Mr Dhafer Alhajim received his B.S. degree in Computer’s Technique Engineering from The Islamic University of najaf, Iraq in 2012, and the MSc degree as scholarship with the first rank in computer engineering in a subfield of computer Architecture from Tabriz university, Iran in 2018. He is currently a PhD candidate in Department of Electrical Engineering, Faculty of Engineering, Shahid Chamran University of Ahvaz, Iran. His research interests are Image processing, Machine Learning, Deep Learning, IOT and Cloud Computing.
RESEARCH, TEACHING, or OTHER INTERESTS
Computer Engineering, Artificial Intelligence, Computer Vision and Pattern Recognition, Hardware and Architecture
4
Scopus Publications
Scopus Publications
Improved lung nodule segmentation with a squeeze excitation dilated attention based residual UNet Dhafer Alhajim, Karim Ansari-Asl, G. Akbarizadeh, M. N. Soorki Scientific Reports, 2025 The diverse types and sizes, proximity to non-nodule structures, identical shape characteristics, and varying sizes of nodules make them challenging for segmentation methods. Although many efforts have been made in automatic lung nodule segmentation, most of them have not sufficiently addressed the challenges related to the type and size of nodules, such as juxta-pleural and juxta-vascular nodules. The current research introduces a Squeeze-Excitation Dilated Attention-based Residual U-Net (SEDARU-Net) with a robust intensity normalization technique to address the challenges related to different types and sizes of lung nodules and to achieve an improved lung nodule segmentation. After preprocessing the images with the intensity normalization method and extracting the Regions of Interest by YOLOv3, they are fed into the SEDARU-Net with dilated convolutions in the encoder part. Then, the extracted features are given to the decoder part, which involves transposed convolutions, Squeeze-Excitation Dilated Residual blocks, and skip connections equipped with an Attention Gate, to decode the feature maps and construct the segmentation mask. The proposed model was evaluated using the publicly available Lung Nodule Analysis 2016 (LUNA16) dataset, achieving a Dice Similarity Coefficient of 97.86%, IoU of 96.40%, sensitivity of 96.54%, and precision of 98.84%. Finally, it was shown that each added component to the U-Net's structure and the intensity normalization technique increased the Dice Similarity Coefficient by more than 2%. The proposed method suggests a potential clinical tool to address challenges related to the segmentation of lung nodules with different types located in the proximity of non-nodule structures.
Application of Optimized Deep Learning Mechanism for Recognition and Categorization of Retinal Diseases Dhafer Alhajim, Ahmed Al-Shammari, Ahmed Kareem Oleiwi International Journal of Computing and Digital Systems, 2024 Retinal disorders are one of the common eye problems and its complication affects the eyes. In some cases, the retinal diseases would not cause any symptoms, or they only show mild vision impairments. Finally, it causes no vision or blindness. So, earlier recognition of symptoms could help to avoid blindness. Routine screening is one of the methods for early diagnosis of retinal disease. Another common way to identify retinal disease is to have an expert evaluate and rate eye photographs for the existence and severity of the illness. Unfortunately, in many parts of the world where retinal disease is common, the medical specialists capable of recognizing DR are scarce. Hence, a novel optimized African Buffalo-based deep Convolutional Neural Network (AB-DCNN) deep learning model is introduced in this article, which could detect the retinal disorders in the earlier stage from the fundus retinal image datasets and classify its stages. The proposed mechanism could detect diseases like Central Serous Retinopathy (CSR), Age-related Macular Degeneration (AMD), Diabetic Retinopathy (DR) and Macular hole (MH) and classify its stages as Severe, Moderate, Mild NPDR, PDR, and normal case. Depending upon the clinical importance, the impact of uncertainty on system performance and the relation between explained ability and uncertainty are examined. The uncertainty evidences make the system more reliable for usage in clinical environments. The proposed methodology increases the operational speed and lessens the computation time of the algorithm. It also reduces the losses and enhances classification accuracy.
A Novel ConvLSTM-Based U-net for Improved Brain Tumor Segmentation Osama Majeed Hilal Almiahi, Alaa Taima Albu-Salih, Dhafer Alhajim IEEE Access, 2024 Using 2D scans or simple 3D convolutions are two limitations of previous works on segmentation of brain tumors by deep learning, which lead to ignoring the temporal distribution of the scans. This study proposes a novel extension to the well-known U-net model for brain tumor segmentation, utilizing 3D Magnetic Resonance Imaging (MRI) volumes as inputs. The method, called ConvLSTM-based U-net + up skip connections, incorporates the ConvLSTM blocks to capture spatio-temporal dependencies in the 3D MRI volumes, and up skip connections to capture low-level feature maps extracted from the encoding path, enhancing the information flow through the network to the standard U-net architecture. A novel intensity normalization technique is used to improve the comparability of scans. This technique normalizes image intensity by subtracting the grey-value of the most frequent bin from the image. The novel method is tested on the Multimodal Brain Tumor Segmentation (BRATS) 2015 dataset, showing that the use of ConvLSTM blocks improved segmentation quality by 1.6% on the test subset. The addition of skip connections further improved performance by 3.3% and 1.7% relative to the U-net and ConvLSTM-based U-net models, respectively. Moreover, the inclusion of up skip connections could enhance the performance by 5.7%, 3.99% and 2.2% relative to the simple U-net, ConvLSTM-based U-net, and ConvLSTM-based U-net with skip connections, respectively. Finally, the novel preprocessing technique had a positive effect on the proposed network, resulting in a 3.3% increase in the segmentation outcomes.
FFDR: Design and implementation framework for face detection based on raspberry pi Dhafer Alhajim, Gholamreza Akbarizadeh, Karim Ansari-Asl Iranian Conference on Machine Vision and Image Processing Mvip, 2022 In today’s world, we are surrounded by data of many types, but the abundance of image and video data available offers the data set needed for face recognition technology to function. Face recognition is a critical component of security and surveillance systems that analyze visual data and millions of pictures. In this article, we investigated the possibility of combining standard face detection and identification techniques such as machine learning and deep learning with Raspberry Pi face detection since the Raspberry Pi makes the system cost-effective, easy to use, and improves performance. Furthermore, some images of a selected individual were shot with a camera and a python program in order to do face recognition. This paper proposes a facial recognition system that can detect faces from direct and indirect images. We call this system FFDR, which is characterized by high speed and accuracy in the diagnosis of faces because it uses the Raspberry Pi 4 and the latest libraries and advanced environments in the Python language.
Industry, Institute, or Organisation Collaboration
