Skip to Content
TomographyTomography
  • Article
  • Open Access

24 June 2025

17 Pages

AI-CAD-Guided Mammographic Assessment of Tumor Size and T Stage: Concordance with MRI for Clinical Staging in Breast Cancer Patients Considered for NAC

,
,
and
1
Department of Radiology, Seoul St. Mary’s Hospital, College of Medicine, The Catholic University of Korea, Seoul 06591, Republic of Korea
2
Division of Medical Oncology, Department of Internal Medicine, Seoul St. Mary’s Hospital, College of Medicine, The Catholic University of Korea, Seoul 06591, Republic of Korea
*
Author to whom correspondence should be addressed.
†
These authors contributed equally to this work.
This article belongs to the Special Issue Imaging in Cancer Diagnosis

Abstract

Objectives: To evaluate the agreement between AI-CAD-guided mammographic and MRI measurements of tumor size and T stage in breast cancer patients being considered for neoadjuvant chemotherapy (NAC). Methods: This retrospective study included 144 women (mean age, 52 ± 11 years) with invasive breast cancer who subsequently received NAC and underwent both AI-CAD mammography (score ≥ 10) and pre-treatment MRI. Tumor sizes from AI-CAD contours were compared with MRI using Pearson correlation, intraclass correlation coefficients (ICCs), and Bland–Altman analysis. Concordance was defined as a ±0.5 cm difference. The contour showing the highest agreement was used to compare T stage with MRI using weighted kappa. Results: The mean AI-CAD abnormality score was 86.3 ± 22.2. Tumor sizes on mammography were 3.0 ± 1.2 cm (inner), 3.8 ± 1.5 cm (middle), and 4.8 ± 2.2 cm (outer), while the MRI-measured tumor size was 4.0 ± 1.9 cm. The middle contour showed the strongest correlation with MRI (r = 0.897; ICC = 0.866), the smallest mean difference (–0.19 cm; limits of agreement, –1.87 to 1.49), and the highest concordance (61.1%). Agreement was higher in mass-only lesions than in NME-involved lesions (ICC = 0.883 vs. 0.775; concordance, 70.9% vs. 46.6%). T stage comparison using the middle contour showed substantial agreement with MRI (κ = 0.743 [95% CI, 0.634–0.852]; agreement, 88.2%), with higher concordance in mass-only lesions (93.0%) than NME-involved lesions (81.0%) and more frequent understaging in the latter (17.2% vs. 2.3%). Conclusions: AI-CAD-guided mammographic assessment using the middle contour demonstrated good agreement with MRI for tumor size and T stage, indicating its value as a supportive tool for clinical staging in MRI-limited settings.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.