Visualization of Lesions in Breast MRI

Motivation and Problem Description

Breast cancer is the most common cancer among women, but has an encouraging cure rate if diagnosed in an early stage. Magnetic resonance (MR) imaging is an emerging and promising new modality for detection and further evaluation of clinically, mammographically and sonographically occult cancers.

Acquisition of temporal sequences of between three to seven MR images depicting the kinetics of contrast agent molecules in the breast tissue allow for detecting and assessing suspicious tissue disorders with high sensitivity, even in the mammographically dense breasts of young women. Yet, the multitemporal nature of the three-dimensional image data poses new challenges to radiologists as the key-information is only perceivable if all images of the temporal sequence are considered simultaneously. Additionally, radiologists need techniques that allow them not to only analyze a single tumor at a time, but enable them to explore many different tumor lesions at the same time. Concurrent tumor inspection is not feasible since it is inter- and intra-observer dependent and the visualization of the entire signal space is limited by the high-dimensionality of the signal space. In MRI, multiple 3D T1-weighted MR images of both breasts are acquired over a period of 5-9 min while a contrast agent (CA) passes through the tissue. A typical sequence of images consists of one precontrast image acquired before injection of a CA bolus and a series of postcontrast images recorded afterwards over.1 Thus, a time-series signal, i.e., a vector reflecting the local signal intensities at the time points of image acquisition, is associated with each voxel. Due to characteristic changes in the structure of benign and malignant tissue influencing the flux of CA molecules between the blood pool and tissue, characteristic timeseries signals can be observed for different tissue types. Interpretation of these time-series signals allows for detecting cancer with high sensitivity, even in the radioopaque breast of young women, as well as for assessing the type of disorders in a non-invasive fashion .

Spatio-Temporal Classification

However, while the presence of a suspicious tissue disorder can already be identified by means of a strong signal enhancement in an early postcontrast image, the course of the entire timeseries signal has to be considered for differentiating benign and malignant tissue. The computer assisted interpretation of time-series signals as measured during a DCE-MRI examination for each image voxel represents one of the major steps in designing CAD systems for breast MRI. Kuhl et al. have shown that the shape of the time-series signals represents an important criterion in differentiating benign and malignant masses . The results indicate that the enhancement kinetics, as represented by the time-series signals visualized in Fig. 2, differ significantly for benign and malignant enhancing lesions and thus represent a basis for differential diagnosis: plateau or washout-time courses (type II or III) prevail in cancerous tissue. Steadily progressive signal intensity time courses (type I) are exhibited by benign enhancing lesions, albeit these enhancement kinetics are shared not only by benign tumors but also by fibrocystic changes.