서브메뉴
검색
Data-Efficient Algorithms for Advancing Medical Artificial Intelligence
Data-Efficient Algorithms for Advancing Medical Artificial Intelligence
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211153051
- ISBN
- 9798346389798
- DDC
- 519.54
- 서명/저자
- Data-Efficient Algorithms for Advancing Medical Artificial Intelligence
- 발행사항
- [Sl] : Stanford University, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 124 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-05, Section: B.
- 주기사항
- Advisor: Yeung-Levy, Serena.
- 학위논문주기
- Thesis (Ph.D.)--Stanford University, 2024.
- 초록/해제
- 요약As artificial intelligence increasingly integrates into medical practice, the need for data-efficient models becomes paramount, especially given the high cost and complexity of obtaining large, annotated datasets in the medical field. This dissertation is structured around three central themes:1. Utilizing Active Learning to Address Label Scarcity: Active learning is employed to strategically select the most informative data points for labeling, thereby maximizing the utility of limited labeling resources. The work introduces a novel adversarial representation active learning framework that effectively combines adversarial learning with active learning to enhance model performance in scenarios with limited labeled data. Additionally, a focused application on medical symptom recognition from patient-reported text demonstrates how active learning can address the challenges of long-tailed, multi-label distributions in telehealth settings.2. Maximizing Unlabeled Data with Semi-Supervised Learning and Domain Adaptation: The dissertation advances the state-of-the-art in semi-supervised learning by developing methods that leverage the vast amounts of unlabeled data available in medical contexts. The proposed AdaEmbed model for semi-supervised domain adaptation addresses the issue of domain shift, enabling AI models trained in one clinical environment to adapt effectively to new, unlabelled domains. This theme is further explored through the adaptation of surgical activity recognition models across different operating rooms, highlighting the importance of robust domain adaptation in achieving generalizable AI solutions.3. Enhancing Robustness and Efficiency with Pre-Trained Foundation Models: To overcome the challenges of limited data and enhance model robustness, the dissertation explores the use of pre-trained models and foundation models in healthcare applications. This is exemplified through the development of a video-based AI system for assessing facial paralysis, which uses pre-trained components to deliver accurate and standardized assessments. Additionally, the structured analysis of trauma care documentation using large language models demonstrates how AI can transform unstructured clinical notes into organized, actionable insights, improving the efficiency and effectiveness of trauma care.
- 일반주제명
- Sensitivity analysis
- 일반주제명
- Labeling
- 일반주제명
- Adaptation
- 일반주제명
- Trauma care
- 일반주제명
- Large language models
- 일반주제명
- Visualization
- 기타저자
- Stanford University.
- 기본자료저록
- Dissertations Abstracts International. 86-05B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017164822
■00520250211153051
■006m o d
■007cr#unu||||||||
■020 ▼a9798346389798
■035 ▼a(MiAaPQ)AAI31643331
■035 ▼a(MiAaPQ)Stanfordjr478gf9209
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a519.54
■1001 ▼aMottaghi, Seyyed Ali.
■24510▼aData-Efficient Algorithms for Advancing Medical Artificial Intelligence
■260 ▼a[Sl]▼bStanford University▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a124 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-05, Section: B.
■500 ▼aAdvisor: Yeung-Levy, Serena.
■5021 ▼aThesis (Ph.D.)--Stanford University, 2024.
■520 ▼aAs artificial intelligence increasingly integrates into medical practice, the need for data-efficient models becomes paramount, especially given the high cost and complexity of obtaining large, annotated datasets in the medical field. This dissertation is structured around three central themes:1. Utilizing Active Learning to Address Label Scarcity: Active learning is employed to strategically select the most informative data points for labeling, thereby maximizing the utility of limited labeling resources. The work introduces a novel adversarial representation active learning framework that effectively combines adversarial learning with active learning to enhance model performance in scenarios with limited labeled data. Additionally, a focused application on medical symptom recognition from patient-reported text demonstrates how active learning can address the challenges of long-tailed, multi-label distributions in telehealth settings.2. Maximizing Unlabeled Data with Semi-Supervised Learning and Domain Adaptation: The dissertation advances the state-of-the-art in semi-supervised learning by developing methods that leverage the vast amounts of unlabeled data available in medical contexts. The proposed AdaEmbed model for semi-supervised domain adaptation addresses the issue of domain shift, enabling AI models trained in one clinical environment to adapt effectively to new, unlabelled domains. This theme is further explored through the adaptation of surgical activity recognition models across different operating rooms, highlighting the importance of robust domain adaptation in achieving generalizable AI solutions.3. Enhancing Robustness and Efficiency with Pre-Trained Foundation Models: To overcome the challenges of limited data and enhance model robustness, the dissertation explores the use of pre-trained models and foundation models in healthcare applications. This is exemplified through the development of a video-based AI system for assessing facial paralysis, which uses pre-trained components to deliver accurate and standardized assessments. Additionally, the structured analysis of trauma care documentation using large language models demonstrates how AI can transform unstructured clinical notes into organized, actionable insights, improving the efficiency and effectiveness of trauma care.
■590 ▼aSchool code: 0212.
■650 4▼aSensitivity analysis
■650 4▼aLabeling
■650 4▼aAdaptation
■650 4▼aTrauma care
■650 4▼aLarge language models
■650 4▼aVisualization
■690 ▼a0800
■71020▼aStanford University.
■7730 ▼tDissertations Abstracts International▼g86-05B.
■790 ▼a0212
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17164822▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


