Medical Image Segmentation Based on Hybrid Deep Learning Techniques
الباحثونMohammed Abdulameer aljanabi
المجلةWasit Journal for Pure sciences
التاريخ2025-03-06
مختصر البحثMedical image segmentation is an integral component of computer-aided treatment planning and diagnosis, enabling accurate analysis of abnormalities and anatomical structures. In this paper, a novel U-Net and transformer encoder-based hybrid deep learning architecture is developed to advance precision and speed in medical image segmentation. The U-Net architecture, for its remarkable capability to capture local details, is integrated in synergy with a transformer encoder, whose power is in capturing data's long-range dependencies. The synergy between such capabilities makes the resultant hybrid architecture robust for segmentation in diverse medical imaging modalities. The approach is rigorously compared on publicly available datasets, including brain MRI, Chest X-Rays, and ISIC, and is shown to have better Dice coefficients than current state-of-the-art. The synergy between U-Net and transformer components, in addition to better segmentation, is shown to result in computational savings, thus becoming efficient for clinical application. The result demonstrates the capability of such a hybrid architecture to advance diagnostic precision and contribute to enhanced treatment planning in clinical settings. The work is in alignment with growing literature on using advanced deep learning to overcome challenging medical image analysis problems, ultimately leading to enhanced clinical outcomes through accurate and reliable tools for segmentation.
A Hybrid Deep Learning Approach for Lung Diseases Classification Using Vision Transformer and DenseNet
الباحثونNoora Abd alrazak Shnain aljanabi ، Mohammed Abdulameer Mohammed
المجلةNJF Intelligent Engineering Journal
التاريخ2025-12-08
مختصر البحثLung disorders, such as pneumonia, tuberculosis, and COVID-19, remain significant global health issues. Accurate and early detection is of paramount significance to ensure effective treatment and control. Conventional CNN models like DenseNet have shown superior performance in the classification of medical images, whereas Vision Transformers (ViTs) have recently gained momentum as formidable models to capture global context information. Each of these models, however, has limitations when operated standalone. This manuscript suggests an innovative hybrid deep learning framework that synergizes the benefits of both ViT and DenseNet to advance lung disease diagnosis using chest X-ray images. The hybrid model uses both ViT to learn long-range context and DenseNet to reuse features and transmit gradients. Our study examines our method using the ChestX-ray14 dataset and uses preprocessing to refine images and boost data. The hybrid model performs excellently in comparison to base case architectures of ViT and DenseNet and shows improvement in accuracy, sensitivity, precision, and F1-score. The outcomes demonstrate the merit of the integrated formulation in refining and improving lung disease diagnosis, thereby potentially contributing to precise and punctual diagnosis.
Design of a Hybrid Metric for Fingerprint Image Matching: A Grediant, Aczil, and SourceAFIS Approach
الباحثونMohammed Abdulameer Aljanabi ، Noor Abd alrazak Shnain
المجلةScientific Visualization
التاريخ2025-07-23
مختصر البحثOne of the most popular biometric matchings is the fingerprint used for personal matching, verification, and authentication. Numerous fingerprint metrics have been developed to detect various types of fingerprint image distortions. The majority of these metrics are based on a single approach. It is feasible to match a specific person's fingerprint image by comparing images of the same person's fingerprint, which we will discuss in detail in this paper. In this work, we provide an alternative technique by interpolating the Aczel metric with gradient matching and Source Automated Fingerprint Matching System (SourceAFIS) algorithm. The Aczel metric demonstrates strong predictive capabilities in discerning relationships among intensity values within fingerprint images while gradient matching is used to gage the change in contrast and structure in images. SourceAFIS describes fingerprints using high-level abstractions …
A HYBRID DEEP LEARNING AND QUALITY METRIC FOR IMPROVING FINGERPRINT IDENTIFICATION ACCURACY
الباحثونNoora Abd alrazak Shnain aljanabi ، Mohammed Abdulameer Mohammed
المجلةNED University Journal of Research
التاريخ2025-01-09
مختصر البحثBiometric systems play a vital role in enhancing security across various fields. Fingerprint recognition has gained significant attention due to the uniqueness and permanence of fingerprint patterns. However, despite this uniqueness, challenges remain in the accurate extraction and matching of fingerprint features. This paper proposes a hybrid system that leverages the strengths of the pretrained Visual Geometry Group model with sixteen convolutional layers (VGG16) and the Universal Quality Index (UQI) metric to improve fingerprint identification accuracy. As VGG16 may not effectively capture ridge structures and minutiae in fingerprints, the proposed VGG-UQI Score (VUS) metric is designed to combine the deep feature representations of VGG16 with the statistical properties of UQI. This fusion aims to enhance discriminatory power and reduce error rates. Extensive experiments conducted on large-scale
Novel Image Quality Index: Integrating UQI and GSI for Enhanced Quality Assessment
الباحثونNoor Abd alrazak Shnain ، Mohammed Abdulameer Aljanabi
المجلةFrontiers in AI and Computational Technologies
التاريخ2024-11-15
مختصر البحثAssessing image or image quality is extremely crucial in various applications in the fields of image processing and computer vision. Numerous methods have been presented to assess the quality of images by comparing between the reference and distorted images. However, most of the previous metrics presented in this field produced a significant amount of correspondence between the reference image and the distorted images. In view of these challenges, we would like to put forward the proposal for a new comprehensive index that gauges the quality of an image with regard to gradient information and statistical coefficients. This index fuses the Gradient Similarity Index (GSI) with Universal Quality Index (UQI) and the gradient-based edge detection method of Sobel. We evaluated the performance of our approach using two well-known databases the LIVE and CSIQ and the results indicate that the proposed approach outperforms the (GSI) and (UQI) even under distortion and noise.
Design of a hybrid measure for image similarity: a statistical, algebraic, and information-theoretic approach
الباحثونMohammed Abdulameer Aljanabi, Zahir M Hussain, Noor Abd Alrazak Shnain, Song Feng، Mohammed Abdulameer Aljanabi, Zahir M Hussain, Noor Abd Alrazak Shnain, Song Feng
المجلةEuropean Journal of Remote Sensing
التاريخ2018-12-03
مختصر البحثImage similarity or distortion assessment is fundamental to a wide range of applications throughout the field of image processing and computer vision. Many image similarity measures have been proposed to treat specific types of image distortions. Most of these measures are based on statistical approaches, such as the classic SSIM. In this paper, we present a different approach by interpolating the information theory with the statistic, because the information theory has a high capability to predict the relationship among image intensity values. Our unique hybrid approach incorporates information theory (Shannon entropy) with a statistic (SSIM), as well as a distinctive structural feature provided by edge detection (Canny). Correlative and algebraic structures have also been utilized. This approach combines the best features of Shannon entropy and a joint histogram of the two images under test, and SSIM with edge …