Improving Virtual Gross Anatomy: Enhancing the Information Content of Cadaveric CT Scans

改善虚拟大体解剖：增强尸体 CT 扫描的信息内容

• 批准号: 10377915
• 负责人: Steven Lewis
• 金额: $ 3.15万
• 依托单位: STATE UNIVERSITY OF NEW YORK AT BUFFALO
• 依托单位国家: 美国
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-04-01 至 2023-03-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10377915
• 关键词: 3-Dimensional; Address; Agreement; Anatomic Models; Anatomy; Artificial Intelligence; Blood coagulation; Buffaloes; Cadaver; Characteristics; Clinical; Computer Models; Contrast Media; Coupled; Cranial Nerves; Data; Development; Diagnostic; Disease; Dissection; Education; Exposure to; Gifts; Goals; Image; Image Analysis; Image Enhancement; Imaging Techniques; Individual; Information Retrieval; Infrastructure; Institution; Kidney; Lead; Libraries; Liver; Machine Learning; Mandible; Manuals; Medical; Medical Students; Methods; Modeling; Modernization; Morphology; Nature; Noise; Organ; Outcome; Pathologic; Pathology; Patients; Population; Preservation Technique; Process; Reading; Research; Research Personnel; Sample Size; Sampling; Scanning; Shapes; Signal Transduction; Site; Specimen; Statistical Data Interpretation; Structure; Students; Testing; Time; Tissues; Training; Universities; Variant; Work; X-Ray Computed Tomography; automated segmentation; base; biomedical imaging; clinically relevant; cohort; contrast enhanced; cost; deep learning; demographics; education research; feature extraction; human imaging; image processing; image registration; improved; innovation; interest; non-invasive imaging; posters; preservation; programs; quantitative imaging; retention rate; segmentation algorithm; shape analysis; soft tissue; success; virtual

Project Summary The long-term goal of this research is to establish a pipeline for automated image processing that enhances cadaveric non-contrast enhanced (NCE) CT data and extracts meaningful models and metrics to improve anatomy research and education. The objective is to develop the necessary toolset for this image processing, and feature extraction. The central hypothesis is that it is possible to enrich the information content of biomedical imaging data, particularly that of cadaveric NCE CT imaging, for use in gross anatomy education and research. The rationale behind this project is that cadaveric dissection, while an important part of anatomy education, is limited due to sample size, infrastructure, cost, and time. Biomedical imaging can preserve specimens for posterity and be used to supplement this material by providing statistical and quantitative information from anatomical structures. This research will attempt to establish a working pipeline for efficient information extraction through the following specific aims: (1) Improving inter-observer anatomical agreement in cadaveric CT scans; (2) Develop an approach to automatically segment anatomical structures from non-contrast enhanced CT images; and (3) Establish normal variation of anatomical structures and its relationship to pathologies. This project is innovative because it applies artificial intelligence to efficiently extract anatomical information from cadaveric NCE CT imaging, which has only been performed with traditional registration- dependent methods that often fail and are domain specific, acting on a single organ at a time. In addition, this project works with multi-species data to enhance human image data. This project is significant because it will allow students to understand anatomical variation better by both expanding student exposure to more samples, while also extracting useful representations and analytics from these samples for education and research. The expected outcome of this project is a toolset that is capable of enhancing anatomy education and research by increasing soft-tissue contrast, automatically segmenting the kidneys, liver, mandible, and intraosseus sites of the cranial nerves, and performing statistical analysis on these organs, including but not limited to statistical shape modelling and shape analysis. This will have a positive impact on anatomical education and student retention because it will provide students with a broader range of sample variability information which will decrease pervading biases in medical training that result from small, limited sample sizes, and improve medical training.

项目摘要 这项研究的长期目标是建立一个自动图像处理的管道， 尸体非对比增强（NCE）CT数据，并提取有意义的模型和指标，以改善 解剖学研究和教育。目标是开发用于这种图像处理的必要工具集， 特征提取。中心假设是，有可能丰富生物医学的信息内容， 成像数据，特别是尸体NCE CT成像数据，用于大体解剖学教育和研究。 这个项目背后的基本原理是，尸体解剖，而解剖学教育的重要组成部分， 由于样本大小、基础设施、成本和时间而受到限制。生物医学成像可以保存标本， 并通过提供统计和量化信息来补充本材料， 解剖结构本研究将尝试建立一个有效的信息提取工作管道 通过以下具体目标：（1）提高尸体CT中观察者之间的解剖一致性 （2）开发一种方法，从非造影剂中自动分割解剖结构 增强CT图像;（3）建立解剖结构的正常变异及其与 病理学该项目具有创新性，因为它应用人工智能来高效提取解剖结构 来自尸体NCE CT成像的信息，仅使用传统配准进行- 依赖的方法往往失败，是特定领域的，一次作用于一个器官。另外这款 该项目使用多物种数据来增强人类图像数据。 这个项目是重要的，因为它将使学生更好地了解解剖变异， 扩大学生接触更多的样本，同时也提取有用的表示和分析， 用于教育和研究。该项目的预期成果是一个工具集， 通过增加软组织对比度，自动分割 肾脏、肝脏、下颌骨和颅神经的骨内部位，并对这些部位进行统计分析 器官，包括但不限于统计形状建模和形状分析。这将有积极的 影响解剖学教育和学生保留，因为它将为学生提供更广泛的 样本可变性信息，这将减少医学培训中普遍存在的偏差， 有限的样本量，并改善医疗培训。