AI-driven hybrid metaheuristic neural network approaches for medical diagnostics Abdullah Khan, Maria Ali, Dzati Athiar Ramli Systems and Soft Computing, 2026 In recent years, metaheuristic algorithms have gained prominence as powerful computational techniques for solving complex optimization problems across various domains, including healthcare. Their ability to effectively explore vast solution spaces enables them to identify optimal or near-optimal solutions to challenging problems. Despite their growing importance, metaheuristic algorithms, inspired by natural and human problem-solving strategies, are increasingly applied in healthcare to address complex optimization challenges such as diagnosis, treatment planning, and resource allocation. But traditional ML models often face problems like getting stuck in training, taking too much time to compute, and not balancing exploration and exploitation properly. These issues make them less effective for solving complex medical classification problems. To overcome these limitations, this study introduces new hybrid metaheuristic models such as Ropalidia Marginata (RM) hybrid with various metaheuristic algorithms such as Ant Colony Optimization (ACO), particle swarm optimization (PSO), Firefly, Artificial Bee Colony (ABC), Grey Wolf Optimizer (GWO), Differential Evolution (DE), and Bat Algorithm. Further all these hybrid metaheuristic algorithms are combined with a Feedforward Neural Network (FFNN) to train the network more efficiently. The proposed method uses the dominance-based behavior of RM wasp to improve the global search ability, is stable during learning, and improves classification accuracy. The performance of proposed frameworks is to check against several start of the art models on three critical medical datasets such as breast cancer, diabetes, and heart disease. The results showed that proposed models gave better results than the traditional algorithms in terms of accuracy, mean squared error (MSE), standard deviation (SD), and AUC-ROC.
A hybrid bio inspired neural model based on Ropalidia Marginata behavior for multi disease classification Maria Ali, Abdullah Khan, Dzati Athiar Ramli, Muhammad Imran, Javed Iqbal Bangash, Arshad Khan Scientific Reports, 2025 Accurate and efficient disease diagnosis remains a critical challenge in the healthcare sector. With the growing availability of biomedical data, machine learning techniques have become invaluable tools for developing intelligent disease detection systems. Researchers have applied various algorithms, including artificial neural networks (ANNs), to improve classification accuracy. To further improve ANN performance, various optimization methods are applied to enhance learning and avoid the local minima problem, as each model demonstrates distinct performance characteristics. Therefore, this paper presents a hybrid Bio inspired Ropalidia Marginata Optimization-based hybrid neural network (RMO-NN) aimed at improving medical data classification. The proposed RMO-NN incorporates biologically inspired task allocation and dominance hierarchy mechanisms from RMO to optimize neural network learning performance effectively and reducing classification errors. To validate its effectiveness, the RMO-NN is tested on three large-scale medical datasets such as breast cancer, diabetes, and blood transfusion datasets and three medical images datasets. The performance of the proposed model is compared against two established metaheuristic neural models: Cuckoo Search Neural Network (CSNN) and Artificial Bee Colony Neural Network (ABCNN). The proposed RMO-NN model outperforms CSNN and ABCNN in terms of accuracy, MSE, SD, and convergence speed. And for medical images datasets the proposed is further validated with various start of art deep learning models. The results highlight the proposed model perform better on biomedical data classification tasks. The Proposed method significantly outperforms baseline approaches, achieving substantial accuracy, while introducing a novel RMO algorithm.
Diagonal Factorization Subspace in I-Vector Extraction for Fast Computation and Memory Efficiency: A Case Study on Frog Sound Identification Dzati Athiar Ramli, Noor Salwani Ibrahim IEEE Access, 2025 In this study, we investigate the performance of low-dimensional identity vectors, called I-vectors, in automatically identifying frog species on the basis of their bio-acoustics. I-vector is one of the state-of-the-art techniques and has greatly improved speaker recognition performance. However, the I-vector method has some limitations, such as slow computation and high memory consumption. To address these issues, this study improves the I-vector algorithm using diagonal techniques. Although the proposed method reduces computational cost, it leads to some accuracy degradation. Therefore, an enhanced version, called Diagonal Factorisation Subspace in I-vector (DFS I-vector), is developed. Additionally, channel compensation fusion is implemented to further improve the accuracy of the I-vector approach. Frog species identification are employed as a case study for this research and experiment data for this study are employed from five databases, i.e. Frog Identification Expert System Database, Frogs of Australia Database, Frog Watch Database, British Library Amphibian Database and Intelligent Biometric Group Universiti Sains Malaysia Database, which consist of 2,656 samples of bio-acoustic syllables. Experimental results prove the promising output of the proposed DFS I-vector-32, achieving 90.67% accuracy, which is five times faster, and reducing memory consumption by one-fifth the memory of standard I-vector-16. Experimental results also demonstrate the advantage of the fusion approach, DFS I-vector-32 with WCCN(LDA), achieving 91.56% accuracy. Both performances are better than that of standard I-vector 16, which achieves 88.00% accuracy.
Enhancing Smart Outdoor Object Navigation for the Visually Impaired via YOLOv10 With Neighbor Coordinates and C2FCIB Attention Mechanism Abdullah Khan, Dzati Athiar Ramli, Muhammad Zubair Rehman, Javed Iqbal Bangash, Muhammad Nouman Atta IEEE Access, 2025 This research introduces an AI-driven outdoor object detection system aimed at enhancing navigation for visually impaired individuals (VIIs). VIIs often face significant challenges in accessing and interpreting visual information. Recent advancements in computer hardware and deep learning techniques have led to notable progress in developing assistive technologies for VIIs. However, existing datasets often focus on single scenarios and lack sufficient annotations to represent the diverse obstacles encountered in real-world settings. This limitation hinders the development of comprehensive object detection systems tailored to the needs of VIIs. The system utilizes advanced models such as YOLOv8 (Nano, Small, Medium), YOLOv9c, and YOLOv10n, with neighbor coordinates and C2FCIB attention modules trained on the WOTR dataset, which includes 20 classes of common outdoor objects. A comparative study evaluated the performance of these models across key metrics. The YOLOv8m model demonstrated balanced performance with an accuracy of 85.53. YOLOv8n showed slightly lower performance, with an accuracy of 77.05%, and the YOLOv8s model recorded an accuracy of 84.99%, precision and recall of 0.99, 0.89, matching YOLOv8n with an F1 score of 0.74. Similarly, YOLOv9c achieved an accuracy of 79.82%, and the proposed YOLOv10n model with neighbor coordinates and C2FCIB attention modules led with the highest accuracy of 89.33%, precision and recall of 0.99, 0.92, with F1 score of 0.79. A comparative analysis revealed that the proposed YOLOv10n with neighbor coordinates and C2FCIB attention modules achieved the highest accuracy with precision and recall indicating its reliability for assistive applications. In the realm of assistive technologies, similar AI-powered devices have been developed to aid visually impaired individuals. These innovations, alongside the described object detection system, exemplify the potential of AI in creating inclusive solutions that empower visually impaired individuals to navigate their environments more safely and independently.
Fractional synchrosqueezing transform for enhanced multicomponent signal separation Yangyang Li, Dzati Athiar Ramli Scientific Reports, 2024 The precise separation of multicomponent signals encounters numerous challenges due to the complexity of signals and widespread interference. Synchrosqueezing Transform (SST) is one of the important technologies for improving the accurate separation of multicomponent signals, but it faces challenges in terms of the difficulty and effectiveness of squeezing. This paper introduces a multicomponent signal separation method based on innovative Fractional Synchrosqueezing Transform (FrSST). FrSST rearranges along the fractional frequency axis, improving the accuracy of time-frequency ridges and, consequently, enhancing the precision of multicomponent signal separation. In the signal reconstruction process, chirp multiplication and energy rearrangement compensate for chirp bases' effects, boosting energy concentration and reconstruction potential. Utilizing improved ridges from FrSST ensures effective signal reconstruction. Simulation comparisons demonstrate that, with varying SNRs from - 5 to 15 dB, the reconstructed components based on FrSST exhibit favorable approximation to the original signal components. Furthermore, as the sample size increases, the proposed algorithm shows satisfactory computational efficiency.
Adaptive Fractional Instantaneous Frequency Estimation using Modulation Maxima and Phase Change Rate Yangyang Li, Dzati Athiar Ramli ACM International Conference Proceeding Series, 2024 Instantaneous frequency estimation plays a crucial role in signal processing; however, existing methods face challenges in achieving high precision, especially when dealing with non-stationary signals and low signal-to-noise ratio environments. To address this, this paper proposes a novel algorithm that combines phase change rate with modulation maxima to enhance the performance of instantaneous frequency estimation. The algorithm synergistically utilizes both time and frequency domain information, overcoming the limitations of traditional methods and significantly improving estimation accuracy and robustness. Through comparative experiments in various signal-to-noise ratio environments, we validate the superior accuracy of the proposed algorithm, particularly demonstrating outstanding performance in low signal-to-noise ratio conditions. This affirms the algorithm's practical superiority, providing an effective approach to enhance accuracy in signal identification.
Person identification using lip motion sequence Salina Abdul Samad, Dzati Athiar Ramli, Aini Hussain Lecture Notes in Computer Science Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics, 2007
