Automated Orthopaedic Implant Identification through Artificial Intelligence

该项目旨在通过人工智能技术自动化在普通X光片上识别骨科植入物的过程。将开发一种软件，其中将提供一张普通的X光照片，并在几秒钟内确定植入物的制造和型号。该项目的目标是通过更好地利用资源来改变临床实践。植入物的自动识别可以导致更好的库存管理，并更好地利用外科和技术人员的时间，而不是花费数小时正确地识别植入物的制造商和型号。我们预计这在每一家关节置换医院都将是无价的，特别是那些位于大量流动人口地区的医院，他们以前可能曾使用不熟悉的外国植入物进行过全膝关节置换术/全髋关节置换术。此外，该项目还着眼于利用NHS未得到满足的需求，在膝关节和髋关节翻修手术的术前阶段快速准确地诊断不熟悉的植入物。这也将大大降低总体治疗成本和发病率。在资源有限或技术人员缺乏经验的领域，自动识别将在患者护理中发挥很大作用。该软件将使骨科外科医生在手术前对植入物的类型更有信心，从而在术中更精确地了解目前存在的植入物类型。这些自动化方法的性能通常与训练有素的外科医生和放射科医生持平，并优于普通医生。最有经验的外科医生在识别植入物类型方面的准确率为85.6%。人工智能在12个植入物模型(来自膝盖和髋关节)上的识别任务上提供了近99%的准确率，并在大约2,000张X光片上进行了训练和验证。然而，这些系统中的许多系统使用有限的训练数据和/或仅尝试识别少量的植入物类型，或将自身限制在一个关节上进行验证。我们的目标是确保对植入物识别系统进行系统验证，根据成本效益、稀有性和接合性对植入物进行优先排序，基于质量保证的、经过精心管理的培训数据集，对每种类型的至少100个实例进行验证。该项目将包括上肢和下肢X光。

The project aims to automate the process of identifying orthopaedic implants on plain radiographs through artificial intelligence techniques. A software will be developed, wherein, a plain radiograph will be fed and the make and model of the implant will be identified within seconds.The objective of the project lies in bringing a change in clinical practise by better utilisation of resources. Automatic identification of implants can lead to better inventory stock management, and better use of surgical and technical staff's time in contrast to spending hours correctly identifying implants' makes and models. We expect this to be invaluable in every joint replacement hospital, particularly those situated in areas with large migrant populations who may have been previously operated for a Total Knee Arthroplasty/Total Hip Arthroplasty with foreign unfamiliar implants. Alongside, the project looks towards tapping into the unmet need in the NHS to rapidly and accurately diagnose unfamiliar implants in the preoperative phase of revision arthroplasty of knee and hip. This will also allow a significant reduction in overall treatment cost and morbidity. In areas where resources are limited, or the technical staff inexperienced, automatic identification will go a long way in patient care.The software will make orthopaedic surgeons more confident of the type of implant prior to surgery, translating to greater precision intra-op, knowing what type of implant is present. Performance of these automated methods is typically on par with trained surgeons and radiologists, and superior to general practitioners. The most experienced surgeons demonstrated an accuracy of 85.6% in recognising the type of implant. Artificial Intelligence has shown to provide nearly 99% accuracy on an identification task on 12 implant models (from the knee and hip), trained and validated on approximately 2,000 radiographs.However, many of these systems use limited training data and/or only attempt to recognise small numbers of implant types, or confine themselves to validate on one joint. Our aim is to ensure a systematic validation of an implant identification system, with implants prioritised by cost-effectiveness, rarity and joint, on a minimum of 100 examples of each type, based on a quality assured, curated training dataset. The project will be encompassing both upper limb and lower limb X-rays.