Incorporation of a spectral model in a convolutional neural network for accelerated spectral fitting.

摘要: Purpose MRSI has shown great promise in the detection and monitoring of neurologic pathologies such as tumor. A necessary component data processing includes quantitation each metabolite, typically done through fitting a model spectrum to data. For high-resolution volumetric brain, which may have ~10,000 spectra, significant time is required for spectral analysis generation metabolite maps. Methods novel unsupervised deep learning architecture that combines convolutional neural network with priori models presented. This architecture, encoder-model decoder (CEMD), strengths adaptive unbiased networks magnetic resonance readily interpretable. Results The CEMD performs accurate patients glioblastoma, provides whole-brain 1 min on standard computer, handles variety artifacts. Conclusion new combining physics domain knowledge been developed able perform rapid Rapid critical step toward routine clinical practice.