In this section, the framework of our multi-dimensional cascaded net and the structures of each part are described in detail. This study is committed to building a more efficient neural network structure to complete the automatic classification of brain tumor in MRI images. U-net is found to be a more effective approach as compared to other network architectures. U-net, evolved from the traditional CNN, was first designed and applied in 2015 to process biomedical images. As a general convolutional neural network, it focuses on image classification, where input is an image and output is a one label, but in biomedical imaging, it requires us not only to distinguish whether there is a disease or not, but also to localize the area of abnormality. U-net provides fast and precise image segmentation and is dedicated in solving this problem as it is able to localize and distinguish borders by doing classification on every pixel, so that the input and output share the same size. This work emphasizes on the classification module of the CAD system, which presents a deep learning-based UNet for automated detection of cancer in MRI images. We propose an efficient attention network with nested connections, as shown in Fig. 1. This architecture is based on the popular UNet, which is designed to work well with a small number of training samples. The network is an encoder-decoder structure.
3.1. NESTED CONNECTION
The success of UNet depends largely on skip connections. The connection combines the coarse-grained features of the decoder with the fine-grained features of the encoder by adding elements. Zhou et al(2020) redesigned the skip connection of UNet and proposed a new model Light weight UNet (LUNet). It uses nested connection instead of the original simple connection, which can capture more abundant features of multi-level, and then integrate the multi-level features by adding elements. In the case of nested connection, the encoder is not directly connected with the decoder after getting the final aggregation feature mapping. Instead, the nested connection is accessed first, so that the rich features captured earlier can be better preserved.
Let’s take a closer look at nested connection. Let x i,j denote the output value of node Xi,j, index i denote the maximum pooling operation from the encoder, and index j denote the transpose convolution operation from the decoder. The formula of the concatenated feature maps is expressed as
![](data:image/png;base64,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)
Where function F (. ) consists of four operations, two convolution layers M (. ) and T (. ) two DropBlock [14] layers arranged alternately, and [ ] represent the Max-pooling layer and transpose convolution layer respectively, and denotes the concatenation layer. When node i = 0 and j = 0, the function F (. ) is executed on the input original image. When nodes i > 0 and j = 0 indicate that there is only one input in the front layer. When nodes i ≥ 0 and j = 0, they will receive j + 1 inputs. Take decoder node X0,2 as an example, it has three inputs, which are the previous decoder node X1,1, the adjacent decoder nested node X0,1 and the skip connection node X0,0. The collection of multiple feature maps will provide richer semantic information for the final segmentation.
3.2. EFFICIENT ATTENTION
Since Vaswani et al (2017) proposed a new structure that connects encoders and decoders through Self-Attention (SA), which has achieved great success in machine translation tasks. Subsequently, many works have applied SA to computer vision tasks and achieved good results. The Squeeze Excitation (SE) (Hu et al., 2020) module starts from the perspective of channels and constructs information features by fusing the channel information of the local receptive fields of each layer. The SE module significantly improves the performance of the current most advanced Convolutional Neural Network (CNN) while slightly increasing the computational cost. Fu et al(2019) proposed a dual attention (DA) network that combines spatial attention and channel attention for adaptive fusion of local and global features. The best performance is achieved in scene segmentation tasks.
Wang et al.(2020) found through in-depth research on channel attention that appropriate cross-channel interaction is very important for learning more effective feature mapping. Local cross-channel interaction can be achieved through one- dimensional convolution, and nonlinear mapping is used to adaptively determine the size of the convolution kernel. The Efficient Attention (EA) module can be flexibly put into the existing CNN, and its structural details are shown in Fig. 3.
The function of efficient attention module is to obtain more abundant channel information. Group convolution is a method to manually adjust the size of convolution kernel according to the number of channels. However, manual adjustment consumes too much resources, so an adaptive adjustment strategy is designed. There is a mapping θ(•) between the size k of convolution kernel and the channel dimension C:
C = θ(k)
The channel dimension C is usually set to a power of 2. Therefore, a nonlinear function θ(•) is designed:
θ(k) = 2(α∗k − β)
When the channel dimension C is given, the convolution kernel size k can be calculated adaptively by
![](data:image/png;base64,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)
Where function O(. ) takes the nearest odd number. The later experiment, we set α and β to 2 and 1 respectively. In this way, higher dimensional channel will get larger convolution kernel size, while low dimensional channel can get smaller convolution kernel size through nonlinear mapping.
3.3. DropBlock regularization
In 2012, the Hinton team proposed an effective regularization method called Dropout (Srivastava et al., 2014) which can be used to prevent overfitting. Dropout is widely used in fully connected layers, but it is usually not effective for convolutional layers. The reason is probably because the adjacent elements in the feature map of the convolutional layer share semantic information in space, so although a unit is discarded, the adjacent elements can still retain the semantic information of the position, and the information can still be Circulate in convolutional networks. The mainstream network model that solves the segmentation problem just does not include a fully connected layer, so Google Brain proposed a DropBlock (Ghiasi et al., 2018) regularization method for this situation, which is a simple method similar to Dropout.
The main difference with Dropout is that it removes adjacent regions from the feature map of a layer instead of discarding independent random units. The first parameter µ of DropBlock is the size of the block to be dropped, and the second parameter γ controls the number of active units to be dropped. Similar to Dropout, we do not apply DropBlock in the inference process. In the experiment, we set a constant for all feature maps, regardless of the resolution of feature maps. When µ = 1, DropBlock and Dropout are equivalent, when covering the entire feature map. For the setting of the value of γ, we assume that we want each activation unit to maintain a probability of pkeep. γ can be calculated as:
![](data:image/png;base64,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)
In order to ensure that the µ2 region is contained in the λ2 region, we adjust the initial binary mask when sampling, so that the size of the effective seed region is (λ-µ + 1)2, and set p keep as 0.9 and µ as 7.
3.4. FUSION LOSS FUNCTION
The most common loss function used for image semantic segmentation tasks is pixel-level cross-entropy loss. This loss function discriminates each pixel, and then compares the result with the one-hot encoding vector. Binary Cross Entropy (BCE) loss is for the case where there are only two categories, and its loss function formula is:
![](data:image/png;base64,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)
Since the cross-entropy loss evaluates the category prediction of each pixel separately, and then averages the loss of all pixels, we essentially learn equally for each pixel in the image. If the distribution of multiple classes in the image is unbalanced, then this may lead to the training process being dominated by classes with a large number of pixels. The model will mainly learn the features of the large number of class samples, and the learned model will be more biased to predict the pixels for this category.
In order to overcome the problem that the foreground region is difficult to detect completely. Milletari et al(2016) proposed a new loss function based on dice coefficient (Dice). The expression of the dice coefficient D is
![](data:image/png;base64,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)
Where pi = P represents the predicted binary vector and gi ∈ G represents the ground truth binary vector. The total number of pixels is N.
![](data:image/png;base64,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)
Using this formula, we can establish an appropriate balance without assigning different weights to the parts of interest and background. In the fundus blood vessel segmentation task, we fuse the two loss functions of BCE loss and Dice loss by simple addition, which we call the fusion loss function.