Leveraging Neural Imaging for Automated Neonatal Infection Diagnosis

项目摘要/摘要 感染后脑积水(PIH)是发展中国家新生儿死亡的主要原因，但 是否有有限的资源用于适当地诊断和监测导致 脑积水。通常缺乏人员和实验室资源可用于采集和 腰穿和血培养的处理是诊断感染性疾病的金标准 在败血症和妊高征中发挥作用的药物。为了克服这一障碍，脑脊液和血液样本从 乌干达Mbale的一群败血症新生儿，以及一群已经 后来患上了妊高征。头颅超声(CRU)取自败血症新生儿队列，头颅CT 扫描是从PIH队列中收集的。这一提议假设病原体从RNA中确定 血液和脑脊液样本的DNA测序可以用于训练有监督的机器学习算法 以识别潜在病原体的影像表型特征。因此，妊高征是可以预防的。 通过在床边为脓毒症提供病原体特异性诊断和有针对性的治疗建议 使用CRU的新生儿。此外，对于妊高征者，使用CT可以优化手术治疗的成功率。 确定潜在病原体并提供管理计划建议的目的。 该项目为对儿童神经外科感兴趣的研究员提供了一个理想的培训环境 重点研究了工程和机器学习在图像分析中的应用。跨学科和 该项目的全球性鼓励开发一种协作和创新的研究方法。这个 宾夕法尼亚州立大学的家庭机构为儿科神经外科的发展提供了多种临床机会 博士/博士项目支持真正的翻译研究努力，发起人和联合发起人更多 而不是做好充分的准备，提供实现这一目标所需的所有方面的培训指导 规划和培养全面发展的内科医生-科学家。

PROJECT SUMMARY/ABSTRACT Post-infectious hydrocephalus (PIH) is a leading cause of neonate mortality in the developing world, but there are limited resources in place for appropriately diagnosing and monitoring the infections that lead to hydrocephalus. There is often a lack of personnel and laboratory resources available for the gathering and processing of lumbar puncture and blood cultures, which are the gold-standard for diagnosing the infectious agents at play in sepsis and PIH. In order to overcome this obstacle, CSF and blood samples were taken from a cohort of septic neonates in Mbale, Uganda, as well as a cohort of neonates and infants who had already progressed to PIH. Cranial ultrasounds (CrUS) were taken from the cohort of septic neonates, and head CT scans were gathered from the PIH cohort. This proposal hypothesizes that the pathogens determined from RNA and DNA sequencing of the blood and CSF samples can be used to train supervised machine learning algorithms to recognize imaging phenotypes characteristic of the underlying pathogen. Therefore, PIH can be prevented by providing pathogen-specific diagnosis and targeted treatment recommendations at the bedside for septic neonates using CrUS. Furthermore, surgical treatment success for PIH can be optimized using CT for the purpose of identifying the underlying pathogen and providing management plan recommendations. This project provides an ideal training environment for a fellow interested in pediatric neurosurgery with a research emphasis on engineering and machine learning applied to image analysis. The interdisciplinary and global nature of the project encourages development of a collaborative and innovative research approach. The home institution of Penn State provides multiple clinical opportunities for growth in pediatric neurosurgery, the MD/PhD program is supportive of truly translational research efforts, and the sponsor and co-sponsor are more than adequately prepared to provide all aspects of training mentorship necessary to accomplish the aims of this project and develop a well-rounded physician-scientist.