Breast MRI data analysis by LLE

• Published in: 2004 IEEE International Joint Conference on Neural Networks (IEEE Cat. No.04CH37541) Vol. 3; pp. 2449 - 2454 vol.3
• Main Authors: Varini, C., Nattkemper, T.W., Degenhard, A., Wismuller, A.
• Format: Conference Proceeding
• Language: English
• Published: Piscataway NJ IEEE 2004
• Subjects: Applied sciences; Artificial intelligence; Benign tumors; Biomedical imaging; Breast cancer; Cancer detection; Computer science; control theory; systems; Data analysis; Exact sciences and technology; Magnetic resonance imaging; Malignant tumors; Mammography; Pattern recognition. Digital image processing. Computational geometry; Physics; Protocols; Human; Data analysis; Image processing; Image databank; Time series; Neural network; Malignant tumor; Nuclear magnetic resonance imaging; Unsupervised learning; Data reduction; Heterogeneity; Cancerology; Image sequence; Imaging; Classification; Mammary gland; Data structure; Learning algorithm; Artificial intelligence; Medical application
• ISBN: 0780383591; 9780780383593
• ISSN: 1098-7576
• DOI: 10.1109/IJCNN.2004.1381012

Locally linear embedding (LLE) has recently been proposed as a powerful algorithm for unsupervised learning and dimensional data reduction. For a first time we apply LLE to a problem of medical data analysis. Magnetic resonance imaging (MRI) is considered as an essential imaging modality in the detection and classification of breast cancer. In dynamic contrast enhanced MRI (DCE-MRI) the data set of each patient is composed of a sequence of images and each data point in the image is associated with one time-series feature vector. Our results show that LLE is capable of revealing the heterogeneity of malignant tumors from the data structure of DCE-MRI signals.