Innovations in cervical cancer diagnosis for low resource settings using advanced optical imaging and machine learning diagnostic algorithms.

该项目的广泛目标是适应便携式，低成本，易于使用的袖珍阴道镜（开发 在其他资金下）用于社区环境，并开发自动化算法联合收割机 新血管形成、糖原消耗和醋酸增白，以向专家提供可比较的诊断。这项工作 将由第三石设计公司、杜克大学和肯尼亚医学研究所合作完成 院这项建议的具体目标是： 目的1（第一阶段）：通过设计连续放大机制和 改进器械工作流程集成，通过使用 一次性光学透明无菌套管。提供者对我们以前开发的袖珍阴道镜的反馈， 一致建议增加一个滑块机制，以控制粗变焦和套筒消耗品， 袖珍阴道镜设计。 目标2（第一阶段）：宫颈癌前病变检测的自动化算法和软件 VIA使用具有嵌入式机器学习诊断算法的新型软件应用程序的特异性， 子宫颈癌自动检查我们将应用并验证VIA和GIVI（绿色）的各个算法 照明血管成像）与从200例患者临床研究中获得的现有图像进行对比 阴道镜然后，我们将比较算法的性能与专家医生对相同算法的解释。 图像，病理学作为金标准。 目标3（第二阶段）：记录肯尼亚袖珍阴道镜的用户体验。我们将开发一个 直接在社区卫生保健环境中提供文化相关的培训包。我们将收集定量和 定性数据，包括调查，深入访谈，以及来自幼稚提供者和患者的临床观察 并利用这些发现来改善肯尼亚袖珍阴道镜的引进， 同时，告知目标4中的临床研究。 目标4（第II阶段）：比较袖珍阴道镜与目视检查的性能， 醋酸用于肯尼亚HPV+妇女的分诊。我们将在400例HPV+患者中开展一项随机分组试验 在肯尼亚的基苏穆，妇女们将标准分类法与袖珍阴道镜进行比较。所有HPV阳性女性 将进行活检，以确定灵敏度，特异性和阳性和阴性预测值的不同 分类策略数据将被用来模拟该算法的性能对专家阴道镜。 目标5（第二阶段）：评估下列活动的成本、增量成本效益和人口健康影响： 基于HPV的宫颈癌筛查计划与建议的分诊策略。康贝特人将以 增量成本效益比以及四种分诊策略的绝对和相对成本， 成本和人口健康结果模型（癌症病例、死亡和残疾调整生命年）。

The broad goal of this project is to adapt a portable, low-cost, easy-to-use Pocket-sized Colposcope (developed under other funding) for use in a community setting, and develop automated algorithms that combine neovascularization, glycogen depletion and acetowhitening to provide comparable diagnosis to an expert. This work will be done in a collaboration between 3rd Stone Design, Inc, Duke University and Kenya Medical Research Institute. The specific aims of this proposal are: Aim 1 (Phase I): Improve Pocket colposcope by designing continuous magnification mechanism and improving device workflow integration to eliminate between-use disinfection through the use of a disposable optically clear sterile sleeve. Provider feedback on our previously developed Pocket colposcope has unanimously suggested the addition of a slider mechanism to control coarse zoom and a sleeve consumable to the Pocket colposcope design. Aim 2 (Phase I): Automated algorithms and software for cervical pre-cancer detection We will improve the specificity of VIA using a novel software application with embedded machine learning diagnostic algorithms for automated cervical cancer screening. We will apply and validate the individual algorithms for VIA and GIVI (green illumination vascular imaging) to existing images obtained from a 200-patient clinical study with the Pocket colposcope. We will then compare the performance of the algorithms to expert physician interpretation of the same images, with pathology serving as the gold standard. Aim 3 (Phase II): Document user experience with Pocket colposcope in Kenya. We will develop a culturally relevant training package directly in the community healthcare setting. We will collect quantitative and qualitative data including surveys, in-depth interviews, and clinic observations from both naive providers and patients and use these findings to and use these findings to improve the introduction of the Pocket colposcope in Kenya and simultaneously, inform the clinical investigations in Aim 4. Aim 4 (Phase II): Compare the performance of the Pocket colposcope to Visual Inspection with Acetic Acid for triage of HPV+ women in Kenya. We will carry out a cluster-randomized trial among 400 HPV+ women to compare the standard triage with that using the Pocket colposcope in Kisumu, Kenya. All HPV+ women will undergo biopsy to determine sensitivity, specificity and positive and negative predictive values of the different triage strategies. Data will be used to model the performance of the algorithm against that of expert colposcopists. Aim 5 (Phase II): Assess the costs, incremental cost-effectiveness and population health impact of HPV-based cervical cancer screening programs with proposed triage strategies. We will determine the incremental cost-effectiveness ratio and the absolute and relative costs for four triage strategies by measuring the costs and model population health outcomes (cancer cases, deaths and disability adjusted life years).