RESUMO
OBJECTIVES: Our objectives were to provide an automated method for spatially resolved detection and quantification of motion artifacts in MR images of the head and abdomen as well as a quality control of the trained architecture. MATERIALS AND METHODS: T1-weighted MR images of the head and the upper abdomen were acquired in 16 healthy volunteers under rest and under motion. Images were divided into overlapping patches of different sizes achieving spatial separation. Using these patches as input data, a convolutional neural network (CNN) was trained to derive probability maps for the presence of motion artifacts. A deep visualization offers a human-interpretable quality control of the trained CNN. Results were visually assessed on probability maps and as classification accuracy on a per-patch, per-slice and per-volunteer basis. RESULTS: On visual assessment, a clear difference of probability maps was observed between data sets with and without motion. The overall accuracy of motion detection on a per-patch/per-volunteer basis reached 97%/100% in the head and 75%/100% in the abdomen, respectively. CONCLUSION: Automated detection of motion artifacts in MRI is feasible with good accuracy in the head and abdomen. The proposed method provides quantification and localization of artifacts as well as a visualization of the learned content. It may be extended to other anatomic areas and used for quality assurance of MR images.
Assuntos
Processamento de Imagem Assistida por Computador/métodos , Imageamento por Ressonância Magnética , Abdome/diagnóstico por imagem , Algoritmos , Artefatos , Automação , Processamento Eletrônico de Dados , Cabeça/diagnóstico por imagem , Humanos , Imageamento Tridimensional , Aprendizado de Máquina , Movimento (Física) , Redes Neurais de Computação , Probabilidade , Garantia da Qualidade dos Cuidados de Saúde , Reprodutibilidade dos Testes , Processamento de Sinais Assistido por ComputadorRESUMO
PURPOSE: Lipid signals measured by (1)H MR spectroscopy cannot be adequately quantified by common fitting routines like VARPRO or AMARES, if lipid spectra are distorted by irregular spatial and temporal inhomogeneities of the static magnetic field during readout. A fully automatic reference deconvolution algorithm is presented that eliminates these distortions before application of fitting routines. METHODS: The measured signal of the dominant methyl resonance is isolated with aid of a spectral estimator (estimation of parameters via rotational invariance techniques) and used as reference signal for estimation of distortions. A Wiener filter is applied to deconvolve those distortions in the lipid spectrum. Performance of the algorithm is assessed for different bandwidths and shapes of distortions, using artificially distorted as well as measured data. RESULTS: Application of the fully automatic reference deconvolution algorithm on simulated spectra yields a distinct increase in quantification accuracy. Deconvolved in vivo spectra of subcutaneous fat indicate reduced spectral overlap after application of the proposed strategy. CONCLUSION: The proposed method is helpful for in vivo magnetic resonance spectroscopy of adipose tissue to correct for effects of field inhomogeneities within the voxel and for inevitable eddy current effects. Quantification accuracy is improved by eliminating distortions before application of fitting routines.