Modern-era developments in authentication systems have changed from traditional methods based on passwords or signatures to new methods based on biometric patterns. Biometric patterns are unique to each person, and identifying individuals has become much more accurate. Biometric cognition uses an intelligent method to identify a person with some unique characteristics of a human being. Unlike traditional methods, these biometric methods are more reliable and safer. Diagnosing blood patterns of retinal images is one of the safest ways to authenticate taking into consideration the monopoly nature of these patterns for each individual and their non-reproducibility and alteration. In the present study, the Convolutional Neural network (CNN) was used to identify the pattern of blood vessels in the retina. DRIVE dataset was used to evaluate results. The images of the retina of different people were stored in this dataset. After extracting the patterns within the retinal layers for each person as a model indicating the identity of these individuals, the patterns related to the training and testing datasets were compared to determine the identity of individuals.  Properly tested samples increase the accuracy of the proposed method, while incorrect detection will cause an error in the proposed method. The results showed that the average accuracy of matching blood vessel patterns for retinal images in the proposed method was 94.83%, which is high and comparable to previous methods.

Accurate liver segmentation on Magnetic Resonance Images (MRI) is a challenging task especially at sites where surrounding tissues such as spleen and kidney have densities similar to that of the liver and lesions reside at the liver edges. The first and essential step for computer aided diagnosis (CAD) is the automatic liver segmentation that is still an open problem. Extensive research has been performed for liver segmentation; however it is still challenging to distinguish which algorithm produces more precise segmentation results to various medical images. In this paper, we have presented a new automatic system for liver segmentation in abdominal MRI images. Our method extracts liver regions based on several successive steps. The preprocessing stage is applied for image enhancement such as edge preserved and noise reduction. The proposed algorithm for liver segmentation is a combined algorithm which utilizes a contour algorithm with a Vector Field Convolution (VFC) field as its external force and perceptron Neural network.By convolving the edge map generated from the image with the user-defined vector field kernel, VFC is calculated.We use trained Neural networks to extract some features from liver region. The extracted features are used to find initial point for starting VFC algorithm. This system was applied to a series of test images to extract liver region.Experimental results showed the promise of the proposed algorithm.

Recent researches on pixel-wise semantic segmentation use deep Neural networks to improve accuracy and speed of these networks in order to increase the efficiency in practical applications such as automatic driving. These approaches have used deep architecture to predict pixel tags, but the obtained results seem to be undesirable. The reason for these unacceptable results is mainly due to the existence of max pooling operators, which reduces the resolution of the feature maps. In this paper, we present a Convolutional Neural network composed of encoder-decoder segments based on successful SegNet network. The encoder section has a depth of 2, which in the first part has 5 Convolutional layers, in which each layer has 64 filters with dimensions of 3×3. In the decoding section, the dimensions of the decoding filters are adjusted according to the Convolutions used at each step of the encoding. So, at each step, 64 filters with the size of 3×3 are used for coding where the weights of these filters are adjusted by network training and adapted to the educational data. Due to having the low depth of 2, and the low number of parameters in proposed network, the speed and the accuracy improve compared to the popular networks such as SegNet and DeepLab. For the CamVid dataset, after a total of 60, 000 iterations, we obtain the 91% for global accuracy, which indicates improvements in the efficiency of proposed method.

Skin detection is a useful and popular method to identify and recognize human body parts, faces, naked people, and skin diseases and retrieve people in multimedia databases. Therefore, finding a suitable method to divide the pixels of an image into different skin groups can be very important. One of the most common methods is based on Convolutional Neural networks (CNN). However, the process of training a CNN is a challenging issue. Various optimization strategies have recently been used to optimize CNN biases and weights, such as the firefly algorithm (FA) and ant colony optimization (ACO). In this study, we use a well-known nature-inspired technique called Chimp optimization algorithm (ChOA) to train a classical LeNet-5 CNN structure for skin detection. The proposed skin classification algorithms operate directly on the RGB and HVS color space. The results clearly show that the proposed algorithm significantly improves the performance of a Convolutional Neural network.