Improving STELLA Linac design by early fault detection and preventative maintenance for reducing Linac downtime using AI and ML

在全球范围内，癌症是非传染性疾病中第二大最常见的死亡原因，据估计，全球癌症年发病率将从2020年的1 930万例和1 000万例死亡增加到2040年的2 750万例和16.3例死亡。其中约65%至70%的增长将发生在低收入和中等收入国家（LMIC）。放射治疗（RT）对于治疗或缓解超过一半的癌症患者至关重要，但全球缺乏这种治疗方法。直线加速器（LINAC）提供最先进的治疗，但该技术的获取、运营和服务成本很高，特别是对于低收入和中等收入国家（LMIC），并且其恶劣的环境通常会对机器性能产生负面影响，导致大量停机时间。与HIC相比，LMIC中的RT设备会遭受更多的停机时间和更高的故障率（高收入国家）部分原因是环境因素，但也缺乏预防性维护和早期发现问题。该项目的想法建立在ITAR之前的研究基础上（创新技术，建立负担得起的和公平的全球放射治疗能力），其中检查了直线加速器的故障和故障。在这里，我们将研究如何将人工智能和机器学习用于早期检测、干预和预防，以减少特别是在发展中国家发生的长时间停机。目前的项目将着眼于使用AI/ML工具来检测最常发生故障的Linacs选定主要组件的故障，并使用数据开发自动化程序，最终可以远程预测故障，以提醒用户及时订购备件并减少设施的停机时间。

Globally, cancer is the second most prevalent cause of death among non-communicable diseases and it is estimated that the annual global cancer incidence will rise from 19.3 million cases and 10 million deaths in 2020 to as many as 27.5 million cases and 16.3 deaths in 2040. About 65 to 70% of this increase will occur in Low- and Middle-Income Countries (LMICs). Radiation therapy (RT) is critical for the treatment or palliation of over half of all patients with cancer, yet there is a global shortage in access to this treatment. Linear accelerators (LINACs) offer state-of-the-art treatment but this technology is high cost to acquire, operate and service, especially for Low- and Middle-Income Countries (LMICs), and often their harsh environment negatively affects machine performance causing large downtimes.RT equipment in LMICs is subject to more downtime and higher failure rates compared to HIC (Higher Income Countries) partly due to environmental factors, but also a lack of preventative maintenance and early recognition of problems.This project idea builds on prior studies under ITAR (Innovative Technologies towards building Affordable and equitable global Radiotherapy capacity) where the faults and breakdowns in Linacs were examined. Here how AI and ML could be used for early detection, intervention, and prevention to reduce the long downtimes that occur especially in developing countries will be examined. The current project will look at the use AI/ML tools to detect faults in selected major components of Linacs that breakdown most often and using the data, develop automated program(s) that can eventually predict faults remotely to alert users to needed repairs in time to order spare parts and reduce downtime of the facility.