Diabetic retinopathy (DR) is the most common diabetic eye disease worldwide and a leading cause of blindness. The number of diabetic patients will increase to 552 million by 2034, as per the International Diabetes Federation (IDF). With advances in computer science techniques, such as artificial intelligence (AI) and deep learning (DL), opportunities for the detection of DR at the early stages have increased. This increase means that the chances of recovery will increase and the possibility of vision loss in patients will be reduced in the future. In this paper, deep transfer learning models for medical DR detection were investigated. The DL models were trained and tested over the Asia Pacific Tele-Ophthalmology Society (APTOS) 2019 dataset. According to literature surveys, this research is considered one the first studies to use of the APTOS 2019 dataset, as it was freshly published in the second quarter of 2019. The selected deep transfer models in this research were AlexNet, Res-Net18, SqueezeNet, GoogleNet, VGG16, and VGG19. These models were selected, as they consist of a small number of layers when compared to larger models, such as DenseNet and InceptionResNet. Data augmentation techniques were used to render the models more robust and to overcome the overfitting problem. The testing accuracy and performance metrics, such as the precision, recall, and F1 score, were calculated to prove the robustness of the selected models. The AlexNet model achieved the highest testing accuracy at 97.9%. In addition, the achieved performance metrics strengthened our achieved results. Moreover, AlexNet has a minimum number of layers, which decreases the training time and the computational complexity.

Cancer is one of the most feared and aggressive diseases in the world and is responsible for more than 9 million deaths universally. Staging cancer early increases the chances of recovery. One staging technique is RNA sequence analysis. Recent advances in the efficiency and accuracy of artificial intelligence techniques and optimization algorithms have facilitated the analysis of human genomics. This paper introduces a novel optimized deep learning approach based on binary particle swarm optimization with decision tree (BPSO-DT) and convolutional neural network (CNN) to classify different types of cancer based on tumor RNA sequence (RNA-Seq) gene expression data. The cancer types that will be investigated in this research are kidney renal clear cell carcinoma (KIRC), breast invasive carcinoma (BRCA), lung squamous cell carcinoma (LUSC), lung adenocarcinoma (LUAD) and uterine corpus endometrial carcinoma (UCEC). The proposed approach consists of three phases. The first phase is preprocessing, which at first optimize the high-dimensional RNA-seq to select only optimal features using BPSO-DT and then, converts the optimized RNA-Seq to 2D images. The second phase is augmentation, which increases the original dataset of 2086 samples to be 5 times larger. The selection of the augmentations techniques was based achieving the least impact on manipulating the features of the images. This phase helps to overcome the overfitting problem and trains the model to achieve better accuracy. The third phase is deep CNN architecture. In this phase, an architecture of two main convolutional layers for featured extraction and two fully connected layers is introduced to classify the 5 different types of cancer according to the availability of images on the dataset. The results and the performance metrics such as recall, precision and F1 score show that the proposed approach achieved an overall testing accuracy of 96.90%. The comparative results are introduced, and the proposed method outperforms those in related works in terms of testing accuracy for 5 classes of cancer. Moreover, the proposed approach is less complex and consume less memory.

Early detection of cancer increases the probability of recovery. This paper presents an intelligent decision support system (IDSS) for the early diagnosis of cancer based on gene expression profiles collected using DNA microarrays. Such datasets pose a challenge because of the small number of samples (no more than a few hundred) relative to the large number of genes (in the order of thousands). Therefore, a method of reducing the number of features (genes) that are not relevant to the disease of interest is necessary to avoid overfitting. The proposed methodology uses the information gain (IG) to select the most important features from the input patterns. Then, the selected features (genes) are reduced by applying the grey wolf optimization (GWO) algorithm. Finally, the methodology employs a support vector machine (SVM) classifier for cancer type classification. The proposed methodology was applied to two datasets (Breast and Colon) and was evaluated based on its classification accuracy, which is the most important performance measure in disease diagnosis. The experimental results indicate that the proposed methodology is able to enhance the stability of the classification accuracy as well as the feature selection.

One attractive research area in the computer science field is soft biometrics. To Identify a person's gender from an iris image when such identification is related to security surveillance systems and forensics applications. In this paper, a robust iris gender-identification method based on a deep convolutional neural network is introduced. The proposed architecture segments the iris from a background image using the graph-cut segmentation technique. The proposed model contains 16 subsequent layers; three are convolutional layers for feature extraction with different convolution window sizes, followed by three fully connected layers for classification. The original dataset consists of 3,000 images, 1,500 images for men and 1,500 images for women. The augmentation techniques adopted in this research overcome the overfitting problem and make the proposed architecture more robust and immune from simply memorizing the training data. In addition, the augmentation process not only increased the number of dataset images to 9,000 images for the training phase, 3,000 images for the testing phase and 3,000 images for the verification phase but also led to a significant improvement in testing accuracy, where the proposed architecture achieved 98.88%. A comparison is presented in which the testing accuracy of the proposed approach was compared with the testing accuracy of other related works using the same dataset. The proposed architecture outperformed the other related works in terms of testing accuracy.

This paper is an extended version of "Deep Galaxy: Classification of Galaxies based on Deep Convolutional Neural Networks". In this paper, a robust deep convolutional neural network architecture for galaxy morphology classification is presented. A galaxy can be classified based on its features into one of three categories (Elliptical, Spiral, or Irregular) according to the Hubble galaxy morphology classification from 1926. The proposed convolutional neural network architecture consists of 8 layers, including one main convolutional layer for feature ex-traction with 96 filters and two principle fully connected layers for classification. The architecture is trained over 4238 images and achieved a 97.772% testing accuracy. In this version, "Deep Galaxy V2", an augmentation process is applied to the training data to overcome the overfitting problem and make the proposed architecture more robust and immune to memorizing the training data. A comparative result is present, and the testing accuracy was compared with those of other related works. The proposed architecture outperformed the other related works in terms of its testing accuracy.