Effects of Enviromental Exposure to Organolead on the Brain
Contact
People
- Dinggang Shen, Ragini Verma , Xiaoying Wu, Chandan Misra
Collaborators
The goal of this project is to develop a computed aided diagnostic (CAD) system for supporting the detection and classification of breast cancer and improving its specificity, using contrast-enhanced magnetic resonance (ce-MR) images. Although current clinical CAD systems provide visualization tools, they do not provide advanced detection or characterization tools. While various CAD approaches have been developed in the literature to support cancer characterization, they are limited by requiring manual detection of potentially suspicious enhancement by radiologists, and extracting simple morphological features from only a single scan such as the first post-contrast scan, which certainly ignores the evolving pattern of the tumor architecture over time.
We propose a detection module for detecting potentially suspicious enhancement in breast MRI, and a classification module for differentiating benign from malignant contrast enhancement.
In the following, the current results on measurement of breast parenchyma, registration of dynamic contrast-enhanced breast MRI, and classification of breast tumor are provided.
Breast Parenchyma Measurement [1]
Breast density has been shown to be an independent risk factor for breast cancer. To segment breast parenchyma which has been proposed as a biomarker of breast cancer risk, we have developed an integrated algorithm for simultaneous T1 map estimation and segmentation, using a series of magnetic resonance (MR) breast images (such as five series of 3D images acquired by using five different sets of inversion time Ti and repetition time TR, i.e., {(Tim, TRm), m=1,…,5}={(1600,1880), (800,1080), (400,680), (200,480), (140,420)} in a unit of ms.). The advantage of using this algorithm is that the step of T1 map estimation (E-Step) and the step of T1 map based tissue segmentation (S-Step) can benefit each other. Since the estimated T1 map can be noisy due to the complexity of T1 estimation method, the tentative tissue segmentation results from S-Step can help perform the edge-preserving smoothing on the estimated T1 map in E-Step, thus removing noise and also preserving tissue boundaries. On the other hand, the improved estimation of T1 map from E-Step can help segment breast tissues in a more accurate and less noisy way. Therefore, by repeating these two steps, it is possible to simultaneously obtain better results for both T1 map estimation and segmentation.
Robust registration [2]
Dynamic enhancement causes serious problems for registration of contrast enhanced breast MRI, due to variable uptakes of agent on different tissues or even same tissues in the breast. We present an iterative optimization algorithm to de-enhance the dynamic contrast-enhanced breast MRI and then register them for avoiding the effects of enhancement on image registration. In particular, the spatially varying enhancements are modeled by a Markov Random Field, and estimated by a locally smooth function with boundaries using a graph cut algorithm. The de-enhanced images are then registered by conventional B-spline based registration algorithm. These two steps benefit from each other and are repeated until the results converge. Experimental results show that our two-step registration algorithm performs much better than conventional mutual information based registration algorithm. Also, the effects of tumor shrinking in the conventional registration algorithms can be effectively avoided by our registration algorithm.
MR-based Breast Tumor Diagnosis [3,4]
We have tested the utility of a computer analysis method, named Spatial-Temporal Enhancement Pattern (STEP), in characterizing breast cancer in contrast-enhanced MR images. STEP captures not only the dynamic enhancement and architectural features of tumor, but also the spatial variations of pixel-wise temporal enhancement of tumor. Although the latter has been widely used by radiologists during diagnosis, it is rarely considered as important feature for computer-aided diagnosis. One of the main premises in STEP is that, by jointly regarding serial contrast-enhanced images as a single spatial-temporal image, it can potentially capture both spatial and temporal image features within a single framework; correlations between spatial and temporal features are also considered in this methodological framework.
To compare the performances of the STEP features with general dynamic and architectural features in tumor diagnosis, we selected some existing features which were shown effective in the literature. STEP features performed best in all experiments. The AUC value, classification accuracy, sensitivity and specificity all show significant improvement.