Detection of suspicious lesions in dynamic contrast enhanced MRI data

Abstract

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) has become an important source of information to aid breast cancer diagnosis. Nevertheless, next to the temporal sequence of 3D volume data from the DCE-MRI technique, the radiologist commonly adducts information from other modalities for his final diagnosis. Thus, the diagnosis process is time consuming and tools are required to support the human expert. We investigate an automatic approach that detects the location and delineates the extent of suspicious masses in multi-temporal DCE-MRI data sets. It applies the state-of-the-art support vector machine algorithm to the classification of the short-time series associated with each voxel. The ROC analysis shows an increased specificity in contrast to standard evaluations techniques.

Performance of a fully automatic lesion detection system for breast DCE‐MRI

“…An automatic lesion detection method based on support vector machine, proposed by Twellmann et al also showed promising results, yielding an area under the ROC curve of 0.98. However, the algorithm was tested on a limited dataset of 12 patients and only on non-fat-sat images (16). The above mentioned methods cannot be applied to fat-sat images as normalization is performed by dividing each enhanced images by the unenhanced one.…”

“…Similarly to the technique used by Twellmann et al (16), a satisfactory segmentation can be obtained by applying morphological operations such as hole filling and dilation (6 steps with a 3 Â 3 Â 3 kernel) to the thresholding results obtained by means of Otsu's method.…”

“…As it is not operator dependent, a fully automatic lesion segmentation process has the potential to reduce reading time and provide more reproducible results. Unfortunately, few studies have addressed automatic lesion detection and segmentation techniques for breast DCE-MRI (14)(15)(16). Furthermore, to our knowledge these methods have been tested only on non-fat-saturated (fat-sat) contrast-enhanced images.…”

Model-free visualization of suspicious lesions in breast MRI based on supervised and unsupervised learning

“…First, the constraints (7) are relaxed by introducing slack variables ξ i : (10) (11) (12) This soft-margin formulation of the support vector machine allows to tolerate a ceratin amount of margin violations, controlled by the regularization parameter C, and leads to reasonable linear classification functions even in the presence of noise or class overlap.…”

“…Automatic selection of such ROIs is possible (see e.g. [26,12]), but not in the scope of this work. The time-series signals underlying the marked voxels are analyzed with respect to fine-grained differences in the amplitude and dynamics.…”