서브메뉴
검색
Meningioma Relational Database Curation Using a PACS-Integrated Tool for Collection of Clinical and Imaging Features
Meningioma Relational Database Curation Using a PACS-Integrated Tool for Collection of Clinical and Imaging Features
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211151052
- ISBN
- 9798382321370
- DDC
- 610
- 저자명
- McLean, Ryan.
- 서명/저자
- Meningioma Relational Database Curation Using a PACS-Integrated Tool for Collection of Clinical and Imaging Features
- 발행사항
- [Sl] : Yale University, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 74 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 85-11, Section: B.
- 주기사항
- Advisor: Aboian, Mariam.
- 학위논문주기
- Thesis (M.D.)--Yale University, 2024.
- 초록/해제
- 요약Meningiomas are the most common primary intracranial neoplasm seen in adults, and despite the largely benign nature of these tumors, they are associated with significant morbidity and mortality. Physicians specialized in CNS pathologies, such as neurosurgeons, neuroradiologists, and neuro-oncologists rely on multiparametric MRI (mpMRI) for diagnosis, guidance of management, and response to therapy of CNS lesions. Today, in the era of artificial intelligence, with high-throughput computing and accessibility to massive amounts of data, we see the emergence of new fields in translational science. Proteomics, genomics, metabolomics, and more recently, radiomics are quantitative and analytical fields that are driven by the accumulation of data. Advancements in these areas, as well as computer and data science in general, have resulted in the development of automated, objective, and quantitative tools that can provide non-invasive assessments of tumors. However, the accessibility to large amounts of medical imaging data for machine learning (ML) and deep learning (DL) purposes is limited, resulting in a shortage of such tools for meningiomas. To bridge this gap, in this thesis we specifically focus on contributing to the development of automated segmentation tools by building a database of meningiomas.The Brain Tumor Segmentation (BraTS) Challenge has formed a multi-institutional coalition to increase the accessibility of automated segmentation tools for CNS tumors. To do this, the BraTS network aims to compile the largest annotated multilabel meningioma dataset. This dataset will serve as the training dataset for participants in the challenge to develop automated segmentation models for meningiomas. Models will then be validated using standardized metrics. The ultimate goal of this challenge is to provide a resource of segmentation tools and to facilitate incorporation of this technology into clinical practice to improve the care and outcomes of meningioma patients.This thesis will have two primary aims. First, compile a series of manually segmented and annotated MR scans of meningioma patients treated at Yale-New Haven Hospital to contribute to the BraTS Challenge. Second, outline a method of database curation that accommodates the storage, organization, and secure transfer of clinical, genomic, and imaging data to streamline. The ultimate objective is to reduce the time and labor burden on researchers interested in data collection and organization for prediction algorithm development via machine learning or deep learning.To accomplish the second aim, we utilized Fast Healthcare Interoperability Resources (FHIR) webforms, a customizable questionnaire that can be integrated into a picture archiving and communication system (PACS). A PACS-integrated tool serves as a bridge between the electronic medical record (EMR) and PACS, where patient medical imaging is organized and stored. This is advantageous in the context of research and data collection because it can reduce or eliminate the requirement of external software that is typically used for the storage of clinical variables. Utilization of external software, particularly in the context of a large research group, can often result in a disjointed workflow and errors making the management of large amounts of data challenging. This approach provides an avenue for researchers interested in developing ensemble models that use image-based measurements and features to augment traditional clinical-only/clinical-weighted staging and treatment response prediction algorithms.
- 일반주제명
- Medicine
- 일반주제명
- Biomedical engineering
- 일반주제명
- Medical imaging
- 키워드
- Deep learning
- 키워드
- Machine learning
- 키워드
- Meningiomas
- 기타저자
- Yale University Yale School of Medicine
- 기본자료저록
- Dissertations Abstracts International. 85-11B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017160634
■00520250211151052
■006m o d
■007cr#unu||||||||
■020 ▼a9798382321370
■035 ▼a(MiAaPQ)AAI31141731
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a610
■1001 ▼aMcLean, Ryan.
■24510▼aMeningioma Relational Database Curation Using a PACS-Integrated Tool for Collection of Clinical and Imaging Features
■260 ▼a[Sl]▼bYale University▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a74 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 85-11, Section: B.
■500 ▼aAdvisor: Aboian, Mariam.
■5021 ▼aThesis (M.D.)--Yale University, 2024.
■520 ▼aMeningiomas are the most common primary intracranial neoplasm seen in adults, and despite the largely benign nature of these tumors, they are associated with significant morbidity and mortality. Physicians specialized in CNS pathologies, such as neurosurgeons, neuroradiologists, and neuro-oncologists rely on multiparametric MRI (mpMRI) for diagnosis, guidance of management, and response to therapy of CNS lesions. Today, in the era of artificial intelligence, with high-throughput computing and accessibility to massive amounts of data, we see the emergence of new fields in translational science. Proteomics, genomics, metabolomics, and more recently, radiomics are quantitative and analytical fields that are driven by the accumulation of data. Advancements in these areas, as well as computer and data science in general, have resulted in the development of automated, objective, and quantitative tools that can provide non-invasive assessments of tumors. However, the accessibility to large amounts of medical imaging data for machine learning (ML) and deep learning (DL) purposes is limited, resulting in a shortage of such tools for meningiomas. To bridge this gap, in this thesis we specifically focus on contributing to the development of automated segmentation tools by building a database of meningiomas.The Brain Tumor Segmentation (BraTS) Challenge has formed a multi-institutional coalition to increase the accessibility of automated segmentation tools for CNS tumors. To do this, the BraTS network aims to compile the largest annotated multilabel meningioma dataset. This dataset will serve as the training dataset for participants in the challenge to develop automated segmentation models for meningiomas. Models will then be validated using standardized metrics. The ultimate goal of this challenge is to provide a resource of segmentation tools and to facilitate incorporation of this technology into clinical practice to improve the care and outcomes of meningioma patients.This thesis will have two primary aims. First, compile a series of manually segmented and annotated MR scans of meningioma patients treated at Yale-New Haven Hospital to contribute to the BraTS Challenge. Second, outline a method of database curation that accommodates the storage, organization, and secure transfer of clinical, genomic, and imaging data to streamline. The ultimate objective is to reduce the time and labor burden on researchers interested in data collection and organization for prediction algorithm development via machine learning or deep learning.To accomplish the second aim, we utilized Fast Healthcare Interoperability Resources (FHIR) webforms, a customizable questionnaire that can be integrated into a picture archiving and communication system (PACS). A PACS-integrated tool serves as a bridge between the electronic medical record (EMR) and PACS, where patient medical imaging is organized and stored. This is advantageous in the context of research and data collection because it can reduce or eliminate the requirement of external software that is typically used for the storage of clinical variables. Utilization of external software, particularly in the context of a large research group, can often result in a disjointed workflow and errors making the management of large amounts of data challenging. This approach provides an avenue for researchers interested in developing ensemble models that use image-based measurements and features to augment traditional clinical-only/clinical-weighted staging and treatment response prediction algorithms.
■590 ▼aSchool code: 0265.
■650 4▼aMedicine
■650 4▼aBiomedical engineering
■650 4▼aMedical imaging
■653 ▼aDeep learning
■653 ▼aMachine learning
■653 ▼aMeningiomas
■653 ▼aSegmentation tools
■690 ▼a0564
■690 ▼a0541
■690 ▼a0574
■71020▼aYale University▼bYale School of Medicine.
■7730 ▼tDissertations Abstracts International▼g85-11B.
■790 ▼a0265
■791 ▼aM.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17160634▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


