서브메뉴
검색
AI-Driven MRI Biomarkers for Breast Cancer Risk and Treatment Response Assessment
AI-Driven MRI Biomarkers for Breast Cancer Risk and Treatment Response Assessment
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105150
- ISBN
- 9798293831098
- DDC
- 616
- 저자명
- Yan, Ran.
- 서명/저자
- AI-Driven MRI Biomarkers for Breast Cancer Risk and Treatment Response Assessment
- 발행사항
- [Sl] : University of California, Los Angeles, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 107 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-03, Section: A.
- 주기사항
- Advisor: Sung, Kyung H.;Wu, Holden H.
- 학위논문주기
- Thesis (Ph.D.)--University of California, Los Angeles, 2025.
- 초록/해제
- 요약Breast cancer remains the most common malignancy in women worldwide, and quantitative biomarkers from breast MRI, including fibroglandular tissue (FGT) and background parenchymal enhancement (BPE), are emerging as important tools for risk assessment and treatment monitoring. To clarify the biological relevance of BPE, we analyzed screening MRI from 535 women stratified by lifetime risk and BRCA mutation status. After breast and FGT segmentation, six quantitative BPE measures were extracted, revealing strong associations with age, BMI, menopausal status, and breast density. Notably, BRCA1/2 carriers showed significantly lower enhancement than high-risk non-carriers after adjustment, suggesting a link between germline mutations and parenchymal physiology. Extending this work, we evaluated the predictive value of BPE for neoadjuvant chemotherapy (NAC) response in a multicenter cohort of 191 patients. Radiomics models combining tumor and BPE features achieved the best accuracy for predicting pathologic complete response, particularly in postmenopausal women and triple-negative cancers, highlighting the added value of BPE beyond tumor features alone. To enhance the reproducibility of breast MRI biomarkers, we developed an anatomy-aware loss function for deep learning-based FGT segmentation that integrates breast density information into the training process. Applied to 180 internal MRIs with independent evaluation on 40 held-out cases and an additional 100 external cases, this approach improved boundary accuracy and reduced biomarker estimation errors by approximately 10-15% compared with conventional loss functions, with the greatest benefit observed in fatty and scattered breasts. In conclusion, this work demonstrates that BPE is shaped by genetic and clinical risk factors, enhances prediction of treatment response, and can be more reliably quantified through anatomically informed segmentation, advancing its use as a robust imaging biomarker for breast cancer risk stratification and therapy monitoring.
- 일반주제명
- Medical imaging
- 일반주제명
- Oncology
- 일반주제명
- Womens studies
- 기타저자
- University of California, Los Angeles Bioengineering 0288
- 기본자료저록
- Dissertations Abstracts International. 87-03A.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017359637
■00520260202105150
■006m o d
■007cr#unu||||||||
■020 ▼a9798293831098
■035 ▼a(MiAaPQ)AAI32242067
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a616
■1001 ▼aYan, Ran.
■24510▼aAI-Driven MRI Biomarkers for Breast Cancer Risk and Treatment Response Assessment
■260 ▼a[Sl]▼bUniversity of California, Los Angeles▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a107 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-03, Section: A.
■500 ▼aAdvisor: Sung, Kyung H.;Wu, Holden H.
■5021 ▼aThesis (Ph.D.)--University of California, Los Angeles, 2025.
■520 ▼aBreast cancer remains the most common malignancy in women worldwide, and quantitative biomarkers from breast MRI, including fibroglandular tissue (FGT) and background parenchymal enhancement (BPE), are emerging as important tools for risk assessment and treatment monitoring. To clarify the biological relevance of BPE, we analyzed screening MRI from 535 women stratified by lifetime risk and BRCA mutation status. After breast and FGT segmentation, six quantitative BPE measures were extracted, revealing strong associations with age, BMI, menopausal status, and breast density. Notably, BRCA1/2 carriers showed significantly lower enhancement than high-risk non-carriers after adjustment, suggesting a link between germline mutations and parenchymal physiology. Extending this work, we evaluated the predictive value of BPE for neoadjuvant chemotherapy (NAC) response in a multicenter cohort of 191 patients. Radiomics models combining tumor and BPE features achieved the best accuracy for predicting pathologic complete response, particularly in postmenopausal women and triple-negative cancers, highlighting the added value of BPE beyond tumor features alone. To enhance the reproducibility of breast MRI biomarkers, we developed an anatomy-aware loss function for deep learning-based FGT segmentation that integrates breast density information into the training process. Applied to 180 internal MRIs with independent evaluation on 40 held-out cases and an additional 100 external cases, this approach improved boundary accuracy and reduced biomarker estimation errors by approximately 10-15% compared with conventional loss functions, with the greatest benefit observed in fatty and scattered breasts. In conclusion, this work demonstrates that BPE is shaped by genetic and clinical risk factors, enhances prediction of treatment response, and can be more reliably quantified through anatomically informed segmentation, advancing its use as a robust imaging biomarker for breast cancer risk stratification and therapy monitoring.
■590 ▼aSchool code: 0031.
■650 4▼aMedical imaging
■650 4▼aOncology
■650 4▼aWomens studies
■653 ▼aFibroglandular tissue
■653 ▼aBackground parenchymal enhancement
■653 ▼aNeoadjuvant chemotherapy
■690 ▼a0574
■690 ▼a0453
■690 ▼a0992
■690 ▼a0769
■71020▼aUniversity of California, Los Angeles▼bBioengineering 0288.
■7730 ▼tDissertations Abstracts International▼g87-03A.
■790 ▼a0031
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17359637▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
