OBJECTIVE: Fluorescence molecular tomography (FMT) is a promising medical imaging technology aimed at the non-invasive, specific, and sensitive detection of the distribution of fluorophore. Conventional sparsity prior-based methods of FMT commonly face problems such as over-sparseness, spatial discontinuity, and poor robustness, due to the neglect of the interrelation within the local subspace. To address this, we propose an adaptive group orthogonal matching pursuit (AGOMP) method. METHODS: AGOMP is based on a novel local spatial-structured sparse regularization, which leverages local spatial interrelations as group sparsity without the hard prior of the tumor region. The adaptive grouped subspace matching pursuit method was adopted to enhance the interrelatedness of elements within a group, which alleviates the over-sparsity problem to some extent and improves the accuracy, robustness, and morphological similarity of FMT reconstruction. A series of numerical simulation experiments, based on digital mouse with both one and several tumors, were conducted, as well as in vivo mouse experiments. RESULTS: The results demonstrated that the proposed AGOMP method achieved better location accuracy, fluorescent yield reconstruction, relative sparsity, and morphology than state-of-the-art methods under complex conditions for levels of Gaussian noise ranging from 5-25%. Furthermore, the in vivo mouse experiments demonstrated the practical application of FMT with AGOMP. CONCLUSION: The proposed AGOMP can improve the accuracy and robustness for FMT reconstruction in biomedical application.
OBJECTIVE: Fluorescence molecular tomography (FMT) is a promising medical imaging technology aimed at the non-invasive, specific, and sensitive detection of the distribution of fluorophore. Conventional sparsity prior-based methods of FMT commonly face problems such as over-sparseness, spatial discontinuity, and poor robustness, due to the neglect of the interrelation within the local subspace. To address this, we propose an adaptive group orthogonal matching pursuit (AGOMP) method. METHODS: AGOMP is based on a novel local spatial-structured sparse regularization, which leverages local spatial interrelations as group sparsity without the hard prior of the tumor region. The adaptive grouped subspace matching pursuit method was adopted to enhance the interrelatedness of elements within a group, which alleviates the over-sparsity problem to some extent and improves the accuracy, robustness, and morphological similarity of FMT reconstruction. A series of numerical simulation experiments, based on digital mouse with both one and several tumors, were conducted, as well as in vivo mouse experiments. RESULTS: The results demonstrated that the proposed AGOMP method achieved better location accuracy, fluorescent yield reconstruction, relative sparsity, and morphology than state-of-the-art methods under complex conditions for levels of Gaussian noise ranging from 5-25%. Furthermore, the in vivo mouse experiments demonstrated the practical application of FMT with AGOMP. CONCLUSION: The proposed AGOMP can improve the accuracy and robustness for FMT reconstruction in biomedical application.