A novel network architecture combining central-peripheral deviation with image-based convolutional neural networks for diffusion tensor imaging studies

Abstract

Brain imaging research is a very challenging topic due to complex structure and lack of explicitly identifiable features in the image. With the advancement of magnetic resonance imaging (MRI) technologies, such as diffusion tensor imaging (DTI), developing classification methods to improve clinical diagnosis is crucial. This paper proposes a classification method for DTI data based on a novel neural network strategy that combines a convolutional neural network (CNN) with a multilayer neural network using central-peripheral deviation (CPD), which reflects diffusion dynamics in the white matter by spatially evaluating the deviation of diffusion coefficients between the inner and outer parts of the brain. In our method, a multilayer perceptron (MLP) using CPD is combined with the final layers for classification after reducing the dimensions of images in the convolutional layers of the neural network architecture. In terms of training loss and the classification error, the proposed classification method improves the existing image classification with CNN. For real data analysis, we demonstrate how to process raw DTI image data sets obtained from a traumatic brain injury study (MagNeTS) and a brain atlas construction study (ICBM), and apply the proposed approach to the data, successfully improving classification performance with two age groups.

Keywords: Concentric circle pooling; convolutional neural network; diffusion tensor image; image classification; multi-layer perceptron.

Figure 1.

An example of FA image from MagNeTS data in RGB color-coded format. The image dimension is $310\times 380$ pixels. The value of FA is between 0–1. The values are color coded as dark blue (lowest), sky blue, green-yellow, orange and yellow (highest). The white has no value (background) or missing (middle).

Figure 2.

Feature extraction example. The first step (first arrow) shows how to convolute an image with a convolutional filter ( $3\times 3$ size). The values in the convolutional filter are applied to the image as ‘weights’ and multiplied by the values at the same location in the receptive field of $3\times 3$ size. The second step (second arrow) describes a max pooling procedure using the $2\times 2$ pooling filter, where the maximum values in the 2 × 2 grid is selected in the convoluted image. The values in the image ranges from 0 (black) to 255 (white). The red circles represent the edge of the strawberry in the center of the cheesecake.

Figure 3.

Examples of FA image in the age groups from MagNeTS. Left: Subjects with age older than and equal to 52. Right: Subjects age younger than 52.

Figure 4.

The proposed network architecture using CPD. The upper layers describe the CNN and the lower layers describe the MLP with 4 layers. The smaller the length of bar indicates the smaller size of the images. The number on top of each bar represents the number of nodes at the layer. The arrows with different colors represent different procedure as shown in the legend.

Figure 5.

The proposed network architecture using CPD. The upper layers describe the CNN and the lower layers describe the MLP with 2 layers. The smaller the length of bar indicates the smaller size of the images. The number on top of each bar represents the number of nodes at the layer. The arrows with different colors represent different procedure as shown in the legend.