Redha Ali has a total of 13 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    IMNets: Deep Learning Using an Incremental Modular Network Synthesis Approach for Medical Imaging Applications

    by Redha Ali, Russell C. Hardie, Barath Narayanan Narayanan, Temesguen M. Kebede

    Deep learning approaches play a crucial role in computer-aided diagnosis systems to support clinical decision-making. However, developing such automated solutions is challenging due to the limited availability of annotated medical data. In this study, we proposed a novel and computationally efficient deep learning approach to leverage small data for learning generalizable and domain invariant representations in different medical imaging applications such as malaria, diabetic retinopathy, and tuberculosis. We refer to our approach as Incremental Modular Network Synthesis (IMNS), and the resulting CNNs as Incremental Modular Networks (IMNets). Our IMNS approach is to use small network modules that we call SubNets which are capable of generating salient features for a particular problem. Then, we build up ever larger and more powerful networks by combining these SubNets in different configurations. At each stage, only one new SubNet module undergoes learning updates. This reduces the computational resource requirements for training and aids in network optimization. We compare IMNets against classic and state-of-the-art deep learning architectures such as AlexNet, ResNet-50, Inception v3, DenseNet-201, and NasNet for the various experiments conducted in this study. Our proposed IMNS design leads to high average classification accuracies of 97.0%, 97.9%, and 88.6% for malaria, diabetic retinopathy, and tuberculosis, respectively. Our modular design for deep learning achieves the state-of-the-art performance in the scenarios tested. The IMNets produced here have a relatively low computational complexity compared to traditional deep learning architectures. The largest IMNet tested here has 0.95 M of the learnable parameters and 0.08 G of the floating-point multiply–add (MAdd) operations. The simpler IMNets train faster, have lower memory requirements, and process images faster than the benchmark methods tested.


  • Open Access

    Fused Deep Convolutional Neural Network for Precision Diagnosis of COVID-19 Using Chest X-Ray Images

    by Hussin K. Ragb, Ian T. Dover, Redha Ali

    With a Coronavirus disease (COVID-19) case count exceeding 10 million worldwide, there is an increased need for a diagnostic capability. The main variables in increasing diagnostic capability are reduced cost, turnaround or diagnosis time, and upfront equipment cost and accessibility. Two candidates for machine learning COVID-19 diagnosis are Computed Tomography (CT) scans and plain chest X-rays. While CT scans score higher in sensitivity, they have a higher cost, maintenance requirement, and turnaround time as compared to plain chest X-rays. The use of portable chest X-radiograph (CXR) is recommended by the American College of Radiology (ACR) since using CT places a massive burden on radiology services. Therefore, X-ray imagery paired with machine learning techniques is proposed a first-line triage tool for COVID-19 diagnostics. In this paper we propose a computer-aided diagnosis (CAD) to accurately classify chest X-ray scans of COVID-19 and normal subjects by fine-tuning several neural networks (ResNet18, ResNet50, DenseNet201) pre-trained on the ImageNet dataset. These neural networks are fused in a parallel architecture and the voting criteria are applied in the final classification decision between the candidate object classes where the output of each neural network is representing a single vote. Several experiments are conducted on the weakly labeled COVID-19-CT-CXR dataset consisting of 263 COVID-19 CXR images extracted from PubMed Central Open Access subsets combined with 25 normal classification CXR images. These experiments show an optimistic result and a capability of the proposed model to outperforming many state-of-the-art algorithms on several measures. Using k-fold cross-validation and a bagging classifier ensemble, we achieve an accuracy of 99.7% and a sensitivity of 100%.

  • Performance analysis of machine learning and deep learning architectures for malaria detection on cell images

    by Barath Narayanan Narayanan, Redha A. Ali, Russell C. Hardie

    Plasmodium malaria is a parasitic protozoan that causes malaria in humans. Computer aided detection of Plasmodium is a research area attracting great interest. In this paper, we study the performance of various machine learning and deep learning approaches for the detection of Plasmodium on cell images from digital microscopy. We make use of a publicly available dataset composed of 27,558 cell images with equal instances of parasitized (contains Plasmodium) and uninfected (no Plasmodium) cells. We randomly split the dataset into groups of 80% and 20% for training and testing purposes, respectively. We apply color constancy and spatially resample all images to a particular size depending on the classification architecture implemented. We propose a fast Convolutional Neural Network (CNN) architecture for the classification of cell images. We also study and compare the performance of transfer learning algorithms developed based on well-established network architectures such as AlexNet, ResNet, VGG-16 and DenseNet. In addition, we study the performance of the bag-of-features model with Support Vector Machine for classification. The overall probability of a cell image comprising Plasmodium is determined based on the average of probabilities provided by all the CNN architectures implemented in this paper. Our proposed algorithm provided an overall accuracy of 96.7% on the testing dataset and area under the Receiver Operating Characteristic (ROC) curve value of 0.994 for 2756 parasitized cell images. This type of automated classification of cell images would enhance the workflow of microscopists and provide a valuable second opinion.


  • Deep Learning Ensemble Methods for Skin Lesion Analysis towards Melanoma Detection

    by Redha Ali, Russell C. Hardie, Barath Narayanan Narayanan, Supun De Silva

    Skin cancer has a significant impact across the world. Melanoma is a malignant form of skin cancer. Skin lesion segmentation is an important step in computer-aided diagnosis (CAD) for automated diagnosis of melanoma. In this paper, we describe our research work and the submission to the International Skin Imaging Collaborations (ISIC) 2018 Challenge in Skin Lesion Analysis Towards Melanoma Detection. We propose Convolutional Neural Network (CNN) based ensemble methods for improving the existing performance of lesion segmentation. The proposed ensemble technique includes VGG19-UNet, DeeplabV3+ and other preprocessing methodologies. Extensive experiments are conducted on the ISIC 2018 challenge dataset to demonstrate the efficacy of the proposed model. For evaluation, we utilize the ISIC 2018 datasets that contains 2,594 dermoscopy images with their ground truth segmentation masks. We randomly divided the dataset into 80% for training and 20% for validation. Our proposed model provided an overall accuracy of 93.6%, average Jaccard Index of 0.815, and dice coefficient of 0.887 on the testing dataset.


  • Convolutional Neural Network for Classification of Histopathology Images for Breast Cancer Detection

    by Barath Narayanan Narayanan, Vignesh Krishnaraja, Redha Ali

    One of the most common subtypes of all breast cancers is Invasive Ductal Carcinoma (IDC). Pathologists typically focus on regions which contain IDC to determine whether a patient suffers from breast cancer or not. We make use of publicly available Breast Histopathology Images dataset provided at the Kaggle for classification. In this dataset, images are delineated to extract the exact regions of IDC. This dataset contains 277,524 patches among which 198,738 belong to the negative class and 78,786 belong to the positive class (images with IDC). We present a novel deep convolutional neural network architecture for classification. Our performance in terms of area under the receiver operating characteristic curve for the detection of IDC is 0.94 on a set of 27,553 test images thereby setting a new benchmark for future research efforts.


  • Aggregate Channel Features Based on Local Phase, Color, Texture, and Gradient Features for People Localization

    by Hussin K. Ragb, Redha Ali, Vijayan Asari

    Detection of human beings in a complex background environment is a great challenge in computer vision. For such a difficult task, most of the time no single feature algorithm is rich enough to capture all the relevant information available in the image. To improve the detection accuracy, we propose a new descriptor that is constructed from the channels of image gradients, texture features, local phase information, and color features. This information is fused together to build one descriptor named Channels of Chromatic domain, local Phase with Gradient and Texture features (CCPGT). The image gradients, and local phase information based on the phase congruency concept are used to extract the human body shape features. Local binary pattern approach (LBP) is used to capture the texture features, and additional significant information for human detection are added by the color channels. The phase congruency magnitude and orientation of each pixel in the input image is computed with respect to its neighborhood and then six channels representing the histogram of oriented phase are generated. Gradient magnitude channel, LBP channel, and three LUV color channels are also computed for the input image. A maximum pooling of the candidate features is applied for these generated channels. All these features are concatenated in one feature vector and fed to a decision tree Adaboost classifier for training and to classify between the different objects classes. Several experiments were conducted to evaluate the proposed approach using challenging INRIA dataset, and a promising performance is observed.


  • Fused Deep Convolutional Neural Networks Based on Voting Approach for Efficient Object Classification

    by Redha Ali, Hussin K. Ragb

    Object classification has been one of the main tasks in computer vision. With the fast development of deep learning, its performance in image classification and object recognition has presented dramatic improvements. In this paper, we propose a new deep convolutional neural network (CNN) architecture for robust object classification. The proposed model is fused with three traditional CNN approaches, Densenet201, Resnet50, and our proposed residual CNN. The fused network architecture allows parallel processing of the multiple networks for keeping the system sped up. A single shot deep convolution network is trained as an object detector to generate all possible candidates of different object classes. The output of each neural network is representing a single vote that is used in the classification process. 3-to-1 voting criteria are applied in the final classification decision between the candidate object classes. Several experiments were conducted to evaluate the performance of the proposed network. The experimental results show that the proposed approach has better performance than the networks used in the fusion process when they act individually. It also has lower miss rates when compared to several states of art methodologies.


  • A Leaf Recognition Approach to Plant Classification Using Machine Learning

    by Redha Ali, Russell Hardie, Almabrok Essa

    The identification of plants is a very important component of workflows in plant ecological research. This paper presents an automated leaf recognition method for plant identification. The proposed technique is simple and computationally efficient. It is based on a combination of two types of texture features, named Bag-of-features (BOF) and Local Binary Pattern (LBP). These features are utilized as inputs to a decision-making model that is based on a multiclass Support Vector Machine (SVM) classifier. The introduced method is evaluated on a publicly available leaf image database. The experimental results demonstrate that our proposed method is the highly efficient technique for plant recognition.


  • Recursive non-local means filter for video denoising with Poisson-Gaussian noise

    by Redha Almahdi, Russell C. Hardie

    In this paper, we describe a new recursive Non-Local means (RNLM) algorithm for video denoising that has been developed by the current authors. Furthermore, we extend this work by incorporating a Poisson-Gaussian noise model. Our new RNLM method provides a computationally efficient means for video denoising, and yields improved performance compared with the single frame NLM and BM3D benchmarks methods. Non-Local means (NLM) based methods of denoising have been applied successfully in various image and video sequence denoising applications. However, direct extension of this method from 2D to 3D for video processing can be computationally demanding. The RNLM approach takes advantage of recursion for computational savings, and spatio-temporal correlations for improved performance. In our approach, the first frame is processed with single frame NLM. Subsequent frames are estimated using a weighted combination of the current frame NLM, and the previous frame estimate. Block matching registration with the prior estimate is done for each current pixel estimate to maximize the temporal correlation. To address the Poisson-Gaussian noise model, we make use of the Anscombe transformation prior to filtering to stabilize the noise variance. Experimental results are presented that demonstrate the effectiveness of our proposed method. We show that the new method outperforms single frame NLM and BM3D.