Fast and robust segmentation of the striatum using deep convolutional neural networks

Abstract

h i g h l i g h t s• We describe a new method for stratium segmentation.• We employ two serial deep convolutional neural networks (CNN).• Segmentation accuracy of deep CNN is comparable with that of previous methods.• The processing time of the new method is much faster than previous methods. a r t i c l e i n f o a b s t r a c tBackground: Automated segmentation of brain structures is an important task in structural and functional image analysis. We developed a fast and accurate method for the striatum segmentation using deep convolutional neural networks (CNN). New method: T1 magnetic resonance (MR) images were used for our CNN-based segmentation, which require neither image feature extraction nor nonlinear transformation. We employed two serial CNN, Global and Local CNN: The Global CNN determined approximate locations of the striatum. It performed a regression of input MR images fitted to smoothed segmentation maps of the striatum. From the output volume of Global CNN, cropped MR volumes which included the striatum were extracted. The cropped MR volumes and the output volumes of Global CNN were used for inputs of Local CNN. Local CNN predicted the accurate label of all voxels. Segmentation results were compared with a widely used segmentation method, FreeSurfer. Results: Our method showed higher Dice Similarity Coefficient (DSC) (0.893 ± 0.017 vs. 0.786 ± 0.015) and precision score (0.905 ± 0.018 vs. 0.690 ± 0.022) than FreeSurfer-based striatum segmentation (p = 0.06). Our approach was also tested using another independent dataset, which showed high DSC (0.826 ± 0.038) comparable with that of FreeSurfer.Comparison with existing method Segmentation performance of our proposed method was comparable with that of FreeSurfer. The running time of our approach was approximately three seconds. Conclusion: We suggested a fast and accurate deep CNN-based segmentation for small brain structures which can be widely applied to brain image analysis.

Regression‐based machine‐learning approaches to predict task activation using resting‐state fMRI

“…CNNs can take 2D or 3D images as input to train a model. They have been used in MRI for image reconstruction, tissue segmentation, tumor segmentation, and fMRI analysis (Choi & Jin, ; Jang, Plis, Calhoun, & Lee, ; Kamnitsas et al, ; Meszlenyi, Buza, & Vidnyanszky, ; Pham, Ducournau, Fablet, & Rousseau, ; Qin et al, ; Valverde et al, ). CNNs may be worth considering in future studies.…”

Deep learning for brains?: Different linear and nonlinear scaling in UK Biobank brain images vs. machine-learning datasets

“…In domains like computer vision or natural language processing, deep neural networks have already dramatically improved state-of-the-art prediction performances (Goodfellow et al 2016;LeCun et al 2015). However, in the application of neuroimaging data analysis, a similar revolution has not materialized for most common prediction goals, despite considerable research effort and few successful exceptions, such as for the goal of image segmentation (Choi & Jin 2016;Kamnitsas et al 2017;Li et al 2018;Kather et al 2019) and image registration (Balakrishnan et al 2018;Yang et al 2017).…”

Deep Learning in der SPECT und PET des Gehirns

“…CNN können auch für die MRT-basierte Segmentierung einzelner Hirnstrukturen eingesetzt werden, z. B. Hippocampus [30] oder Striatum [31]. Die Rechenzeit liegt dabei im Bereich weniger Sekunden (im Vergleich zu mehreren Stunden mit bisherigen Standardmethoden wie FreeSurfer).…”

“…Die Rechenzeit liegt dabei im Bereich weniger Sekunden (im Vergleich zu mehreren Stunden mit bisherigen Standardmethoden wie FreeSurfer). Damit können diese Verfahren problemlos in der täglichen Routine eingesetzt werden [31]. Die Segmentierung einzelner Hirnstrukturen kann u. a. als Vorbereitung vollständig reproduzierbarer ROI-Auswertung von Hirn-SPECT oder -PET erfolgen.…”