Development of the initial prototype of a pill sensor to detect colonic polyps and early bowel cancer

肠癌每年导致近100万人死亡，超过一半的病例是致命的。尽管早期癌症检测可以显著改善患者的预后，但英国超过一半的肠癌病例是在晚期诊断出来的。目前，对肠癌和癌前息肉的检测主要是通过内窥镜(结肠镜)期间对结肠粘膜的肉眼检查(这是一种侵入性程序)，或者通过横断面成像进行，对于不易观察到的小病变来说，横断面成像的可靠性较低。如果这种息肉没有在早期阶段被发现和切除，它们就有可能癌变。最近，使用结肠胶囊内窥镜直接对结肠进行可视化检查已被引入，但由于担心这种方法会遗漏病变，临床医生的摄取一直受到限制，这可能会给患者带来灾难性的后果。利用我们在可控胶囊技术领域的开创性工作，我们的团队开发了一种新型的、非束缚的、振动的、人工智能辅助的可食用药丸传感器，旨在通过一种新的方法检测微小的结肠病变(DOI：10.1109/LRA.2023.3251853)。吞下这种基于药丸的设备后，它将通过肠道运动和蠕动通过患者的胃肠道(GI)。可以从与原位肠道病变接触的药丸中获取动态信号，并分析其特征，以指示组织中的生物力学变化，以推断良性或恶性病变。设计创新包括用于鼓励药丸-病变相互作用的振动机制、用于相互作用信号采集的便携式平台和一次性部件：每次手术后，药丸的外壳被丢弃，而主要部件无需再加工即可回收，从而大大降低了成本。该平台体积小、重量轻，与标准结肠镜检查相比，资本成本显著降低。我们实验室的早期实验表明，模拟结肠病变的平均检测准确率为96.5%(DOI：10.1109/LRA.2023.3251853)。因此，这种新型的药丸传感器计划用于替代结肠镜检查和目前的结肠胶囊内窥镜检查。该项目旨在开发这种药丸传感器的初始原型，将其与人工智能相融合，以帮助检测难以可视化的肠道病变。这一新型诊断工具将在这项为期18个月的研究计划中进行开发、优化和测试。到这个项目结束时，我们将在TRL 3交付一个概念验证原型，准备进行体内测试。为此，我们将通过(1)集成所有必要的组件，包括车载可控震源和传感器(例如，加速度计)、数据存储和电源，(2)优化药丸的可控性和完整性，(3)使用车载传感器获取的动态信号来执行自主检测病变，(4)在实验室环境中的多个验证场景中使用标准胶囊内窥镜对所提出的技术进行基准测试，(5)基于临床反馈优化初始原型用于临床。从长远来看，这项工作将启动一种新的、微创、为患者和临床医生提供舒适、安全、可靠、准确和具有成本效益的检查模式，用于检测癌前病变和早期癌病变。

Bowel cancer causes nearly a million deaths per annum, and more than half of cases are fatal. Although early cancer detection can significantly improve patients' outcomes, more than half of bowel cancer cases in the UK are diagnosed at a late stage. Detection of bowel cancer and pre-cancerous polyps is currently predominantly performed by either visual inspection of the colonic mucosa during endoscopy (colonoscopy), which is an invasive procedure, or by cross-sectional imaging, which is less reliable for small-sized lesions that are not easily visualised. If such polyps are not detected and removed at an early stage, there is a chance that they may become cancerous. Recently, direct visualisation of the colon using colonic capsule endoscopy has been introduced, but uptake by clinicians has been limited due to concerns about missed lesions with this modality, with potentially catastrophic outcomes for patients.Leveraging our pioneering work in the field of controllable capsule technologies, our team has developed a novel, untethered, vibrational, AI-assisted, ingestible pill sensor, with the aim of detecting small colonic lesions by a new modality (DOI:10.1109/LRA.2023.3251853). After swallowing this pill-based device, it will pass through the patient's gastrointestinal (GI) tract through gut motility and peristalsis. Dynamic signals from the pill in contact with in-situ bowel lesions can be acquired and analysed for features that are indicative of biomechanical changes in the tissues, to infer benign or malignant lesions. Design innovations include the vibrational mechanism for encouraging pill-lesion interactions, a portable platform for interaction signal acquisition, and disposable components: after each procedure, the pill's outer shell is discarded, whilst the main components are reclaimed without reprocessing, dramatically reducing the cost. The platform is small and lightweight, with significantly reduced capital costs compared to standard colonoscopy. Early experiments in our laboratory demonstrate an average accuracy of 96.5% in successful detection of simulated colonic lesions (DOI:10.1109/LRA.2023.3251853). This novel pill sensor is therefore planned to facilitate an alternative to both colonoscopy and current colonic capsule endoscopy.This project aims to develop an initial prototype of this pill sensor in fusion with artificial intelligence to aid the detection of hard-to-visualise bowel lesions. This novel diagnostic tool will be developed, optimised, and tested in this 18-month research programme. By the end of this project, we will deliver a proof-of-concept prototype at TRL 3 that is ready for in-vivo testing. To this end, we will pursue this aim by (1) integrating all the required components, including on-board vibrator and sensor (e.g., accelerometer), data storage, and power supply, (2) optimising the controllability and integrity of the pill, (3) using the dynamic signals acquired by the on-board sensor to perform autonomous detection of lesions, (4) benchmarking the proposed technology with standard colonic capsule endoscopy in multiple validation scenarios in a laboratory setting, (5) optimising the initial prototype for clinical use based on clinical feedback.In the long term, this work will initiate a new, minimally invasive, investigative modality for patients and clinicians that is comfortable, safe, reliable, accurate and cost-effective in the detection of pre-cancerous and early cancerous lesions.