Detecting Giant Cell Tumor of Bone Lesions Using Mueller Matrix Polarization Microscopic Imaging and Multi-Parameters Fusion Network

Giant cell tumor of bone (GCTB) is potentially malignant with a high risk of recurrence. Conventional GCTB diagnosis is often empirical based on histopathology with low efficiency. Reliable auxiliary diagnosis system can improve the efficiency, but it need accurate GCTB detection. Changes of micro-structure of GCTB tissue can alter the scattering properties that lead to the variation of the polarization state of GCTB tissue. This paper proposed a GCTB detection method by using Mueller matrix polarization microscopic (MMPM) imaging and multi-parameters fusion network. First, a MMPM image dataset of GCTB tissue is set up, and based on this dataset the effectiveness of MMPM imaging in the GCTB detection is verified for the first time. Then a multi-parameters fusion network(MPFN) model is proposed to fuse the deep features and hand craft features, which leverages the advantages of both hand-crafted features and deep learning based systems. Experiment results show that the MPFN outperformed state-of-the-art for GCTB lesions detection.