Accurate preoperative estimation of tumour size is vital in the management of breast cancer, particularly when the lesion presents as an architectural distortion (AD). These distortions, defined as abnormal arrangements of breast tissue without a definite mass, can represent both benign and malignant processes. Determining the precise extent of invasive carcinoma within ADs informs critical treatment decisions, such as surgical planning and eligibility for conservative interventions. Misjudging the clinical tumour size may result in incomplete resections or unnecessarily aggressive procedures.
A recent review published in European Radiology has evaluated the performance of five imaging modalities—Digital Mammography (DM), Digital Breast Tomosynthesis (DBT), Ultrasound (US), Contrast-Enhanced Mammography (CEM) and Magnetic Resonance Imaging (MRI)—in estimating pathological tumour size in ADs. A secondary aim is to assess the impact of including thin spicules in DM and DBT measurements on accuracy.
Comparative Performance of Imaging Modalities
Among the 63 ADs analysed from 59 female patients, the imaging techniques showed considerable variation in their ability to predict final pathological size. With pathology considered the gold standard, all modalities except US showed a tendency to overestimate tumour dimensions. DM and DBT, when measured including thin spicules, significantly exaggerated tumour size. In contrast, DBT excluding spicules, along with CEM and MRI, produced measurements most consistent with histological findings. These three techniques did not present statistically significant differences when compared to pathology.
Must Read: Ultrasound-Based Radiomics for Predicting HER2-Low Breast Cancer
The highest rate of concordance—defined as a difference within ±5 mm of the pathological size—was observed in CEM at 75%, followed by MRI at 67.6% and US at 65.1%. However, US also had the highest rate of underestimation at 25.4%. DM and DBT, when including thin spicules, yielded excessive overestimation rates of 93.7% and 93.1%, respectively, while the exclusion of spicules from these measurements significantly reduced this tendency. The inclusion of thin spicules, often representative of stromal response rather than tumour tissue, proved detrimental to accuracy.
Histological Subtypes and Imaging Concordance
The cohort comprised four histological subtypes: invasive ductal carcinoma (69.8%), invasive lobular carcinoma (23.8%) and smaller percentages of tubular (4.8%) and micropapillary carcinoma (1.6%). An analysis of the two most common subtypes—ductal and lobular—showed no significant differences in tumour size estimation when using DBT excluding thin spicules, CEM, MRI or US. For ductal carcinoma, DM excluding thin spicules significantly overestimated tumour size, whereas no significant difference was observed for lobular carcinoma. This suggests that DM performs more accurately in lobular subtypes when thin spicules are not measured. Similarly, DBT including thin spicules overestimated tumour size in ductal carcinoma and showed a nearly significant difference in lobular cases.
These findings underscore the variability in imaging accuracy depending on histological subtype and measurement method. Notably, thin spicules contributed little to accurate tumour delineation and were associated with systematic overestimation, particularly in mammographic techniques.
Clinical Implications and Limitations
These findings highlight the importance of selecting appropriate imaging modalities and measurement techniques for accurate preoperative staging. The high concordance of CEM and MRI with pathological size suggests their suitability for treatment planning, though limitations in access or patient suitability may preclude their use in some cases. In such instances, DBT excluding thin spicules offers a reliable alternative. The consistent overestimation associated with thin spicules supports their exclusion from measurement protocols in mammography and tomosynthesis. Given the potential impact on treatment outcomes, avoiding overestimation could reduce the risk of overtreatment and facilitate more conservative surgical approaches where appropriate.
Limitations of the study include its retrospective nature, single-centre design and reliance on measurements by a single radiologist with two years’ experience. Furthermore, the relatively small sample size and lack of interobserver variability assessment suggest the need for future multi-centre, prospective studies to confirm these findings. Nevertheless, the consistent trends observed provide valuable insight for imaging protocol refinement.
In cases of breast cancer presenting as architectural distortions, DBT excluding thin spicules, CEM and MRI provide the most accurate preoperative estimates of invasive tumour size. Thin spicules, while visible on mammography and tomosynthesis, often represent fibrous stromal response rather than true tumour extension, leading to substantial overestimation when included in measurements. US remains useful but tends to underestimate tumour size. DM, particularly when excluding thin spicules, can still be a valuable tool, especially in lobular carcinoma cases. Overall, excluding thin spicules in mammographic assessments significantly improves accuracy and should be adopted as standard practice to support more precise and appropriate clinical decision-making.
Source: European Radiology
Image Credit: Freepik
