Rapid Sperm Capture

不孕不育，即在一年的努力想要孩子后无法怀孕，影响到大约六分之一的夫妇。精子的问题--例如，快速游泳的精子数量较少，或者精子形态不佳(可能损坏了DNA)--约占所有病例的一半。IVF或ICSI(将精子注射到卵子中)等疗法被用来治疗与精子有关的问题，但它们无效，非常昂贵，并给夫妇，特别是女性带来身体和情感上的压力。令人担忧的是，使用DNA受损的精子可能会导致流产或由此产生的孩子的健康问题。一些患者最好继续尝试自然生育，但要改变一些生活方式(戒烟，改善饮食)，或者进行侵入性较小的治疗(子宫内受精)。其他患者应该尽快转移到试管婴儿，一些人只会通过ICSI受孕。治疗决策的困难在于，确定精子问题的类型和严重程度的方法是不准确的。主要的方法是手动‘计数’游泳的精子，并将它们放在玻片上烘干以检查它们的形状。这并没有利用近年来计算机和摄像技术的巨大飞跃--事实上，即使是20世纪80年代的“计算机辅助精液分析”方法也没有在临床上普遍使用(部分原因是它们的准确性不令人满意)。想想典型智能手机中的技术-高清晰度(可能是快速帧速率)摄像头、模式识别和大容量数据处理/存储-这是需要引入生育临床的21世纪技术类型。我们将开发一种新的方法来检查精子，使用快速数码相机成像和基于计算机的模式识别。其目标将是能够自动、准确和重复地检查精液样本，收集关于细胞如何游泳以及它们的形状的同步数据。这项技术的目标是寻找形状合适、游动能力强的少数细胞--这样在自然受精的情况下，它们就能够通过宫颈、子宫和输卵管使卵子受精。将有许多实际问题需要解决。例如，我们是否需要让细胞发出荧光，这样我们才能更好地看到它们的形状，或者我们可以用标准类型的显微镜来实现我们的目标？我们可以处理任何浓度的样本吗，或者我们需要稀释它们以正确识别细胞吗？我们是不是应该先用微流体室来分离出游动细胞--我们是不是应该用一种高粘度的液体来更好地代表女性生殖道中精子所面临的生理挑战？一个关键问题将是如何将我们可以测量的大量信息转化为医生和患者可以使用的信息。我们将应用一种基于原型的机器学习，基于我们可以从快速精子视频中提取的许多“特征”来表示典型类型的患者。随着更多的患者通过该系统，这些原型将被逐步修改，使模型更加准确。从长远来看，通过我们可以训练的模型整合大量数据的可能性将导致一种非常强大的方式来整合国内甚至国际上的临床信息。随着该系统投入实际应用，患者将更快地接受正确的治疗，节省资源，患者将不那么困难的生育治疗体验，并将更快取得成功。更长期的好处将是帮助临床研究和毒理学--该系统将为研究人员提供一种强大的方法来测试生育治疗的新进展，例如药物、生活方式的改变，他们还将能够检查其他化学物质对精子的意外毒性影响。

Infertility, not being able to conceive after a year of trying for a baby, affects around one in six couples. Problems with sperm - for example low numbers of rapid swimming sperm, or poorly-formed sperm (which may have damaged DNA) contribute in around half of all cases. Therapies such as IVF, or ICSI (injection of sperm into an egg) are used to treat sperm-related problems, however they are ineffective, are very expensive and put physical and emotional strain on the couple, particularly the woman. Worryingly, using sperm with damaged DNA may contribute to miscarriage or health problems in any resulting children. Some patients would be better off continuing to try for a baby naturally, but with some lifestyle changes (stopping smoking, improving diet), or with a less invasive treatment (insemination into the womb). Other patients should be quickly moved to IVF, some will only conceive through ICSI. The difficulty with treatment decisions is that methods to determine the type and severity of sperm problems are imprecise. The main methods are manually 'counting' swimming sperm, and drying them out on a slide to examine their shape. This does not take advantage of the huge leaps in computer and camera technology made in recent years - indeed even 1980s 'Computer-Aided Semen Analysis' methods are not generally used in clinics (partly because of their unsatisfactory accuracy). Think of the technology in a typical smartphone - a high definition (possibly rapid framerate) camera, pattern recognition, and high volume data processing/storage - this is the type of 21st Century technology that needs to be brought into the fertility clinic.We will develop a new way to examine sperm, using both rapid digital camera imaging, and computer-based pattern recognition. The aim will be to be able to automatically, accurately and repeatably examine a semen sample, collecting simultaneous data on how cells swim, and what their shape is like. The target of this technique will be to look for the 'special' few cells that have the right shape, and can swim well - so that in natural fertilisation they would be able to travel through the cervix, womb and fallopian tubes and fertilise the egg. There will be a number of practical issues that will need to be solved. For example, do we need to make the cells fluorescent so we can see their shape better, or can we achieve our aims with a 'standard' type of microscopy? Can we work with samples at any concentration, or do we need to dilute them to recognise the cells properly? Should we use a 'micro-fluidic' chamber to separate out the swimming cells first - and should we use a high viscosity ('sticky') fluid that better represents the physical challenge sperm face in the female reproductive tract? A key question will be how to convert the large volume of information we can measure into information that doctors and patients can make use of. We will apply a type of machine learning based on prototypes, representations of typical types of patients based on the many 'features' we can extract from rapid sperm videos. These prototypes will be progressively modified as more patients come through the system, making the model more accurate. In the long term the possibility of integrating the large volume of data through a model we can train will lead to a very powerful way to bring together clinical information nationally or even internationally. As the system makes its way into real application, patients will receive the right treatment more quickly, saving resources, patients will have a less difficult experience of fertility treatment and will achieve success more quickly. A longer-term benefit will be by helping clinical research and toxicology - the system will provide researchers with a powerful method to test new advances in fertility treatment, for example drugs, lifestyle changes, and they will also be able to check for unintended sperm-toxic effects of other chemicals.