An unsupervised Kalman filter-based linear mixing approach to MRI classification

Kalman filter-based linear mixing method (KFLM) approach has shown success in magnetic resonance image classification. It employs an auxiliary equation, called abundance state equation (ASE), to trace the signature abundance. The signature abundance can be estimated and updated recursively by the Kalman filter and an abrupt change in signature abundance can be detected via the abundance state equation. Therefore, it requires a complete knowledge of the desired target signature and the unwanted signatures present in the images. In this paper, an unsupervised Kalman filter-based linear mixing method (UKFLM) approach is presented, this didn't know how many target signatures were present in the image and where are these signatures. UKFLM comprises two processes, target generation process (TCP) and target classification process (TCP). The objective of TCP is to generate a set of potential target signatures from an unknown background, which can be subsequently classified by TCP. As a result, UKFLM can be used to search for a specific target in unknown scenes. Finally, the effectiveness of UKFLM in target detection and classification is evaluated by several MR images experiments. The method has been evaluated through several experiments. All experiments were under supervision of the expert radiologist. Results show that the UKFLM have the capability for multispectral images segmentation and robustness to the noise, indicating the possible usefulness of this method.