and layer thickness predicted by the Monte Carlo simulations.

4.3.

Methodology

Tracing the location where the biopsy was taken on the fluorescence image was done by hand. Two different imaging devices were used and a topographical match between the two images was made based on visual clues and landmarks, like the portio and the edge of the cervix. Although a “double blind” approach was used in this procedure, it remains the “Achilles heel” of the study. This problem could very simply be avoided by using an integrated device thereby performing both photographic documentation and DR imaging through the same optics. Technically this should not be a major problem, and the preliminary results obtained in this study certainly warrant such a development.