DIFFERENCE: DIFFusion magnetic resonance imaging with Enhanced Resolution ENCoding - Precision Imaging in Cancer

癌症导致了英国三分之一的死亡，其管理仍然是一个挑战。例如，七分之一的男性一生中会被诊断出患有前列腺癌；10个男性中有8个能活10年，然而，一开始很难判断谁对治疗的反应是好是坏，实际上，谁可能根本不需要治疗。我们目前诊断癌症的方法是用切割针对前列腺进行取样（活检）。这可能是痛苦的，是侵入性的，重要的是，活组织检查可能会遗漏前列腺癌，因为它们没有针对正确的区域，或者因为在多达三分之一的患者中遗漏了癌症最具侵略性的部分而产生错误的诊断。成像技术已经在癌症诊断中发挥了关键作用。使用磁共振扫描仪的扩散成像对癌症成像特别好。它利用磁场在微观层面上扫描水分子的运动。由于这种水的运动在癌症中减少了，它们在背景中很显眼，很容易被发现。然而，扩散成像的问题是图像可能会出现模糊和畸形（“低质量”），因为在扫描身体器官运动的过程中需要时间。这些图像也只是二维的，2D的，而不是三维的，3D的，并且缺乏细节（“低分辨率”）。如果我们有精细的3D图像和更精确的数值测量，我们可以使用扩散成像来检测甚至是小的癌症，计划新的治疗方法，需要一个精确的癌症图片，并通过新的计算技术（“人工智能”）揭示当前隐藏的癌症信息，可能预测其未来的行为。然而，扩散成像数据的鲁棒性和灵敏度需要显著提高，以使这些复杂的分析工作。我的目标是把扩散成像变成一种精确的诊断工具，首先集中在前列腺癌上。具体来说，我将解决目前阻碍我们获得高质量、高分辨率3D扩散图像的技术挑战。挑战1：扩散成像耗时太长：呼吸、肠道收缩和心脏运动的运动模糊了我们扫描时试图看到的微观水运动。挑战2：来自图像的数据质量不够详细，无法对特定癌症区域进行精确治疗。挑战3：目前的分析只能实现非常简单的“扩散长度”测量。通过将扩散成像与人工智能（深度学习）相结合，开发更复杂的分析，可以更深入地了解癌症的微观结构以及癌症的行为方式。我将开发一种获取扩散图像的新方法，该方法将使用详细的磁共振信号“模拟”和一种新颖的3D“相位导航信号”来跟踪、建模并消除干扰我们希望捕获的扩散信号的运动。我还将开发更复杂的分析方法来更好地量化和显示这些信息。我将把新的成像方法与“人工智能”计算结合起来，看看它们是否能帮助预测单个患者癌症的侵袭性。我的研究将在伦敦国王学院进行，那里有一个独特的跨学科环境，有人工智能研究人员、临床医生和行业合作伙伴。在研究结束时，我相信新的高分辨率扩散技术和新的分析方法将提高临床显著前列腺癌的早期发现，避免不必要的活检，并在需要时指导活检，并有助于确定是否需要治疗。结合人工智能方法，我的工作将开辟新的机会，通过实现个性化的精确治疗，如磁共振引导放射治疗，来改变患者的治疗结果。

Cancer causes a third of all UK deaths and its management remains a challenge. For example, 1 in 7 men will be diagnosed with prostate cancer in their lifetime; 8 in 10 men will survive for >10 years, yet, it is difficult to tell at the beginning who is likely to respond well or poorly to treatment and indeed, who may not need treatment at all. Our current way of diagnosing cancer is to sample the prostate using cutting needles (a 'biopsy'). This can be painful, is invasive, and importantly, biopsies may miss prostate cancer because they are not aimed at the right area or yield a false diagnosis because they miss the most aggressive part of the cancer in up to a third of patients.Imaging already plays a key role in cancer diagnostics. Diffusion imaging using a magnetic resonance scanner is particularly good for imaging cancer. It uses magnetic fields to scan the motion of water molecules at a microscopic level. As this water motion is reduced in cancers, they stand out against the background and are easily detected. However, the problem with diffusion imaging is that the images can appear blurred and misshapen ('low quality') as it takes time to scan during which the body organs move: These images are also only two dimensional, 2D not three-dimensional, 3D, and lack fine detail ('low resolution').If we had 3D images with fine detail and more accurate numerical measurements instead, we could use diffusion imaging to detect even small cancers, to plan new treatments that require an exact picture of the cancer, and with new computing techniques ('artificial intelligence') unravel currently hidden information about the cancer that may forecast its future behaviour. However, the robustness and sensitivity of diffusion imaging data needs to be improved significantly for these sophisticated analyses to work.My aim is to turn diffusion imaging into a precise diagnostic tool, focussing first on prostate cancer. Specifically, I will tackle the technical challenges that currently prevent us from having high-quality, high-resolution 3D diffusion images.Challenge 1: Diffusion imaging takes too long: movement from breathing, bowel contraction and heart motion blurs the microscopic water movement we are trying to see when we scan.Challenge 2: The quality of data from images is not detailed enough for precise treatments to specific cancer areas.Challenge 3: Current analysis only achieves a very simple measurement of 'diffusion length '. Deeper insight into cancer microstructure, and how the cancer may behave, could be gained by developing more sophisticated analysis combining diffusion imaging with artificial intelligence (deep learning).I will develop a new method to acquire diffusion images that will use detailed magnetic resonance signal 'simulations' and a novel 3D 'phase navigator signal' to track, model and get rid of the movement that interferes with the diffusion signal we wish to capture. I will also develop more sophisticated analysis methods to better quantify and show this information. I will combine the new imaging method with 'artificial intelligence' computing to see if together they can help forecast the aggressiveness of an individual patient's cancer.My research will take place at King's College London, where there is a unique interdisciplinary environment with artificial intelligence researchers, clinicians, and industry partners. At the end of this fellowship, I believe the new high-resolution diffusion technique together with new analysis methods will improve the early detection of clinically significant prostate cancer, avoid unnecessary biopsy and guide biopsy when needed, and help determine whether treatment is required or not. Combined with artificial intelligence methods, my work will open up new opportunities to transform patient outcomes by enabling personalised precision treatment such as MR-guided radiotherapy treatment.