Page 32 - Research and Innovation Pulse

Task offloading and optimization methods in UAV-enabled mobile BioOtoFusionNet – A bio-attention & frequency-aware hybrid
edge computing: A comprehensive survey network for ear disease classification

Author(s): Author(s):

Tesfay, C.H. (School of Telecommunications Engineering, Xidian University), Xiang, Z. (School of Telecommunications Engineering, Jain, E. (Centre for Research Impact and Outcome, Chitkara University Institute of Engineering and Technology, Chitkara
Xidian University), Yang, L. (School of Telecommunications Engineering, Xidian University), Mahmood, J. (State Key Laboratory of University), Kukreja, V. (Centre for Research Impact and Outcome, Chitkara University Institute of Engineering and
Blockchain and Data Security, School of Cyber Science and Technology, and College of Computer Science and Technology, Zhejiang Technology, Chitkara University), Kaushik, P. (Centre for Research Impact and Outcome, Chitkara University Institute
University; Department of Computer Engineering, Faculty of Engineering and Architecture, Istanbul Gelisim University), Ren, M. of Engineering and Technology, Chitkara University), Ati, M. (Computer Science and Information Technology, College
(School of Telecommunications Engineering, Xidian University), Zhang, S. (School of Telecommunications Engineering, Xidian of Engineering, Abu Dhabi University), Hariharan, S. (Department of Computer Science and Engineering, Vel Tech
University), Das, A.K. (Center for Security, Theory and Algorithmic Research, International Institute of Information Technology; Rangarajan Dr. Sagunthala R&D Institute of Science and Technology)
Department of Computer Science and Engineering, College of Informatics, Korea University), Chaudhry, S.A. (Department of
Computer Science and Information Technology, College of Engineering, Abu Dhabi University; Department of Software Engineering,
Faculty of Engineering and Architecture, Nisantasi University) Index Terms:

Index Terms: Attention Mechanisms; Deep Learning; Dual-Stream Fusion; Ear Disease Classification; Interpretability Metrics; Otoscopic
Image Analysis
Antennas; Data mining; Edge computing; Energy efficiency; Energy utilization; Green computing; Surveying; Unmanned aerial vehicles
(UAV); Aerial vehicle; Communications channels; Edge computing; Energy-consumption; Optimisations; Optimization method; Abstract:
Sensitive application; Task offloading; Unmanned aerial vehicle-enabled mobile edge computing; Optimization

Objective Automated interpretation of otoscopic images is challenging due to subtle textural variations, anatomical
Abstract: complexity, and inconsistent acquisition conditions. This study aims to develop an accurate and interpretable
deep learning framework for ear disease classification. Methods This work presents BioOtoFusionNet, a clinically
The surge in compute-intensive and latency-sensitive applications such as autonomous driving and smart cities is straining motivated dual-branch architecture integrating a Frequency-Aware Stream based on Discrete Wavelet Transform
traditional cloud computing. Recent advancements in Mobile Edge Computing (MEC) and the emergence of Beyond 5G (B5G)
technologies have catalyzed the integration of Unmanned Aerial Vehicles (UAVs), heralding UAV-enabled MEC as a transformative (DWT) sub-bands and a Shape-Aware Stream that captures anatomical structures using edge-based features and
paradigm. By utilizing UAVs as flexible aerial edge servers or relays, MEC can bring computational resources closer to end-users, capsule attention. An Adaptive Cross-Fusion Module (ACFM) and Multi-Scale Attention Pooling (MSAP) are employed
thereby reducing latency and enhancing service quality. Despite their advantages, challenges remain regarding task offloading, to effectively fuse complementary representations across spatial resolutions. Results BioOtoFusionNet achieved an
specifically, determining when, how, and whether to transfer computational tasks from user devices to UAV-MEC servers for overall accuracy of 96.8%, an F1-score of 95.6%, and an AUC-ROC of 98.4%, outperforming all ablated variants. High
remote processing. UAV constraints, offloading strategies, communication channels, and energy consumption complicate the task class-wise accuracy was observed for Earwax Plug (97.8%), Myringosclerosis (94.5%), Chronic Otitis Media (93.2%),
offloading process. Therefore, sophisticated optimization strategies are essential to balance latency, energy consumption, task and Normal Ear (97.1%). Interpretability and Robustness Clinically motivated interpretability metrics demonstrated
completion rates, and overall system efficiency. This survey paper focuses on task offloading strategies and optimization methods balanced reliance on texture and structure (FHI = 0.60, SAR = 1.25), strong attention localization (ALS = 0.72), and
for UAV-enabled MEC networks. We discuss potential UAV-enabled MEC architectures, focusing on the interactions between UAVs stable multi-scale behaviour (MSAC = 0.87). Robustness analysis showed resilience to illumination variations (RII = 0.15)
operating within the MEC framework and ground users, as well as communication channel access. We present task offloading and low diagnostic ambiguity (DCS = 0.31). Evaluation Protocol Evaluation was conducted on a four-class otoscopic
methods and optimization techniques. Although UAV-enabled MEC holds great promise, it still faces challenges that need to dataset using stratified five-fold cross-validation with strict separation between training and evaluation samples. Data
be addressed to support a broader range of applications. We highlight research issues and future opportunities, divided into augmentation was applied only to training subsets to prevent information leakage. Conclusion BioOtoFusionNet
three levels based on the real-world deployment of UAV-enabled MEC, moving from those most critical to those important for provides accurate, interpretable, and robust ear disease classification from otoscopic images, highlighting its potential
optimization and scale.
for clinical decision support and telemedicine-based otologic screening.
Read the paper
Read the paper

Abu Dhabi University | Research and Innovation Pulse Newsletter May Edition | 33