Diagnosing Cervical Neoplasia in Rural Brazil Using a Mobile Van Equipped withIn VivoMicroscopy: A Cluster-Randomized Community Trial

Abstract

Cervical cancer is a leading cause of death in underserved areas of Brazil. This prospective randomized trial involved 200 women in southern/central Brazil with abnormal Papanicolaou tests. Participants were randomized by geographic cluster and referred for diagnostic evaluation either at a mobile van upon its scheduled visit to their local community, or at a central hospital. Participants in both arms underwent colposcopy, microscopy, and cervical biopsies. We compared rates of diagnostic follow-up completion between study arms, and also evaluated the diagnostic performance of microscopy compared with colposcopy. There was a 23% absolute and 37% relative increase in diagnostic follow-up completion rates for patients referred to the mobile van (102/117, 87%) compared with the central hospital (53/83, 64%; = 0.0001; risk ratio = 1.37, 95% CI, 1.14-1.63). In 229 cervical sites in 144 patients, colposcopic examination identified sites diagnosed as cervical intraepithelial neoplasia grade 2 or more severe (CIN2+; 85 sites) with a sensitivity of 94% (95% CI, 87%-98%) and specificity of 50% (95% CI, 42%-58%). microscopy with real-time automated image analysis identified CIN2+ with a sensitivity of 92% (95% CI, 84%-97%) and specificity of 48% (95% CI, 40%-56%). Women referred to the mobile van were more likely to complete their diagnostic follow-up compared with those referred to a central hospital, without compromise in clinical care. microscopy in a mobile van provides automated diagnostic imaging with sensitivity and specificity similar to colposcopy..

A comparison of screening tests for detection of high-grade cervical abnormalities in women living with HIV from Cameroon

“…Promising new technologies such as Table 2 Distribution of results for provider-collected (Provider) or self-collected (Self) specimen tested for HPV or visual inspection after acetic acid (VIA) by severity of the cytologic interpretation for women living with HIV. Abbreviations: N+, number of positives; %+, percent positive automated visual evaluation [27] and in vivo microscopy [28] may be better at distinguishing between clinically important and benign HPV infections, an important consideration for management of WLWH who typically have a very high prevalence of HPV [23,[29][30][31][32][33][34] that is related to the degree of HIV control [29,32,33,35]. Table 4 Sensitivity and positive predictive value (PPV), with 95% confidence interval (95%CI), of triage strategies for high-grade cervical abnormalities among women living with HIV who tested HPV positive on the provider-collected (Provider) or self-collected (Self) specimen for women living with HIV.…”

Development of Low-Cost Point-of-Care Technologies for Cervical Cancer Prevention Based on a Single-Board Computer

“…Previous versions of the HRME have incorporated image analysis software that relies on image segmentation to quantify changes in nuclear morphometry due to high-grade cervical precancer and early cancer [10], [15], [16]. This is consistent with current clinical guidelines for screening and treatment of cervical cancer and its precursors [32].…”

“…Due to limitations in processing speed and availability of proprietary software frameworks on the Raspberry Pi, deployment of real-time image analysis using previously developed nuclei segmentation algorithms was not practical. As a proof of principle, real-time image analysis was accomplished on the PiHRME by employing a transfer learning approach to retrain a light weight convolutional neural network (CNN) (MobileNetV2) [33] to predict the output of the pre-existing HRME image analysis algorithm being validated in previous and ongoing clinical trials [10], [11].…”

“…Before imaging, LEEP specimens were stained with 0.01% proflavine solution, a topical antiseptic solution that has been used to fluorescently stain cervical nuclei for imaging, and Lugol's iodine, to increase image contrast. Similar staining has been performed previously to acquire high-resolution images of the cervix using HRME imaging systems [10], [11]. All images were analyzed using the CNN image analysis algorithm.…”

“…For example, the high-resolution microendoscope (HRME) is one low-cost technology that has been developed to provide in vivo imaging of the cervix at the point-of-care [12]- [14]. Image segmentation algorithms have been developed to characterize the size and shape of nuclei within the field-ofview, [8], [15], [16] demonstrating diagnostic performance on par with expert colposcopy for detecting high-grade cervical precancer and cancer [10]. However, these algorithms are often implemented on Windows PC systems that rely on proprietary and computationally heavy software frameworks (LabVIEW/MATLAB) and contribute significantly to the overall cost of the device.…”