Innovative multi-channel AI method for the improved diagnosis of upper extremity bone fractures in the trauma setting.

该项目将开发一种创新的多通道人工智能方法的概念验证，以改进脆性骨折的预防和诊断。数字放射学(DR)仍然是创伤后骨折检测的标准成像协议，虽然大多数病理被正确识别，但有相当大一部分被遗漏，导致诊断延迟，增加疼痛和痛苦，并因重复诊断测试和治疗而增加成本。放射科医生评估标准DR图像的许多特征，以预测是否存在病理或未来骨折的风险，这需要对受影响区域的骨结构进行更详细的检查。然而，创伤放射科医生在根据预先确定的因素操纵图像对比度和确定患者风险方面的时间有限，如果某些病理情况在射线照片中没有立即显现出来，就会被遗漏。该项目将建立在独特的IBEX TrueView(R)技术的基础上，该技术是在先前InnovateUK资助的项目的支持下开发的。TrueView使用标准DR图像生成更高的图像质量和独特的基于构图的输出，以抑制软组织对比度并增强骨骼精细细节，从而提高对细微骨折和其他病理情况的可见性。它还能够提供基于骨密度的骨骼健康的准确测量。TrueView包括一种专有的基于人工智能的骨分割算法，可以在有限的训练数据下准确地从软组织中分割出骨。该项目将有三个主要目标：1)进一步增强专有的“IBEX XNET”基于AI的骨分割方法，以生成额外的定量骨形态输出；2)开发新的成分制导的后处理方法，以创建独立的骨增强和组织增强诊断图像；以及3)自动化检测或确定风险或未来脆性骨折的过程。

This project will develop the proof of concept of an innovative multi-channel AI method to improve the prevention and diagnosis of fragility fractures.Digital radiology (DR) remains the standard imaging protocol for fracture detection after trauma and whilst the majority of pathologies are correctly identified, a significant proportion are missed, leading to a delay in diagnosis, increased pain and suffering and additional costs from repeated diagnostic tests and treatment. Radiologists assess many features of a standard DR image to predict the presence of pathologies or the risk of future fractures that requires a more detailed inspection of bone structure in the affected region. However, trauma radiologists have limited time to manipulate image contrast and determine patient risk based on pre-determined factors, and certain pathologies will be missed if they are not immediately obvious in the radiograph.The project will build upon the unique IBEX Trueview(r) technology which has been developed with the support of previous InnovateUK funded projects. Using standard DR images, Trueview generates improved image quality and unique composition-based outputs to suppress soft tissue contrast and boost bone fine detail, thereby improving the visibility of subtle fractures and other pathologies. It is also able to provide an accurate measure of bone health based on bone density. Trueview includes a proprietary AI-based bone segmentation algorithm to accurately segment bone from soft tissue with limited training data.The project will have three primary aims: 1) To further enhance the proprietary "IBEX XNET" AI-based bone segmentation method to generate an additional quantitative bone morphology output; 2) To develop new composition-guided post processing methods to create independent bone-enhanced and tissue-enhanced diagnostic images; and 3) To automate the process of detecting or determining risk or future fragility fractures.