Multiscale Optical Imaging Fusion for Cervical Precancer Diagnosis: Integrating Widefield Colposcopy and High-Resolution Endomicroscopy.

OBJECTIVE: Early detection and treatment of cervical precancers can prevent disease progression. However, in low-resource communities with a high incidence of cervical cancer, high equipment costs and a shortage of specialists hinder preventative strategies. This manuscript presents a low-cost multiscale in vivo optical imaging system coupled with a computer-aided diagnostic system that could enable accurate, real-time diagnosis of high-grade cervical precancers. METHODS: The system combines portable colposcopy and high-resolution endomicroscopy (HRME) to acquire spatially registered widefield and microscopy videos. A multiscale imaging fusion network (MSFN) was developed to identify cervical intraepithelial neoplasia grade 2 or more severe (CIN 2+). The MSFN automatically identifies and segments the ectocervix and lesions from colposcopy images, extracts nuclear morphology features from HRME videos, and integrates the colposcopy and HRME information. RESULTS: With a threshold value set to achieve sensitivity equal to clinical impression (0.98 [p = 1.0]), the MSFN achieved a significantly higher specificity than clinical impression (0.75 vs. 0.43, p = 0.000006). CONCLUSION: Our findings show that multiscale optical imaging of the cervix allows the highly sensitive and specific detection of high-grade precancers. SIGNIFICANCE: The multiscale imaging system and MSFN could facilitate the accurate, real-time diagnosis of cervical precancers in low-resource settings.

Multi-task network for automated analysis of high-resolution endomicroscopy images to detect cervical precancer and cancer.

Cervical cancer is a public health emergency in low- and middle-income countries where resource limitations hamper standard-of-care prevention strategies. The high-resolution endomicroscope (HRME) is a low-cost, point-of-care device with which care providers can image the nuclear morphology of cervical lesions. Here, we propose a deep learning framework to diagnose cervical intraepithelial neoplasia grade 2 or more severe from HRME images. The proposed multi-task convolutional neural network uses nuclear segmentation to learn a diagnostically relevant representation. Nuclear segmentation was trained via proxy labels to circumvent the need for expensive, manually annotated nuclear masks. A dataset of images from over 1600 patients was used to train, validate, and test our algorithm; data from 20% of patients were reserved for testing. An external evaluation set with images from 508 patients was used to further validate our findings. The proposed method consistently outperformed other state-of-the art architectures achieving a test per patient area under the receiver operating characteristic curve (AUC-ROC) of 0.87. Performance was comparable to expert colposcopy with a test sensitivity and specificity of 0.94 (p = 0.3) and 0.58 (p = 1.0), respectively. Patients with recurrent human papillomavirus (HPV) infections are at a higher risk of developing cervical cancer. Thus, we sought to incorporate HPV DNA test results as a feature to inform prediction. We found that incorporating patient HPV status improved test specificity to 0.71 at a sensitivity of 0.94.

Topical Imiquimod for the Treatment of High-Grade Squamous Intraepithelial Lesions of the Cervix: A Randomized Controlled Trial.

OBJECTIVE: To evaluate the histologic response rate of high-grade squamous intraepithelial lesions (HSIL) of the cervix after topical application of 5% imiquimod cream. METHODS: In this phase II trial, women with cervical HSIL (cervical intraepithelial neoplasia [CIN] 2-3) were randomly assigned to 250 mg of 5% imiquimod cream applied to the cervix weekly for 12 weeks, followed by loop electrosurgical excision procedure (LEEP) without preceding treatment. The sample size was calculated based on the HSIL regression rates previously reported by Grimm et al. The primary outcome was rate of histologic regression (to CIN 1 or less) in LEEP specimens. Prespecified secondary endpoints included surgical margin status and adverse events. Outcomes were stratified by human papillomavirus type and lesion grade (CIN 2 or CIN 3). Results were reported according to per protocol (PP) and intention-to-treat (ITT) analyses. RESULTS: Ninety women were enrolled: 49 in the experimental group and 41 in the control group. In the PP population, histologic regression was observed in 23 of 38 participants (61%) in the experimental group compared with 9 of 40 (23%) in the control group (P=.001). Surgical margins were negative for HSIL in 36 of 38 participants (95%) in the experimental group and 28 of 40 (70%) in the control group (P=.004). In the ITT population, rates of histologic regression also were significantly higher in the experimental group. Rates of adverse events in the experimental group were 74% (28/38) in the PP population and 78% (35/45) in the ITT population. Adverse events were mild, with abdominal pain being the most common. Three patients in the experimental group had grade 2 adverse events, including vaginal ulcer, vaginal pruritus with local edema, and moderate pelvic pain. CONCLUSION: Weekly topical treatment with imiquimod is effective in promoting regression of cervical HSIL. CLINICAL TRIAL REGISTRATION: ClinicalTrials.gov, NCT03233412.

High-resolution microendoscopy: a point-of-care diagnostic for cervical dysplasia in low-resource settings.

Cervical cancer is the third leading cause of cancer-related death among women in low-to-middle income countries. Pap testing and pathological services are difficult to implement under these settings. Alternative techniques for the diagnosis of cervical precancer in these settings are needed to reduce the burden of the disease. The objective of this study was to evaluate the diagnostic accuracy of a low-cost, high-resolution microendoscope imaging system in identifying precancerous lesions of the cervix in vivo. A retrospective study of 59 patients undergoing colposcopy for an abnormal Pap test was performed at Hospital de Câncer de Barretos in Brazil. All patients underwent colposcopy as per standard of care, and acetowhite lesions were recorded. High-resolution microendoscopy (HRME) images were obtained from one colposcopically normal region and from all lesions observed on colposcopy. Biopsies of abnormal areas were obtained and reviewed by three independent, blinded pathologists and compared with HRME findings. The mean nuclear area and the median nuclear eccentricity were calculated from HRME images acquired from each site. A diagnostic algorithm to distinguish histopathologically diagnosed cervical intraepithelial neoplasias of grade 2 or more severe lesions (high grade) from less severe lesions (low grade) was developed using these parameters. A test of trend was used to analyze the relationship between HRME positivity and severity of histopathogical diagnosis. Fisher's exact test was used to analyze differences in HRME positivity between high-grade and low-grade lesions. Evaluable images were obtained from 108 of 143 discrete sites. Of these, 71 sites were colposcopically normal or low grade according to histopathology and 37 were diagnosed as high grade on the basis of histopathology. Using the mean nuclear area and the median nuclear eccentricity, HRME images from 59 colposcopically abnormal sites were classified as high grade or low grade with 92% sensitivity and 77% specificity compared with histopathological findings. Increasing HRME positivity showed a significant trend with increasing severity of diagnosis (Ptrend<0.001). We found a strong association (P<0.001) between HRME positivity and a histopathological diagnosis of cervical intraepithelial neoplasia of grade 2 or higher. HRME demonstrated an accurate in-situ diagnosis of high-grade dysplasia. In low-resource settings in which colposcopy and histopathology services are severely limited or unavailable, HRME may provide a low-cost, accurate method for diagnosis of cervical precancer without the need for biopsy, allowing for a single 'screen-and-treat' approach.

The Accuracy of p16/Ki-67 and HPV Test in the Detection of CIN2/3 in Women Diagnosed with ASC-US or LSIL.

The objective of this study was to compare the accuracies of double staining for p16/Ki-67 and the molecular test for high-risk HPV (hr-HPV) to identify high-grade cervical intraepithelial neoplasia (CIN2/CIN3) in women with cervical cytology of atypical squamous cells of undetermined significance (ASC-US) and low-grade squamous intraepithelial lesion (LSIL). Data were collected from 201 women who underwent cervical cytology screening in the Barretos Cancer Hospital and their results were categorized as ASC-US (n=96) or LSIL (n=105). All patients underwent colposcopy with or without cervical biopsy for diagnosis of CIN2/CIN3. The hr-HPV test (Cobas 4800 test) and immunocytochemistry were performed to detect biomarkers p16/Ki-67 (CINtec PLUS test). Two samples (1 ASC-US/1 LSIL) were excluded from the analysis due to inconclusive results of the histologic examination. There were 8 cases of CIN2/CIN3 among 95 women with ASC-US (8.4%), and 23 cases of CIN2/CIN3 among 104 women with LSIL (22.1%). In the group of women with ASC-US, the sensitivity and specificity in diagnosing CIN2/CIN3 were 87.5% and 79.5% for the HPV test and 62.5% and 93.1% for p16/Ki-67. Among women with LSIL, the sensitivity and specificity in the diagnosis of CIN2/CIN3 were 87% and 34.7% for the HPV test and 69.6% and 75.3% for immunocytochemistry. Superior performance was observed for p16/Ki-67 double staining, especially among women under 30 for whom the test had an area under the ROC curve of 0.762 (p<0.001). Both p16/Ki-67 double staining and the hr-HPV DNA test had similar performance in predicting high-grade cervical intraepithelial neoplasia among women with ASC-US. The best performance was observed in women aged >30 years. In younger women (≤30 years) with LSIL, p16/Ki-67 had greater accuracy in identifying precursor lesions. Among women >30 years diagnosed with LSIL, the two methods showed similar performance.