MRI Radiomic Signatures of DCIS to Optimize Treatment

DCIS 的 MRI 放射学特征可优化治疗

• 批准号: 10537149
• 负责人: Despina Kontos
• 金额: $ 59.75万
• 依托单位: UNIVERSITY OF WASHINGTON
• 依托单位国家: 美国
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-07-01 至 2027-06-30
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10537149
• 关键词: Address; Affect; Aftercare; American College of Radiology Imaging Network; Anxiety; Biological; Biological Assay; Biology; Breast; Breast Cancer Detection; Breast Magnetic Resonance Imaging; Clinical; Clinical Data; Clinical Markers; Collaborations; Data; Databases; Detection; Development; Diagnosis; Disease; Eastern Cooperative Oncology Group; Gene Expression; Genomics; Gold; Heterogeneity; Histopathology; Image; In Situ Lesion; Incidence; Individual; Institution; Interobserver Variability; Ipsilateral; Link; Local Therapy; MRI Scans; Machine Learning; Magnetic Resonance Imaging; Malignant Neoplasms; Mammographic screening; Mammography; Measures; Medical; Modeling; Molecular; Molecular Profiling; Morbidity - disease rate; Newly Diagnosed; Noise; Noninfiltrating Intraductal Carcinoma; Normal tissue morphology; Oncology; Operative Surgical Procedures; Outcome; Pathologic; Pathology; Patients; Penetration; Pennsylvania; Performance; Phenotype; Physicians; Plant Roots; Prognosis; Prognostic Factor; Public Health; Radiation therapy; Radiosurgery; Recurrence; Reproducibility; Risk; Risk Assessment; Sampling; Science; Semantics; Signal Transduction; Staging; Standardization; Statistical Data Interpretation; Survival Rate; Systemic Therapy; Testing; Thick; Tissue Sample; Tissues; Universities; Unnecessary Surgery; Washington; Woman; Work; aggressive therapy; angiogenesis; base; breast cancer diagnosis; breast imaging; calcification; cancer invasiveness; clinical database; clinical prognostic; cohort; combat; experience; health goals; high risk; hormone therapy; imaging biomarker; improved; indexing; inter-institutional; malignant breast neoplasm; molecular marker; multidimensional data; non-invasive imaging; novel; novel strategies; oncotype; open source; overtreatment; phenomics; phenotypic data; prognostic; prognostic index; prognostic model; radiomics; risk prediction model; risk stratification; side effect; software development; standard of care; statistical center; tool; treatment optimization; tumor; user-friendly

Abstract/Project Summary: The purpose of this study is to determine whether breast MRI radiomic features can be utilized to optimize treatment of ductal carcinoma in situ (DCIS), the earliest form of breast cancer diagnosed. Although DCIS survival rates approach 100%, there is concern that its management generally results in overtreatment, exposing many of the 50,000 U.S. women diagnosed each year to unnecessary anxiety and morbidity. The vast majority of DCIS is detected in asymptomatic women in whom suspicious calcifications are identified on mammography and characterized using limited tissue histopathology. Unfortunately, conventional imaging and pathology have not proven reliable for distinguishing low vs. high-risk DCIS. Specifically, it is unclear at diagnosis which forms of DCIS will upstage to invasive disease or have an ipsilateral breast recurrence (IBR) after treatment. This limited risk-stratification is due in part to inadequate sampling of the entire DCIS lesion and an inability to account for peritumoral microenvironment features. This results in unnecessary surgery, radiation therapy, and medical therapy for as many as half of women diagnosed with DCIS. Breast MRI is commonly and easily performed, able to best depict DCIS span, and can assess tumor and peritumoral heterogeneity rooted in biological features such as angiogenesis, making it an appealing choice for a radiomics assay to improve DCIS risk assessments. The Quantitative Breast Imaging Lab at the University of Washington has shown that quantitative MRI features are associated with DCIS grade, a molecular marker of recurrence (Oncotype DX DCIS Score), and IBR. The Computational Biomarker Imaging Group at the University of Pennsylvania has pioneered breast MRI radiomic phenotyping and shown radiomic measures of breast cancers correlate with genomic features and recurrence. The Center for Statistical Sciences at Brown University has expertise with radiomics, machine learning, and statistical analyses for imaging trials from ECOG-ACRIN. In this collaborative application, we hypothesize that breast MRI radiomic signatures of DCIS will result in distinct phenotypes that are prognostic and can be integrated with clinical, molecular, and pathologic markers to optimize DCIS treatment. To test this hypothesis, we will create a multi-institutional database of over 1400 MRIs, including exams from the ECOG-ACRIN E4112 trial, with curated outcomes (e.g., upstage to invasion, DCIS Score, and IBR). Leveraging a novel approach to harmonize multicenter data (nested-Combat radiomic feature standardization), we will discover and validate MRI radiomic phenotypes and assess those phenotypes’ associations with invasive upstaging, Oncotype DX DCIS Score, and 5- and 10-year IBR. Finally, we will determine whether integration of these phenotypes into existing clinical prognostic indices (e.g., Van Nuys Prognostic Index) can provide more precise estimates of IBR. If successful, this study will help clinicians de-escalate DCIS therapy in low-risk patients and address an important public health goal: decreasing breast cancer overtreatment.

摘要/项目摘要：本研究的目的是确定乳腺 MRI 放射学特征是否 可用于优化导管原位癌 (DCIS)（乳腺癌的最早形式）的治疗 确诊。尽管 DCIS 存活率接近 100%，但人们担心其管理普遍存在问题。 导致过度治疗，每年诊断出 50,000 名美国女性，其中许多女性面临不必要的治疗 焦虑和发病。绝大多数导管原位癌是在无症状的女性中发现的，其中可疑 钙化通过乳房X线照相术进行鉴定，并使用有限的组织病理学进行表征。 不幸的是，传统的影像学和病理学尚未证明能够可靠地区分低风险与高风险 导管原位癌。具体来说，在诊断时尚不清楚哪些形式的导管原位癌会抢先发展为侵袭性疾病或具有 治疗后同侧乳房复发（IBR）。这种有限的风险分层部分是由于不充分 对整个 DCIS 病变进行采样，并且无法考虑瘤周微环境特征。这 导致多达一半的女性接受不必要的手术、放射治疗和药物治疗 诊断为导管原位癌 (DCIS)。乳房 MRI 是常见且容易执行的，能够最好地描述 DCIS 跨度，并且可以 评估源于血管生成等生物学特征的肿瘤和瘤周异质性，使其成为 放射组学检测是改善 DCIS 风险评估的一个有吸引力的选择。定量乳房成像 华盛顿大学的实验室表明，定量 MRI 特征与 DCIS 等级相关， 复发的分子标志物（Oncotype DX DCIS 评分）和 IBR。计算生物标志物成像 宾夕法尼亚大学的研究小组开创了乳腺 MRI 放射组学表型分析的先河，并展示了放射组学 乳腺癌的测量与基因组特征和复发相关。统计中心 布朗大学的科学专业拥有放射组学、机器学习和统计分析方面的专业知识 ECOG-ACRIN 的成像试验。在此协作应用中，我们假设乳腺 MRI 放射组学 DCIS 的特征将导致不同的表型，这些表型具有预后意义，并且可以与 临床、分子和病理标志物以优化 DCIS 治疗。为了检验这个假设，我们将 创建包含 1400 多个 MRI 的多机构数据库，包括 ECOG-ACRIN E4112 试验的检查， 具有精心策划的结果（例如，入侵的前期、DCIS 评分和 IBR）。利用新颖的方法 协调多中心数据（嵌套战斗放射组学特征标准化），我们将发现并验证 MRI 放射组学表型并评估这些表型与侵入性上期的关联，Oncotype DX DCIS 评分以及 5 年和 10 年 IBR。最后，我们将确定这些表型是否整合到 现有的临床预后指数（例如范奈斯预后指数）可以提供更精确的估计 IBR。如果成功，这项研究将帮助临床医生降低低风险患者的 DCIS 治疗并解决 重要的公共卫生目标：减少乳腺癌过度治疗。