서브메뉴
검색
Advancing Machine Learning for Precision Medicine and AI-Driven Clinical Decision Support for Bladder Cancer Treatment Response Assessment and Survival Prediction
Advancing Machine Learning for Precision Medicine and AI-Driven Clinical Decision Support for Bladder Cancer Treatment Response Assessment and Survival Prediction
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105233
- ISBN
- 9798291567494
- DDC
- 004
- 저자명
- Sun, Di.
- 서명/저자
- Advancing Machine Learning for Precision Medicine and AI-Driven Clinical Decision Support for Bladder Cancer Treatment Response Assessment and Survival Prediction
- 발행사항
- [Sl] : University of Michigan, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 168 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-02, Section: B.
- 주기사항
- Advisor: Hadjiyski, Lubomir M.;Matuszak, Martha M.
- 학위논문주기
- Thesis (Ph.D.)--University of Michigan, 2025.
- 초록/해제
- 요약Bladder cancer is a common and clinically challenging malignancy with a substantial impact on global health, making it crucial to develop advanced methodologies for diagnosis, prognosis, and treatment planning. This dissertation explores the application of advanced machine learning (ML) and artificial intelligence (AI) techniques to improve clinical decision-making, enhance assessment of treatment response, and predict patient outcomes for bladder cancer. The research developed and evaluated several AI-driven models, focusing on integrating multi-modal data sources, including imaging, clinical, and genomic information, to build predictive tools that can guide personalized treatment strategies.A key contribution of this work was the development of a deep-learning convolutional neural network model combined with radiomics features from CT urography, demonstrating the potential of AI in assessing bladder cancer treatment response to neoadjuvant chemotherapy. This model achieved an area under the receiver operating characteristic curve (AUC) of 0.77-0.80, highlighting its feasibility and potential for clinical implementation.Additionally, the dissertation introduced a multi-modal model (termed CRD) for five-year survival prediction of bladder cancer patients post-cystectomy. This model, which integrated clinical, radiomics, and deep-learning features, achieved an AUC of 0.87, demonstrating the importance of combining diverse data sources for more accurate survival predictions. An enhanced CRD model was proposed by incorporating large language models (LLMs) into the modeling to improve efficiency by automating clinical information extraction from unstructured electronic medical records. The results showed the LLM-assisted model achieved survival prediction AUCs comparable to those based on manually curated data, emphasizing the potential of LLMs to automate clinical data extraction. LLM reliability and its impact on downstream prediction model were further investigated. The results showed GPT-4 and Llama held outstanding consistency in information retrieval for predictive modeling.Furthermore, an observer study was conducted and its result demonstrated that AI-based decision support systems (CDSS-T) significantly improved diagnostic accuracy of 17 clinicians when assessing bladder cancer treatment response. The use of AI tools to assist in clinical decision-making resulted in a notable increase in diagnostic performance, with AUC improving from 0.73 to 0.77 (p = 0.002). These findings highlight the practical benefits of integrating AI into clinical practice, suggesting that AI can complement and enhance clinician expertise, particularly in complex diagnostic scenarios.The dissertation also explored methods for confidence estimation of AI outputs and examined the influence of case complexity and physician characteristics on AI-assisted assessments. This is crucial for providing the user with an estimate of the trustworthiness of AI output, allowing the user to properly weight the AI recommendation with their own assessment, thus maximizing the benefits of AI-assisted decision-making.In conclusion, this work advances the integration of AI and ML in bladder cancer prognosis and treatment response assessment. By developing robust predictive models, evaluating AI in a pre-clinical multi-institutional observer study, and investigating the reliability and trustworthiness of ML models, this research underscores the transformative potential of AI in improving patient care. Future research will focus on integrating multi-institutional data to enhance model generalizability, developing longitudinal models to track disease progression, and conducting prospective study of the impact of AI tools in clinical settings. Ethical considerations and human-AI collaboration will be essential to ensure AI supports, rather than replaces, clinician judgment. This dissertation contributes to integrating AI into clinical workflows for more personalized, accurate, and timely interventions to improve outcomes for bladder cancer patients worldwide.
- 일반주제명
- Computer science
- 일반주제명
- Medical imaging
- 일반주제명
- Oncology
- 일반주제명
- Biomedical engineering
- 키워드
- Bladder cancer
- 키워드
- Radiology
- 키워드
- Computer vision
- 기타저자
- University of Michigan Nuclear Engineering & Radiological Sciences
- 기본자료저록
- Dissertations Abstracts International. 87-02B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017359899
■00520260202105233
■006m o d
■007cr#unu||||||||
■020 ▼a9798291567494
■035 ▼a(MiAaPQ)AAI32271917
■035 ▼a(MiAaPQ)umichrackham006268
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a004
■1001 ▼aSun, Di.
■24510▼aAdvancing Machine Learning for Precision Medicine and AI-Driven Clinical Decision Support for Bladder Cancer Treatment Response Assessment and Survival Prediction
■260 ▼a[Sl]▼bUniversity of Michigan▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a168 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-02, Section: B.
■500 ▼aAdvisor: Hadjiyski, Lubomir M.;Matuszak, Martha M.
■5021 ▼aThesis (Ph.D.)--University of Michigan, 2025.
■520 ▼aBladder cancer is a common and clinically challenging malignancy with a substantial impact on global health, making it crucial to develop advanced methodologies for diagnosis, prognosis, and treatment planning. This dissertation explores the application of advanced machine learning (ML) and artificial intelligence (AI) techniques to improve clinical decision-making, enhance assessment of treatment response, and predict patient outcomes for bladder cancer. The research developed and evaluated several AI-driven models, focusing on integrating multi-modal data sources, including imaging, clinical, and genomic information, to build predictive tools that can guide personalized treatment strategies.A key contribution of this work was the development of a deep-learning convolutional neural network model combined with radiomics features from CT urography, demonstrating the potential of AI in assessing bladder cancer treatment response to neoadjuvant chemotherapy. This model achieved an area under the receiver operating characteristic curve (AUC) of 0.77-0.80, highlighting its feasibility and potential for clinical implementation.Additionally, the dissertation introduced a multi-modal model (termed CRD) for five-year survival prediction of bladder cancer patients post-cystectomy. This model, which integrated clinical, radiomics, and deep-learning features, achieved an AUC of 0.87, demonstrating the importance of combining diverse data sources for more accurate survival predictions. An enhanced CRD model was proposed by incorporating large language models (LLMs) into the modeling to improve efficiency by automating clinical information extraction from unstructured electronic medical records. The results showed the LLM-assisted model achieved survival prediction AUCs comparable to those based on manually curated data, emphasizing the potential of LLMs to automate clinical data extraction. LLM reliability and its impact on downstream prediction model were further investigated. The results showed GPT-4 and Llama held outstanding consistency in information retrieval for predictive modeling.Furthermore, an observer study was conducted and its result demonstrated that AI-based decision support systems (CDSS-T) significantly improved diagnostic accuracy of 17 clinicians when assessing bladder cancer treatment response. The use of AI tools to assist in clinical decision-making resulted in a notable increase in diagnostic performance, with AUC improving from 0.73 to 0.77 (p = 0.002). These findings highlight the practical benefits of integrating AI into clinical practice, suggesting that AI can complement and enhance clinician expertise, particularly in complex diagnostic scenarios.The dissertation also explored methods for confidence estimation of AI outputs and examined the influence of case complexity and physician characteristics on AI-assisted assessments. This is crucial for providing the user with an estimate of the trustworthiness of AI output, allowing the user to properly weight the AI recommendation with their own assessment, thus maximizing the benefits of AI-assisted decision-making.In conclusion, this work advances the integration of AI and ML in bladder cancer prognosis and treatment response assessment. By developing robust predictive models, evaluating AI in a pre-clinical multi-institutional observer study, and investigating the reliability and trustworthiness of ML models, this research underscores the transformative potential of AI in improving patient care. Future research will focus on integrating multi-institutional data to enhance model generalizability, developing longitudinal models to track disease progression, and conducting prospective study of the impact of AI tools in clinical settings. Ethical considerations and human-AI collaboration will be essential to ensure AI supports, rather than replaces, clinician judgment. This dissertation contributes to integrating AI into clinical workflows for more personalized, accurate, and timely interventions to improve outcomes for bladder cancer patients worldwide.
■590 ▼aSchool code: 0127.
■650 4▼aComputer science
■650 4▼aMedical imaging
■650 4▼aOncology
■650 4▼aBiomedical engineering
■653 ▼aBladder cancer
■653 ▼aRadiology
■653 ▼aComputer vision
■653 ▼aArtificial intelligence
■653 ▼aLarge language models
■690 ▼a0574
■690 ▼a0984
■690 ▼a0992
■690 ▼a0541
■71020▼aUniversity of Michigan▼bNuclear Engineering & Radiological Sciences.
■7730 ▼tDissertations Abstracts International▼g87-02B.
■790 ▼a0127
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17359899▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


