Machine Learning Inspired Physical Models in Organs

机器学习启发了器官的物理模型

• 批准号: 10544288
• 负责人: Bilyana Tzolova
• 金额: $ 4.68万
• 依托单位: RICE UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-09-01 至 2024-08-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10544288
• 关键词: 3-Dimensional; Adenomyosis; Algorithms; Arteriovenous malformation; Automation; Benign Prostatic Hypertrophy; Blood Vessels; Blood flow; Breast; Cardiovascular system; Catheters; Cerebral Aneurysm; Chemicals; Clinical; Colorectal; Computer Models; Computing Methodologies; Coupled; Data; Deep Vein Thrombosis; Development; Diagnosis; Diagnostic; Diagnostic Imaging; Disease; Drainage procedure; Fistula; Freedom; Goals; Hemodialysis; Human; Hyperplasia; Image; Individual; Islet Cell Tumor; Joints; Laws; Liver; Location; Machine Learning; Malignant Neoplasms; Malignant neoplasm of liver; Manuals; Mathematics; Medical; Medical Imaging; Metastatic to; Methods; Modeling; Motivation; Nature; Noise; Operative Surgical Procedures; Organ; Patient Care; Patient-Focused Outcomes; Patients; Physician' s Role; Physicians; Physics; Play; Primary carcinoma of the liver cells; Procedures; Process; Pulmonary Embolism; Research; Rest; Rice; Role; Scanning; Stenosis; Structure; Techniques; Therapeutic; Therapeutic Embolization; Time; Training; Translating; Transportation; Universities; University of Texas M D Anderson Cancer Center; Uterine Fibroids; Vascular System; Vascular blood supply; Venous; Work; X-Ray Computed Tomography; accurate diagnosis; base; cost; deep learning; deep learning algorithm; deep neural network; disease diagnosis; experience; image processing; imaging Segmentation; imaging modality; improved; insight; method development; models and simulation; neural network; neural network algorithm; neural network architecture; novel; physical model; radiologist; solute; treatment planning; tumor

PROJECT SUMMARY The vascular system plays a crucial role in diagnostics, treatment, and surgical planning in a wide array of diseases. Historically, practitioners locate vessel manually on each image of a CT scan. This is a tedious process that can vary highly depending on the individual's experience and ability. Recently, there has been motivation to automate this process to save time and increase accuracy. This process, vessel segmentation, is challenging because of the small size of the vessel structure and the varying contrast and noise in medical images. Current image processing techniques have not been successful in resolving the full vascular systems in humans because of these challenges. However, a novel neural network algorithm has shown potential to reduce training times and increase accuracy per degree of freedom in medical imaging segmentation. Applying this algorithm in the liver vessel segmentation, and eventually other organs' vascular system segmentation shows great promise. In addition to achieving successful vessel segmentation of the full vascular system, there is motivation to create a model that simulates blood flow and mass transportation in the vascular system. This is accomplished by using coupled multidimensional computational models for the flow and transport within the blood vessels. The combination of these two aims will give a complete overview of the location and function of a patient's circulatory system. This research will be completed by the joint effort of the Computational and Applied Mathematics Department at Rice University and the Department of Imaging Physics, Division of Diagnostic Imaging at The University of Texas MD Anderson Cancer Center. The collaborative nature of this project allows mathematicians to work with physicians who are experienced in the diagnosis and treatment of many diseases. Leveraging everyone's strengths and background will allow for a successful development and implementation of this project.

项目摘要 血管系统在广泛的阵列中在诊断，治疗和手术计划中起着至关重要的作用 疾病。从历史上看，从业者将船只手动定位在CT扫描的每张图像上。这是一个乏味的 可以根据个人的经验和能力来高度变化的过程。最近，有 动机自动化此过程以节省时间并提高准确性。这个过程，船只细分， 由于容器结构的尺寸很小以及医疗中的对比度和噪音，因此具有挑战性 图像。当前的图像处理技术尚未成功解决完整的血管系统 人类是因为这些挑战。但是，一种新型的神经网络算法显示了减少的潜力 训练时间并提高医学成像细分中的自由度的准确性。应用此 肝容器分割中的算法，有时其他器官的血管系统分割显示 伟大的承诺。除了实现完整血管系统的成功船舶分割外，还有 创建模拟血管系统中血流和质量转运的模型的动机。这是 通过在血液中使用耦合的多维计算模型来完成 船只。这两个目标的结合将详细概述位置和功能 患者的电路系统。这项研究将通过计算的共同努力来完成 以及莱斯大学的应用数学系和成像部 德克萨斯大学医学博士安德森癌症中心的诊断成像。此的协作性质 项目允许数学家与经验丰富的医生一起诊断和治疗 许多疾病。利用每个人的优势和背景将允许成功发展， 该项目的实施。