MRI brain morphometry for computer-aided detection of neurological disorders

 描述（由申请人提供）：神经放射科医生常规使用的技术落后于世界各地成像研究中使用的技术数年或数十年。目前神经放射学的临床决策是基于不足以检测到的细微变化，预示着临床上重要的条件，如肿瘤生长，扩大脑室，或海马萎缩的信息。 研究中经常使用的工具在临床上无法使用，原因有很多：执行速度慢、对扫描仪制造商敏感、场强和脉冲序列以及缺乏FDA许可。在这个项目中，我们试图通过进一步开发FreeSurfer（世界上首屈一指的自动化脑形态测量软件包）来消除这些障碍，以（1）将新型深度学习和基于随机森林的块匹配图像合成技术集成到FreeSurfer中，使其对扫描仪平台和采集参数的变化具有鲁棒性，（2）使用现代并行处理将执行时间缩短到临床可行的长度，以及（3）获得FDA对这些工具的许可，允许它们在直接患者护理中常规使用。 由此产生的产品将极大地改善神经放射科医生可以获得的信息，用于患者诊断，分期和评估包括AD，肿瘤，脑积水和癫痫等疾病的治疗干预措施的疗效。

 DESCRIPTION (provided by applicant): The technology that is routinely available to neuroradiologists is years or decades behind that which is used in imaging research throughout the world. Current clinical decision-making in neuroradiology is based on information that is insufficient for detecting the subtle changes that presage clinically significant conditions such a a growing tumor, enlarging ventricle, or shrinking hippocampus. Tools that are regularly used in research are not accessible in the clinic due to a number of factors: slow execution speed, sensitivity to scanner manufacturer, field strength and pulse sequence, and lack of FDA clearance. In this project we seek to remove these barriers by further developing FreeSurfer, the world's premier automated brain morphometry package, to (1) integrate novel deep-learning and Random Forest- based patch-matching image synthesis technology into FreeSurfer to make it robust to variations in scanner platform and acquisition parameters, (2) use modern parallel-processing to reduce execution time to a clinically-feasible length, and (3) obtain FDA clearance on the tools allowing them to be routinely used in direct patient care. The resulting product would dramatically improve the information that neuroradiologists have access to for patient diagnosis, staging, and the assessment of the efficacy of therapeutic interventions for disorders including AD, tumor, hydrocephalus, and epilepsy, to name a few.