Risk stratification of ductal carcinoma in situ using whole-lesion histogram analysis of the apparent diffusion coefficient

To investigate the value of the whole-lesion histogram apparent diffusion coefficient (ADC) metrics for differentiating low-risk from non-low-risk ductal carcinoma in situ (DCIS).

The authors identified 93 women with pure DCIS who had undergone preoperative MR imaging and diffusion-weighted imaging from 2013 to 2016. Histogram analysis of pixel-based ADC data of the whole tumour volume was performed by two radiologists using a software tool. The results were compared between low-risk and non-low-risk DCIS. Associations between quantitative ADC metrics and low-risk DCIS were evaluated by receiver operating characteristics (ROC) curve and logistic regression analyses.

In whole-lesion histogram analysis, mean ADC and 5th, 50th and 95th percentiles of ADC were significantly different between low-risk and non-low-risk DCIS (1.522, 1.207, 1.536 and 1.854 × 10⁻³ mm²/s versus 1.270, 0.917, 1.261 and 1.657 × 10⁻³ mm²/s, respectively; p = .004, p = .003, p = .004 and p = .024, respectively). ROC curve analysis for differentiating low-risk DCIS revealed that 5th percentile ADC yielded the largest area under the curve (0.786) among the metrics of whole-lesion histogram, and the optimal cut-off point was 1.078 × 10⁻³ mm²/s (sensitivity 80%, specificity 75.9%, p = .001). Multivariate regression analysis revealed that a high 5th percentile of ADC (> 1.078× 10⁻³ mm²/s; odds ratio [OR] = 10.494, p = .016), small tumour size (≤ 2 cm; OR = 12.692, p = .008) and low Ki-67 status (< 14%; OR = 10.879, p = .046) were significantly associated with low-risk DCIS.

Assessment with whole-lesion histogram analysis of the ADC could be helpful for identifying patients with low-risk DCIS.