«Novel Biophotonic Imaging Techniques for Assessing Women’s Reproductive Health by Tyler Kaine Drake Department of Biomedical Engineering Duke ...»
The ROC analysis for nuclear diameter in the basal layer showed that the optimum decision line for discrimination existed at 10.46 µ;m, and this is represented by the dashed line in the Figure 6.6.
By using this line, the dysplastic biopsies can be classified retrospectively with a sensitivity of 100% (3/3), a specificity of 90% (18/20), and an overall accuracy of (21/23) 91.3%.
If just normal and dysplastic data are considered, a sensitivity of 100% (3/3), specificity of 94% (15/16), and accuracy of 95% is seen.
The dataset has a negative predictive value (NPV) of 100% (18/18), which means that all subjects with a negative 106 test results, or dysplastic sites, were correctly diagnosed.
Positive predictive value (PPV), or the portion of positive results that are true positives, was found to be 60% (3/5).
10.50 9.50 8.50 7.50
Scatter plot showing nuclear size versus nuclear density for the three tissue types at the basal layer.
Each point represents a single biopsy and the dashed line is at 10.46 µ;m, which is the optimal decision line for this pilot study.
6.7 Discussion In this a/LCI pilot study of cervical epithelium, the nuclear diameter in the basal layer depth segment (~200 µ;m) provided diagnostic value in identifying CIN with high sensitivity (100%) and specificity (90%).
This result is consistent with previous a/LCI studies of both animal and human epithelial tissue.34- 37, 39, 40, 84 A statistically significant increase in nuclear density at 100 µ;m depth was also found when comparing dysplastic tissue to metaplastic tissue.
However, the diagnostic utility of this metric is not clear, as 107 no such relationship existed between normal and dysplastic tissue, and it had a lower level of statistical significance than nuclear diameter at the basal layer.
Therefore, in future studies, nuclear diameter at the basal layer should be used as the discriminating factor in assessing CIN.
a/LCI was successful in classifying dysplastic cervical tissues with a high NPV in this study.
A high NPV is useful in guiding biopsies because it will identify specific areas of interest, yet ignore healthy tissue.
This study had a NPV of 100%, meaning no dysplastic tissues were classified as healthy, i.e., no precancerous lesions were deemed normal tissue (false negative).
The PPV (60%) was modestly lower than NPV, meaning that areas of normal tissue were diagnosed as dysplastic (false positive).
When guiding biopsies, however, a modest PPV is preferable to a modest NPV.
It is better to take an extra biopsy of normal tissue than to miss an area of dysplasia entirely.
It was also discovered that a/LCI measurements are useful in determining epithelial tissue types in cervical measurements.
A- scans of ectocervix and endocervix obtained by a/LCI have distinguishing features, as shown in Figure 6.7.
A- scans of ectocervical tissue have a falloff in signal after the basal layer, or at about 200- 250 µ;m in depth.
However, endocervical tissue has a sharp rise at the tissue surface, and then a slow decline in signal out to depths of 800- 1000 µ;m.
Figure 6.8 shows a representative histological section of the cervix, with endocervical tissue on the left, the transformation zone (T- zone) in the middle and endocervical tissue on the right of the image.
a/LCI A- scans of cervical tissue types.
Ectocervical epithelium is shown in (a) and endocervical epithelium is shown in (b).
Histology of the cervix.
The ectocervix is made up of nonkeratinized- stratified squamous epithelium (left) and the endocervix is comprised of simple columnar epithelium (right).
The transformation zone is shown in the middle.
The composite graphic created from images taken from
Histology of the Female Reproductive Tract.85 109 It is hypothesized that the nonkeratizined- stratified squamous epithelium has more scattering in its layered epithelium and dense basal layer than ectocervical tissue.
A scans such as Figure 6.7(b), from the ectocervix, have a small region of scattering from the columnar epithelium layer, and then a slow falloff in the stroma.
Differentiating tissue types is important in cervical imaging because many CIN lesions and cancers start in the transformation zone.
The a/LCI device offers insight to a method of identifying the transformation region in real- time for screening of CIN, as its Fourier- domain LCI A- scans can be used for tissue classification.
The mLCI technique, described in Chapter 3, would be ideal for this application, as its multiplexing would allow increased tissue coverage for scanning, as compared to the single point scan of the a/LCI device.
The two could be used in tandem, where mLCI imaging would localize the transformation zone on the cervix, and then a/LCI would be used to screen for CIN lesions.
Early detection of CIN offers the greatest opportunity for therapeutic intervention.
Typically, a woman is initially screened with a Papanicloaou (Pap) test in her first visit.
However, Pap tests are limited in both sensitivity (68- 80%) and specificity (75- 90%).33 If a Pap test is positive, a 2nd visit is needed for a colposcopy and/or biopsy, scheduled weeks later.
Sensitivity and specificity of the colposcopy examination have a wide range of values, 50- 96% and 34- 86%, respectively.30- 32 Finally, if a physical biopsy is taken during colposcopy, and it is positive, a 3rd visit is scheduled for treatment, again 110 weeks later.
Treatment usually entails either a cervical conization or hysterectomy.
This entire process is subjective, time consuming, and labor intensive.
Sensitivities and specificities of the tests suffer from inadequate tissue sampling and human fatigue when interpreting results.
Interestingly, a/LCI could benefit the CIN screening and treatment procedure at a variety of stages.
It could be used as a primary screening tool as a replacement of the Pap test for real- time diagnosis and surveillance, eliminating the need for a follow up visit.
a/LCI could also be used to guide a physical biopsy during colposcopy by providing feedback of tissue morphology to the physician to select the tissue region most likely to harbor dysplasia.
Finally, dysplastic tissues and margins could be identified by a/LCI during the treatment procedure.
Possible roles of a/LCI in the CIN screening and treatment process.
6.8 Summary In this pilot study, a/LCI was able to identify dysplasia with a high sensitivity and specificity when compared to pathological diagnosis.
Nuclear morphology measurements from the basal layer (200- 250 µ;m) of epithelium show a statistically significant increase in nuclear diameter of dysplastic tissue compared to normal tissue.
The optimal decision line of 10.46 µ;m nuclear diameter is consistent with studies of human epithelial tissue by Terry et al.
and Pythilia et al.36, 37 Future in vivo studies will be conducted to further investigate a/LCI as a novel optical cervical screening platform.
An instrument description, as used in the trial, is first presented in Section 6.2.
Relevant data processing for measuring nuclear morphology with a/LCI is then given in Section 6.3.
Section 6.4 describes the study design.
The trial scanned 23 tissues, and a/LCI was able to detect CIN with high sensitivity (100%) and specificity (90%).
This pilot study validated the ability of a/LCI in detecting cervical dysplasia using ex vivo tissue.
The average nuclear diameter in the basal layer of ectocervical epithelium showed a statistically significant increase in size in dysplastic tissue, which is consistent with a/LCI results from other tissue types.36, 37, 82 Also, Section 6.7 discussed the ability of a/LCI depth scans in identifying ectocervical and endocervical tissue, based off of A- scan features that are related to the epithelial structure of the different tissue types.
Several possible roles of a/LCI in the 112 CIN screening and treatment process are also discussed.
Future in vivo studies will be conducted to further investigate a/LCI as a novel optical cervical screening platform.
113 7 Conclusions and future directions The work presented in this dissertation chronicled the development and application of two optical imaging techniques for women’s reproductive health applications.
A dual- modality optical imaging instrument which features simultaneous measurements from mLCI and fluorimetry was described from the first benchtop sytem through the completion of a clinically capable device.
Also, an a/LCI device was applied in detecting cervical dysplasia to show the feasibility of using a/LCI in this novel application.
The evolution of the dual- modality instrument was first described in Chapter 3 which focused on the instrumentation of the system.
The expansion of a single channel LCI device into a multiplexed, Fourier- domain LCI instrument was detailed, including the evaluation of its optical performance and initial validation measurements of gel layers on tissue phantoms.
Once the mLCI instrument was characterized, steps were taken in order to combine mLCI and fluorimetry into a single endoscopic instrument, as detailed in Section 3.2.2.
Again, the optical performance of this new, dual- modality optical imaging instrument was characterized, and initial in vivo human data were obtained to show the feasibility of simultaneous optical imaging from two modalities in a single probe.
A clinical study was then performed with the dual- modality instrument in Chapter 4 to assess the ability of mLCI in measuring microbicide gel thickness 114 distributions as compared to fluorimetry.
This study concluded that mLCI captured statistically similar fraction of vaginal surface with coating measurements as fluorimetry, without the need for exogenous contrast agents.
A second study was completed with the dual- modality device, as described in Chapter 5, in which the difference in the instrument’s modalities was exploited to measure dilution of a placebo microbicide gel.
It was found that as a gel becomes more dilute, its mLCI signal remains constant as its fluorimetry measurement decreases.
A calibration curve was created which allowed the slope- ratio between the measurements to be used to calculate the extent of gel dilution.
An example calculation using the methodology from in vivo data was presented in order to illustrate the complete process of measuring dilution with the dual- modality optical imaging instrument.
Chapter6 then confronted another important issue in women’s reproductive health, cervical cancer.
A clinical a/LCI instrument was applied in an ex vivo pilot study of cervical tissue in order to determine the feasibility of a/LCI in identifying cervical dysplasia.
The study found that a/LCI successfully identified dysplasic tissues with a sensitivity of 100% and specificity of 90% when compared retroactively with traditional histopathological assessment.
The study also found that the depth- resolving LCI aspect of a/LCI allowed for identification of ectocervical and endocervical tissues.
While these studies are extremely encouraging, the dual- modality optical imaging instrument will be used again in a study on optical imaging and user 115 perception of vaginal gels.
The disparities in clinical trial outcomes of a Tenofovir- based gel (as discussed in Section 1.1) between the CAPRISA- 004 and VOICE trials indicate that effectiveness of microbicide gel use is related to the dose regimen and user- experience of the gels.12- 14 If the dose regimen and/or product characteristics are not acceptable to women, then adherence suffers, and even highly potent APIs will be ineffective.
One such determinate of adherence is inserted gel volume.
Gel volume also plays a crucial role in the pharmacokinetics of candidate products.
A successful microbicide will be one that is biologically effective and acceptable by women’s perceptions.
Therefore, a new study will be completed to study the linkage between prophylactic potential of the gel as a delivery vehicle for an API, and user perception of the gel.
The study will focus on the safe application of the dual- modality device and attempt to determine whether there are differences in vaginal distributions and user sensory perceptions between two gel volumes.
The study will also be used to explore whether bipophysical computations of vaginal gel spreading correlate with image- based measures of coating, and to again assess if there are any local differences in measured gel coating between fluorimetry and mLCI.
These local differences could suggest gel dilution, so if there are any found, the dilution method presented in Chapter 5 could be used to help interpret results.
116 More studies are also needed to verify the ability of a/LCI in detecting cervical dysplasia.
A higher powered study, i.e.
more co- located biopsy points, is needed to determine if a/LCI can classify dysplasia from metaplasia in the basal layer of cells.
The pilot study had a p- value of 0.058 in which statistical significance was not achieved.
It would be advantageous to demonstrate the ability of a/LCI in separating the tissue types because metaplastic tissue is common at the transformation zone of the cervix, and does not necessarily indicate a pre- cancerous change in tissue structure.
The transformation zone undergoes metaplasia throughout various times in a women’s life naturally such as puberty, during the menstrual cycle, and at post- menopause.
Since a/LCI depth- scans were able to resolve endocervical and ectocervical tissues, an imaging platform that combines a/LCI and mLCI could be created to exploit advantages of both modalitites.
mLCI could be used to efficiently map the cervical epithelium since it allows simultaneous measurements with broad tissue coverage.