Most individuals at risk for cervical cancer reside in the developing world [4], which in the majority of cases, lack the resources, infrastructure and/or access to trained pathologists for effective cervical cancer screening. An affordable and easy-to-use diagnostic device could have a significant impact on cervical cancer screening in this setting, especially in remote areas of developing countries [5, 6].
The Foldscope has shown its ability to image parasites such as Schistosoma haematobium [14], as well as cervical squamous epithelial cells [10]. This latter study reported a 100% correlation between cytologic characteristics found by the Foldscope and conventional optical microscopy, with a global Cohen kappa index of 0.7, and values of 0.8 for chromatic value, 0.8 for nuclear membrane continuity, and 0.7 for cytoplasmic morphology. The referred study, however, utilized an image station, projecting the image from a Foldscope with an additional condensation system into a dark room [10] thus requiring additional technology and resources that limit its potential application in remote and/or rural settings.
Overall, the Foldscope images were comparable to those obtained with conventional microscope and digital camera setup. For most of the cytology slides (75%), an exact match was observed. In these exact matches, cytologic changes were more clear and classic, compared to discordant cases. The mismatches observed between Normal and LSIL could be attributed to problems with focusing (due to dense cell clusters), limited ROI, as well as subjectivity. Image clarity was the likely reason for the single HSIL/Malignant to Normal discordance observed. Normal cells clumped together with reduced clarity and inspection of a single ROI may have also led to the misclassification.
The limitations of a ROI and subjectivity with the cytologic features may account for the discordances we observed in our study. Most of the mismatches occurred with HSIL/Mal (6 out of 10), and LSIL (5 out of 8). This group of mismatches is the most concerning because of the significant difference in treatment regimens for these two categories. Nevertheless, the sensitivity and specificity for LSIL and HSIL were over 80%. Generally, pathologists screen the entire slide, along with examination of multiple ROIs prior to rendering a diagnosis, which helps in accuracy of classification. The differentiation between LSIL and HSIL may have been difficult with the Foldscope due to the limitation of a single or a few ROIs, and lack of whole slide screening.
In addition, the transition from LSIL to HSIL can itself be subjective in certain groups of cells. The visual evaluation of N:C ratio is inherently subjective and not a quantitative value. Subjectivity and poor reproducibility are not uncommon within pathological classification schemes [15]. In certain classification schemes some categories might not have significant change in the initial diagnostic work-up. This is the case when differentiating between HSIL and malignant/cancer, in which for both categories, the initial diagnostic plan calls for an immediate colposcopy and biopsy. For these reasons and the limited ROI in our project we elected to combine the HSIL and Malignant cytology categories into a single HSIL/Malignant category.
The small visual field, limitations in focusing, along with the inability to make fine movements of the slide under the Foldscope, were the main drawbacks of the Foldscope. The current paper stage does not provide sufficient focusing or slide scanning opportunities, making it a very strenuous task to completely scan/screen a cytology slide. Focusing problems were apparent in the Foldscope as evidenced by large areas with variation in sharpness and clarity even within a single cluster of cells. Lack of sharpness around the focal point and focus constrictions have been previously reported as restrictions of the device [10].
The ball lens microscope is a powerful innovation which may potentially allow for affordable diagnosis of not only cervical cytology but other cytopathological conditions especially with the use of specific Foldscope designs [16]. The lens of the Foldscope is powerful and its ability for classification can be further exploited through computerized analysis. Phone applications with morphometry may allow for rapid classification of cytology and reduce subjective classifications. The development of a staging system as well as improvement in the current focusing system may reduce the scanning obstacles.