Scholarlink

<p>Dynamic contrast-enhanced (DCE) breast MRI is a highly sensitive modality for detecting breast cancer, but its limited specificity often leads to false-positive findings and unnecessary biopsies. Radiomics offers a quantitative approach to improve lesion characterization by capturing tumor-related image-based features. In this study, we explored the application of treatment response assessment maps (TRAMs), a technique based on delayed-contrast MRI originally developed for neuro-oncology, to breast imaging. TRAMs are sensitive to late contrast clearance dynamics influenced by vascular and tissue microenvironment properties. By capturing differences in permeability and contrast retention, TRAMs may enable improved differentiation between malignant tumors and abnormal non-tumor or benign tissues, thus enhancing diagnostic precision in breast MRI. Radiomic features were extracted from post-contrast DCE-MRI phases using fat-suppressed T1-weighted MR images (T1-MRI), capturing the dynamic enhancement patterns of 243 breast lesions. TRAMs were generated by subtracting delayed post-contrast images from early post-contrast images (after image pre-processing), and corresponding TRAMs-based radiomic features were calculated. Various machine learning classifiers were trained to differentiate between malignant and benign lesions. The diagnostic performances were evaluated using area under the curve (AUC), sensitivity, specificity, positive predictive values (PPV), negative predictive values (NPV), and accuracy, with 95% confidence intervals. TRAMs-based radiomic models yielded diagnostic performance comparable to conventional T1-MRI models across evaluated metrics. While T1-MRI models achieved AUCs between 0.79 and 0.85, several TRAMs feature combinations reached AUC point estimates of 0.87. A univariate model using the TRAMs feature BlueClustV (representing the volume of the largest cluster of pixels showing contrast clearance at the delayed time point) alone achieved an AUC of 0.86, with balanced sensitivity (0.82) and specificity (0.85), demonstrating high diagnostic value using a single interpretable feature. These results were further supported by competitive PPV, NPV, and accuracy values. ROC analysis demonstrated that TRAMs models maintained favorable sensitivity–specificity tradeoffs, with the univariate BlueClustV feature showing the most balanced curve, emphasizing its diagnostic utility in distinguishing malignant from benign lesions. In summary, TRAMs-based classifiers provide complementary physiological information to standard imaging, suggesting their potential to reduce benign biopsies and aid decision making in breast cancer diagnosis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Application of treatment response assessment maps (TRAMs), based on delayed-contrast MRI for radiomic characterization of breast lesions

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月5日

查看全文
  • Dianne Daniels,
  • Kfir Cohen,
  • David Last,
  • Shirley Sharabi,
  • Maayan Zuniga,
  • Nora Lahat,
  • Renata Faermann,
  • Osnat Halshtok,
  • Anat Shalmon,
  • David Samoocha,
  • Michael Gotlieb,
  • Yael Mardor,
  • Miri Sklair-Levy

摘要

Dynamic contrast-enhanced (DCE) breast MRI is a highly sensitive modality for detecting breast cancer, but its limited specificity often leads to false-positive findings and unnecessary biopsies. Radiomics offers a quantitative approach to improve lesion characterization by capturing tumor-related image-based features. In this study, we explored the application of treatment response assessment maps (TRAMs), a technique based on delayed-contrast MRI originally developed for neuro-oncology, to breast imaging. TRAMs are sensitive to late contrast clearance dynamics influenced by vascular and tissue microenvironment properties. By capturing differences in permeability and contrast retention, TRAMs may enable improved differentiation between malignant tumors and abnormal non-tumor or benign tissues, thus enhancing diagnostic precision in breast MRI. Radiomic features were extracted from post-contrast DCE-MRI phases using fat-suppressed T1-weighted MR images (T1-MRI), capturing the dynamic enhancement patterns of 243 breast lesions. TRAMs were generated by subtracting delayed post-contrast images from early post-contrast images (after image pre-processing), and corresponding TRAMs-based radiomic features were calculated. Various machine learning classifiers were trained to differentiate between malignant and benign lesions. The diagnostic performances were evaluated using area under the curve (AUC), sensitivity, specificity, positive predictive values (PPV), negative predictive values (NPV), and accuracy, with 95% confidence intervals. TRAMs-based radiomic models yielded diagnostic performance comparable to conventional T1-MRI models across evaluated metrics. While T1-MRI models achieved AUCs between 0.79 and 0.85, several TRAMs feature combinations reached AUC point estimates of 0.87. A univariate model using the TRAMs feature BlueClustV (representing the volume of the largest cluster of pixels showing contrast clearance at the delayed time point) alone achieved an AUC of 0.86, with balanced sensitivity (0.82) and specificity (0.85), demonstrating high diagnostic value using a single interpretable feature. These results were further supported by competitive PPV, NPV, and accuracy values. ROC analysis demonstrated that TRAMs models maintained favorable sensitivity–specificity tradeoffs, with the univariate BlueClustV feature showing the most balanced curve, emphasizing its diagnostic utility in distinguishing malignant from benign lesions. In summary, TRAMs-based classifiers provide complementary physiological information to standard imaging, suggesting their potential to reduce benign biopsies and aid decision making in breast cancer diagnosis.