Optimizing Hospital Workflows Through Artificial Intelligence Karthic Sundaram, S. Vinu, Arunkumar Ramamoorthy, S. Venkatesan Evaluation and Assessment of AI Driven Systems in Hospitals, 2025 The integration of latest technologies like Artificial Intelligence (AI) plays a vital role in improving the Healthcare systems right from real time monitoring, health record maintenance and efficient disease diagnosis and appropriate treatment. AI systems incorporate machine learning (ML) models, Deep learning models (DL) models, IOT devices and sensor data to provide real time monitoring of patient health. AI models assess lung imaging and real-time oxygen saturation data to predict respiratory deterioration. The power of AI in predictive analytics helps in identifying disease at the early stage to plan for faster treatment and recovery. Automated documentation and AI-powered EHR enhances clinical workflows reducing errors and improves patient data management. Generally, the hospitals have large amount of patient data. The usage of AI helps to derive actionable insights from these data that aids in better and faster recovery reducing mortality rates
Public Safety Management in Smart Society 5.0: A Blockchain-Based Approach P.N. Senthil Prakash, S. Karthic, M. Saravanan Networked Sensing Systems, 2025 Society 5.0 represents a framework to indicate representation of cyberspace and physical space, which is aimed to provide solutions to the societal challenges, promoting economic growth using technology. In the recent years, significant advancements across various technological domains have produced various benefits to the human community. For instance, traffic management systems is capable of dynamically adjusting the signal timings considering the real-time vehicle flow, which helps to eliminate traffic in smart cities. Another scenario is the innovations in autonomous vehicles including cars and other passenger vehicles. These are capable of providing more secure traveling experience reducing the need for human drivers. Similarly, smart healthcare systems efficiently provide essential treatments to patients and support doctors with detailed reports helping in further diagnosis and treatment more accurately. Similarly, the innovation in technology plays a significant role in industries, such as automotive, construction, and food, which provide improvements in terms of efficiency and productivity. This leads to streamlined process of the industries to meet the evolving demands of the market. For instance, consider the manufacturing sector; regulatory bodies are essential to ensure the production activities by conducting periodic audits to ensure compliance with safety, environmental, and quality standards. Despite such supervision, deviation incidents still occur causing risks to industry workers and the public residing near the industries. One of the areas where blockchain technology provides a promising solution is to improve the regulatory overhead and overcome industrial risks. Data are captured from industrial operations and stored in blocks and transmitted to regulatory bodies in real time utilizing blockchain technology. This ensures that data stored in the distributed ledger remains unaltered enhancing transparency and accountability of the industrial information. Smart contracts are a key feature of blockchain technology, which is utilized to enforce compliance with regulatory standards, since smart contracts execute automatically and perform the specified tasks when the specific criteria are met. From the huge amount of data generated, information, like quality of product, regulatory compliance, and environmental impact, can be obtained. Applying blockchain with these data can ensure regulatory compliance and quality of the product. The integration of blockchain into an industry process can provide significant improvement and contribute positively toward Society 5.0. The industries can utilize blockchain technology to improve environment practices being followed. This also enables the industries to have proper planning toward waste disposal, pollution emission, and resource consumption effectively. Thus, integrating blockchain into industrial processes contribute to sustainability and enhance the quality of life.
Blockchain Based Decentralized Adaptive Client Selection for Heterogeneous IIoT Environments Federated Learning Vinu S, Maheswaran N, Karthic Sundaram, Karthick S 4th International Conference on Applied Artificial Intelligence and Computing Icaaic 2025, 2025 Federated Learning (FL) has emerged as a potential approach for privacy-preserving collaborative learning within Industrial Internet of Things (IIoT) systems, where devices generate vast amounts of sensitive and diverse data. Nevertheless, traditional federated learning models encounter significant challenges, such as inefficiencies in handling non-independent and identically distributed (non-IID) data, communication costs, privacy issues, and vulnerability to adversarial attacks. This paper introduces an innovative distributed federated learning system that integrates blockchain technology with active client selection and differential privacy methodologies to address these challenges. The proposed approach eliminates reliance on a central server by employing blockchain technology to disseminate and safeguard the consolidation of model updates. By giving customers with highquality updates top priority, dynamic client selection based on uncertainty sampling and entropy measurements helps to speed up model convergence and improve communication efficiency. Differential privacy adds controlled noise to local model updates, which keeps data private while also making sure it is accurate. The system also uses a lightweight Proof-of-Contribution (PoC) consensus method to check client updates, give reputation ratings, and get people to be honest. Tests on benchmark datasets including MNIST, CIFAR-10, and LEAF synthetic reveal that the proposed system works better than current methods like FedAvg, clustered FL (CFL), and blockchain-based FL. More accurate (96.8 % on MNIST and 88.7 % on CIFAR-10), faster convergence (up to 25 % less rounds), and improved protection against attacks from bad actors The proposed approach only results in a 2.5 % loss in performance when 30 % of clients are malevolent. This study solves the problems of scalability, privacy, and security, making it possible to create reliable and large-scale decentralized learning systems. This makes it a robust and efficient solution for federated learning in IIoT environments.
A Trust-Enabled Edge-Assisted Depth-Based Protocol for Reliable and Secure Data Transmission in Underwater IoT S Sivsakthiselvan, S Pavithra, Karthic Sundaram, Han Yi Chiew, K Rama Abirami 2025 International Conference on Green Energy Computing and Sustainable Technology Gecost 2025, 2025 The Internet of Underwater Things (IoUT) enables critical applications such as ocean monitoring, seismic prediction, and defense operations. However, underwater acoustic communication faces unique challenges, including limited bandwidth, high latency, energy constraints, and frequent link disruptions. These factors, coupled with the absence of robust trust mechanisms, make IoUT networks highly vulnerable to malicious attacks such as packet dropping, false data injection, and energy depletion, which degrade reliability and network lifetime. To overcome these challenges, this paper proposes a Trust-Enabled Edge-Assisted Depth-Based Protocol (TE-EA-DBP) for secure and reliable data transmission in IoUT. The protocol incorporates a trust evaluation model to detect and isolate untrustworthy nodes based on forwarding behavior, reliability, and residual energy. To reduce the computational load on constrained nodes, edge-assisted processing is employed for trust assessment and Cluster Head (CH) election. Furthermore, a depth-based CH selection strategy balances energy consumption and ensures stable routing paths. Simulation results show that TE-EA-DBP achieves higher packet delivery ratio, energy efficiency, and resilience to adversarial behavior compared with existing protocols, offering a scalable and dependable solution for underwater communication.
Advancing building energy efficiency: A deep learning approach to early-stage prediction of residential electric consumption Karthic Sundaram, K.R. Sri Preethaa, Yuvaraj Natarajan, Akila Muthuramalingam, Ahmed Abdi Yusuf Ali Energy Reports, 2024 Enhancing building energy efficiency underscores the critical need for innovative predictive models to mitigate environmental issues from high energy consumption, especially in residential areas with air-conditioning and heating ventilation systems. This study introduces the use of Long Short-Term Memory (LSTM) networks for early prediction of residential electric consumption, representing a significant innovation in the field. Unlike traditional Deep Neural Network (DNN) and Artificial Neural Network (ANN) models, Long Short-Term Memory networks efficiently process time-series data, predicting future energy usage with unmatched accuracy. The Long Short-Term Memory model exhibited superior training efficiency, requiring only 2.69 s for over 500 test cases, outperforming Deep Neural Network and Artificial Neural Network models, which took 5.26 and 3.88 s, respectively. Its predictive accuracy, evidenced by an R-squared value of 0.97, surpasses the 0.95 and 0.92 of Deep Neural Network and Artificial Neural Network models, respectively. This breakthrough enables accurate predictions of annual energy usage before construction starts and aids in identifying energy efficiency improvements early in the design process. Applying Long Short-Term Memory networks in this context marks a substantial advancement in predictive modeling for building energy consumption, equipping architects and engineers with a vital tool for designing energy-efficient buildings from the beginning. The innovation and quantitatively proven effectiveness of the Long Short-Term Memory model highlight its potential to revolutionize early-stage building design strategies, filling a crucial gap in the existing literature.
Bias and fairness in generative AI Mani Deepak Choudhry, M. Sundarrajan, Karthic Sundaram, Abirami K. Rama Generative AI and Llms Natural Language Processing and Generative Adversarial Networks, 2024
Deep learning methods for high-level control using object tracking Urban Air Mobility Intelligent Safe and Sustainable Systems for Future Transportation, 2024
Securing Electronics Gadgets using Blockchain based Framework P. N. Senthil Prakash, B. Lanitha, Karthic Sundaram 2nd International Conference on Emerging Research in Computational Science Icercs 2024, 2024 The usage of electronic gadgets has grown dramatically in the current digital era. Devices like mobile phones and smart watches are unavoidable and has ingrained themselves into every person's daily routine. Almost everyone owns at least one portable electronic device and these devices are usually expensive. Due to this widespread use, there is also an increase in the number of lost or stolen cases reported. In a recent study, it has been revealed that the number of devices lost is mostly because of misplacing in public places like home, office, public transport. Such incidents will not only cause financial loss but it may also lead to data breaches stored in the portable devices. Hence there is a need for a secure eco-system which helps to identify the genuine owners of electronic devices. A blockchain-based framework is deployed on an Ethereum-based network and adopts the Proof of Stake (PoS) mechanism to protecting gadgets. Further, it is designed using the smart contract defined using solidity programming language. The system enables the owner's details to be recorded upon purchase of a device. The complete specification details of the devices along with owner details is contained in a block and then appended to network. Later on, if the owner wants to resale the device, then the transfer of ownership information is stored in a block and added to the network. The proposed system provides an appropriate interface to check the genuine owner. The proposed framework has been implemented using solidity language, Remix IDE and Sepolia testnet. Obtained results highlights the suitability of the method and its performance in terms of gas consumption and execution latency.
