A 3-D optical imaging system based on active stereo vision and motion tracking is built to track the motion of patient and to register the time-sequenced images of cervix recorded during colposcopy. The imaging system is evaluated by examining human subjects in vivo before routine colposcopy examination procedures. The system tracks the motion of patient accurately. The temporal kinetics of the acetowhitening process in the area of interest can be quantitatively measured. The results demonstrate that the kinetics of acetowhitening may be potentially used for accurately differentiating the precancerous lesions from the normal and benign lesions, and grading the precancerous lesions.
© 2008 Optical Society of America
Colposcopy is a standard technique to detect early neoplastic growth in cervical tissue. A fundamental part of the colposcopy is the use of acetic acid, to induce acetowhitening effect in tissues and to produce the contrast between normal tissues and precancerous lesions for diagnosis . The diagnostic criteria of colposcopy based on acetowhitening are subjectively related to the following parameters: the rapidity and length of the acetowhitening processes, the degree of acetowhiteness when the change of color reaches maximum, and the sharpness of the demarcation line between the precancerous lesions and normal tissues. The diagnostic accuracy depends on how an individual colposcopist applies these guidelines during the diagnosis. Therefore, an objective diagnostic procedure based on quantitative measurement of acetowhitening process is desirable. It was found that the dynamic process of acetowhitening can assist in discriminating the normal and abnormal cervical tissue, even for distinguishing the different grades of cervical intraepithelial neoplasia (CIN) [2,3]. However, as the patients do not remain completely stationary and the camera may be moved by the colposcopist during the examination, the pixel coordinates of the area of interest in time-sequenced images will change. To quantify the kinetics of acetowhitening in an area of interest, it is crucial to track the movement of patient or camera and register the time-sequenced images during colposcopy procedure accurately. Recently, we developed a combined 3-D imaging and motion tracking technique for the quantitative measurement of the kinetics of acetowhitening . The 3-D surfaces of a cervix model measured at different times are used to track the motion and to register the time-sequenced 2-D images. We demonstrated that this imaging system holds the potential to enable quantitative mapping of acetowhitening kinetics over the cervical surface. In this study, we built a 3-D colposcopic imaging system based on the same principles. The goals of this study are twofold: (i) to evaluate the accuracy of tracking patient’s motion and co-registration of the time-sequenced 2-D images of cervix recorded during colposcopy in vivo, (ii) to demonstrate that the temporal kinetics of acetowhitening in the imaged tissue can be quantitatively mapped based on the co-registered time-sequences images of cervix.
2. Methods and materials
2.1 3-D colposcopic imaging system based on active stereo vision
We built a compact imaging system to measure the 3-D topology of cervix surface based on active stereo vision. The system consists of an imaging channel, an illumination channel and a projection channel, as shown in Fig. 1. The iterative closest point (ICP) algorithm was used to track the motion of the object by using the information of 3-D surfaces. Time-sequenced images of cervix recorded during the examination were registered by using the information of 3-D motion tracking. The details of the technology were described in previous report .
The imaging channel was designed with a similar configuration as that of a standard digital colposcope. A three-CCD color camera (Model No. DXC-390P, Sony Co., Japan) and a fast frame grabber (Genesis Plus, Matrox Inc., Canada) was used to capture images with the resolution of 768×576 pixels in the imaging channel. The illumination channel consisted of a ring illuminator and a linear polarizer. The ring illuminator with 30 blue-light LEDs produced uniform illumination on the object surface. The polarization direction of the linear polarizer was set to be perpendicular to the polarizer in front of the camera. The cross-polarization arrangement eliminated the specular reflection from the surface of the examined object. In the projection channel, a structured light with 33 laser stripes was generated by a laser diode at 660 nm and a holographic grating (Lasiris SNF and SNF-533L-660S-100-5-SD, StockerYale Inc., Canada). To make sure that most of the 33 laser stripes can be projected through the speculum used in colposcopy, the angle between the imaging channel and projection channel was set to about 5°. The actual number of the projected laser stripes on cervix was dependent on the operation of the colposcopist and the position of the patient.
Because the patient’s motion is continuous, the images of cervix and the projected laser stripe pattern on cervix must be taken simultaneously. We used blue and red channels of the three-CCD color camera to capture the images of cross-polarized reflection and the laser stripe pattern, respectively. Typical image of a cervix surface illuminated by the blue-light ring illuminator and the image of laser stripes on cervix are shown in Fig. 2(a&b). The field-of-view of the image is about 37 mm by 28 mm. The resolution of the imaging system is about 0.05mm/pixel. Though the cross-talk between red and blue channels of the camera was not significant, small number of saturation pixels caused by the specular reflection of red laser stripes indeed appeared in the image of cross-polarized reflection. The cross-polarization arrangement was not used in the structured light imaging because it blurs the image of laser stripes and causes errors in the measurement of 3-D topology of cervix surface. The color image from the same subject imaged by a commercial colposcope is shown in Fig. 2(c) as a reference.
2.2 3-D surface measurements and 2-D image registration
The projection and imaging channels of imaging system were calibrated before the measurement of 3-D surface of cervix. The detailed calibration procedures have been reported in [4–6]. The 3-D topology of imaged cervical surface was calculated based on the triangulation of a ray from the imaging channel and a plane from the projection channel . A reconstructed 3-D cervix surface is shown in Fig. 2(d). After the 3-D information of the cervix surface was known, the motion of surface at different locations could be tracked by using ICP algorithm . The registration of 2-D images taken during the examination was achieved by using the 3-D transformation provided by the motion tracking .
Ten human subjects originally scheduled for the examination with colposcopy and loop electrosurgical excision procedure (LEEP) were enrolled in this in vivo study. To avoid interference from the mucus discharge, cervix was first cleaned with PBS before applying 5% acetic acid. The images of cervix and structured light were recorded over 5 minutes after the application of acetic acid. The images in the blue and red channels were stored for the analysis of acetowhitening kinetics. A routine colposcopy was followed up after 3-D imaging. Colposcopic diagnosis of examined tissue sites was provided by two colposcopists. This study was approved by the Ethical Committee of the Prince of Wales Hospital, the Chinese University of Hong Kong.
3. Results and discussions
The 3-D colposcopic imaging system was evaluated in vivo. Representative results of motion tracking and image registration are shown in Fig. 3 and Fig. 4. The movie in Fig. 3(a) is the superposition of unregistered target images and a reference image. The target images are the time-sequenced images of cervix taken over certain period of time. The reference image is the first image. To visualize the registration accuracy of the 3-D imaging technique, the target images and reference image are displayed in green and red color, respectively. It should be noticed that all the time-sequenced reflectance images were recorded from the blue channel of the color CCD camera. In this case, the maximal movement of the human cervix was about 10mm. Similarly, the movie shown in Fig. 3(b) is the superposition of registered target images and the reference image to show the accuracy of image registration.
In general, there are no reliable landmarks on cervix surface as the cervix model used in previous study for quantitative measurement of the accuracy of motion tracking and image registration . In this work we compare the intensity profiles along a few representative lines measured from target images and reference image. The comparison provides a qualitative evaluation on how accurately the target images are registered to the reference image. As shown in Fig. 3(c), the profile measured from unregistered target image along line A-B (green line) and the line profile measured from the reference image (red line) are not correlated because of large motion of cervix. However, the profiles measured along line A′-B′ from registered and reference images shown in Fig. 3(d) highly overlap with each other, indicating the target images are accurately registered to the reference image. The mean distance between the corresponding profile peaks is 1.16 ± 0.92 pixels after image registration. It should be emphasized that the effectively registered area in the target images are that sampled by the structured light in the reference image. The image out of the sampling area can not be accurately registered due to lacking of surface information. The movie shown in Fig. 4 presents the results of tracking and registration of relatively small movements of patient. Similarly, the mean distance between the corresponding profile peaks is 1.36 ± 1.06 pixels after image registration. Again, it was demonstrated that the 3-D colposcopic imaging technique can accurately track the patient’s motion and register the 2-D images recorded during the examination.
The representative results of the measurement of temporal kinetics of acetowhitening are shown in Fig. 5. The movies in Fig. 5 (a&b) are formed in the same way as Fig. 3&4. The target images were taken over 5 minutes after the application of acetic acid to the cervix. The reference image was the image taken at 95 second after the application of acetic acid because the sampling area on cervix was maximal. We measured the changes of acetowhitening as a function of time at 5 different sites of cervix from the unregistered and registered target images, respectively. The sampling area at each measurement site was 6×6 pixels in image, equivalent to 0.3×0.3 mm on cervix. It is sufficiently small compared with the CIN lesion of normal size of a few mm. The results are shown in Fig. 5(c&d). As can be seen, the reflection intensity as a function of time shown in Fig. 5(c) do no provide information on kinetics of acetowhitening at all because of patient’s movement, while the curves displayed in Fig. 5(d) demonstrate that reflection signal as a function of time follows a smooth pattern. It should be noted that unlike the results of the measurements without the application of acetic acid shown in Fig. 3&4, the color in certain areas of the superposition images in Fig. 5(b) changes over the measurement time period because the reflection intensity of target images are a function of time due to the temporal kinetics of acetowhitening.
Normally, the grade of CIN lesion varies over the entire cervix surface. The results shown in Fig. 5(d) demonstrate multiple kinetics of acetowhitening measured at different tissue sites. The diagnoses of the tissues at 5 different measurement sites labeled in Fig. 5(b) by two colposcopists showed that the site A, B are CIN I (low grade lesion), site C, D are CIN II/III (high grade lesion) and site E is normal.
To investigate possible correlation between the acetowhitening kinetics and tissue pathology, we measured the kinetics of acetowhitening from registered time-sequenced images recorded from ten patients and related the kinetics to the diagnoses made by two colposcopists. For each patient the kinetics of acetowhitening were measured from 4–5 sites that were diagnosed as four different types of tissues: normal tissue, human papillomavirus (HPV) infected tissue, low grade lesion (CIN I) and high grade lesion (CIN II/III), respectively. Based on the measurements from the ten patients, the kinetics of acetowhitening associated with four types of tissues were calculated and the statistical results are displayed in Fig. 6. As can be seen, the kinetics of acetowhienting measured from different types of tissues are distinct. The reflection intensity from normal tissue remains almost unchanged over the measurement period time of 5 minutes. The signal from HPV infected tissue increases to the maximum in 60 to 100 second and stays at the level over the measurement period time. The signal from low grade lesion increases quickly after the application of acetic acid and reaches to maximum in about one minute. The signal then decreases quickly after reaching to the peak. Similar to the signal from low grade lesion, the signal from high grade lesion increases quickly and reach to peak in about one minute after the application of acetic acid. However, the decay of the acetowhitening in high grade lesion is much slower than low grade lesion. Statistically, significant differences between the curves of four different kinetics are found. To quantify the similarity between four different kinetics of acetowhitening, the correlation coefficient (r) between each two of them are calculated . The ∣r∣ between normal and HPV infected tissue is 0.54. The ∣r∣ between normal tissue and low/high grade lesions are 0.65 and 0.03, respectively. The ∣r∣ between HPV infected tissue and low/high grade lesions are 0.20 and 0.38, respectively. The ∣r∣ between low grade lesions and high grade lesions is 0.70. Normally, the absolute value of a correlation coefficient over 0.8 implies a strong linear relationship . The lower r-values suggest that the kinetics of acetowhitening provide clear differentiation between four types of tissues.
We investigated a quantitative imaging method for the diagnosis of cervical precancerous lesions. The imaging system is equipped with the capabilities of 3-D vision and motion tracking. The results demonstrate that the imaging technique can accurately track the patient’s motion and register the 2-D images of cervix recorded during colposcopy in vivo. The temporal kinetics of acetowhitening in the imaged tissue can be quantitatively mapped based on the registered time-sequences images of cervix. The distinct differences in acetowhitening kinetics between normal tissue, HPV infection and low/high grade lesions provide solid basis for quantitative diagnosis and grading of cervical precancerous lesions. The quantitative imaging method holds the potential to be used in conjunction with the traditional colposcopy and to improve the diagnostic accuracy of colposcopy. In addition, the system shown in Fig. 1 can be used for the standard colposcopy and the mapping of acetowhitening kinetics over the imaged tissue surface simultaneously. Therefore, the diagnostic instrument and methods could be inserted into the physician’s workflow with minimal interference to routine colposcopy procedure. The measurement of acetowhitening kinetics requires about five minutes, which means that the examination procedure may be longer than conventional colposcopy. The major limitation of current system is that the data processing is time-consuming . We will improve the image requisition and processing techniques to achieve the real-time measurements. Finally, it should be pointed out that though the temporal kinetics of acetowhitening is highly correlated with the colposcopists’ diagnosis, the final conclusion on the correlation between the kinetics and tissue pathology must be based on the histological analysis of examined tissue. In the future work, a large scale of clinical study will be conducted to investigate the correlation between the kinetics of acetowhitening and the pathology of examined tissue.
The authors gratefully acknowledge financial support by the Hong Kong Research Grants Council through the grant HKUST6408/05M.
References and links
1. M. C. Anderson, J. A. Jordan, A. R. Morse, and F. Sharp, A Text and Atlas of Integrated Colposcopy: for Colposcopists, Histopathologists and Cytologists (Chapman & Hall Medical, London, 1992).
2. B. W. Pogue, H. B. Kaufman, A. Zelenchuk, W. Harper, G. C. Burke, E. E. Burke, and D. M. Harper, “Analysis of acetic acid-induced whitening of high-grade squamous intraepithelial lesions,” J. Biomed. Opt. 6, 397–403 (2001). [CrossRef] [PubMed]
3. T. T. Wu, J. Y. Qu, T. H. Cheung, S. F. Yim, and Y. F. Wong, “Study of dynamic process of acetic acid induced-whitening in epithelial tissues at cellular level,” Opt. Express 13, 4963–4973 (2005). [CrossRef] [PubMed]
5. R. Y. Tsai, “A versatile camera calibration technique for high-accuracy 3D machine vision metrology using off-the-shelf TV camera and lenses,“ IEEE J. Rob. Autom. 3, 323–344 (1987). [CrossRef]
6. D. Q. Huynh, R. A. Owens, and P. E. Hartmann, “Calibrating a structured light stripe system: a novel approach,” Int. J. Comput. Vis. 33, 73–86 (1999). [CrossRef]
7. P. J. Besl and N. D. McKay, “A method for registration of 3-D shapes,” IEEE Trans. on Patt. Anal. Mach. Intell. 14, 239–256 (1992). [CrossRef]
8. D. S. Moore and G. P. McCabe, Introduction to the Practice of Statistics, 5th ed. (W. H. Freeman, New York, 2006).