Harnessing Coded Ptychography to Deliver AI-powered Evaluation of Unstained Lung Biopsies at the Point-Of-Care

项目摘要 该小型企业创新研究(SBIR)第一阶段项目旨在开发 成像平台分析肺活检标本的护理点。系统将 立即将未染色的标本数字化，并利用计算机辅助检测/诊断 使病理学家能够对样本进行快速有效的评估。广泛的、长期的 这项拟议研究的目标是实现对病理样本的实时诊断 护理点。 肺癌的早期诊断和治疗是必不可少的；存活率低， 高度依赖于疾病的阶段。镜检初步诊断 有20%-40%的活检(FNAs)由于标本不足而失败率。因此，重复 必须进行活组织检查，导致诊断和治疗延误长达90天，即 足够的时间让癌症抢占风头，并可能使存活率降低高达20%。快速- 由病理学家现场(ROSE)评估充分性可以保证诊断 活检的质量，但由于财务和操作的原因，10%的肺活检是进行的 障碍。 在这个项目中，我们建议克服历史上的障碍，使ROSE标准化 肺活组织检查的程序。这可以通过两项关键创新来实现：(I) 一种新的数字化显微模式(编码相变显微镜-CPM)的应用 病理学；和(Ii)未染色的肺FNAs的成像和自动分析使用人工 智能(AI)对象检测算法。对甲状腺FNAs的初步研究表明 CPM可以产生可视化良好的超分辨率定量相位图像(QPI 未染色的玻片上的细胞密度。将建造一个多波长紧凑型原型并进行测试 优化未染色肺FNAs的图像质量和速度。在展示了可重现性之后 高质量的图像、目标检测算法将在未染色的QPI上进行训练和验证， 并将引导病理学家到感兴趣的区域进行快速分析。所做的充分性评估 将对比使用和不使用人工智能对象检测辅助，以演示临床 有效性。在此可行性研究中产生的数据将快速跟踪产品开发和服务 作为临床验证的基础。这项工作的圆满完成是 长期目标是在全球范围内增加病理服务的可得性，共同努力 降低癌症相关死亡率。

Project Summary This Small Business Innovation Research (SBIR) Phase I project aims to develop an imaging platform to analyze lung biopsy samples at the point of care. The system will immediately digitize unstained specimens and utilize computer-assisted detection/diagnostics to enable quick and efficient evaluations of samples by pathologists. The broad, long-term objective of this proposed research is to enable real-time diagnostics of pathology samples at the point-of-care. Early diagnosis and treatment of lung cancer is essential; the survival rate is low and is highly dependent on the stage of the disease. Primary diagnoses through microscopic analysis of biopsies (FNAs) have a 20-40% failure rate due to inadequate specimens. As a result, repeat biopsies must be performed causing delays in diagnosis and treatment up to 90 days, which is enough time for cancer to upstage and can reduce survivability by as much as 20%. Rapid- onsite (ROSE) assessment of the adequacy by a pathologist can guarantee the diagnostic quality of biopsies but is performed in <10% of lung biopsies due to financial and operational barriers. In this project, we propose to overcome the historical barriers to ROSE to standardize the procedure for lung biopsies. This can be accomplished with two key innovations: (i) the Application of a novel microscopy modality (coded ptychography microscopy - CPM) for digital pathology; and (ii) imaging and automated analysis of unstained lung FNAs using Artificial Intelligence (AI) object detection algorithms. Preliminary work on thyroid FNAs indicates that CPM can produce super-resolution quantitative phase images (QPIs) with well-visualized cellularity on unstained slides. A multiwavelength compact prototype will be built and tested to optimize image quality and speed on unstained lung FNAs. After demonstrating reproducible high-quality images, object detection algorithms will be trained and validated on unstained QPIs, and will direct pathologists to areas of interest for quick analysis. Adequacy assessments made with and without AI object detection assistance will be compared to demonstrate the clinical validity. The data produced in this feasibility study will fast-track product development and serve as the foundation for clinical validation. Successful completion of this work is a key step in the long-term goal of increasing the availability of pathology services worldwide in a concerted effort to reduce cancer-associated mortality.