The "INDIAai" National AI Portal of India offers weekly articles that showcase the AI research contributions of universities across the country. Each article provides an in-depth report on the work of a specific university, allowing researchers and students to present concise explanations of their research. 

This week, the portal highlights the top AI research contributions from G H Raisoni College of Engineering and Management, Nagpur- Maharashtra, India.

Raisoni Education, a prominent educational network in Central India, has a longstanding tradition of cultivating future leaders across diverse fields. Founded in 1998 with the establishment of G H Raisoni College of Engineering in Nagpur, the network has expanded to include 19 institutes and 4 universities across six cities: Nagpur, Jalgaon, Pune, Amravati, Pandhurna, and Bhandara. The institution encourages students to interpret its motto in a way that inspires them—whether as a vision beyond sight, dreams, or excellence—motivating them to pursue their goals with determination and to reach new heights through the power of knowledge.

Normalized Feature Plane Alteration for Dental Caries Recognition

Authors

• Shashikant Patil, EXTC Department, MPSTME, SVKMs NMIMS, Shirpur, Maharashtra, India
• Smita Nirkhi, AI Department, GHRIET, Nagpur, Maharashtra, India
• Suresh Kurumbanshi, EXTC Department, MPSTME, SVKMs NMIMS, Shirpur, Maharashtra, India
• Mayank Kothari, EXTC Department, MPSTME, SVKMs NMIMS, Shirpur, Maharashtra, India  
• Sachin Sonawane, EXTC Department, MPSTME, SVKMs NMIMS, Shirpur, Maharashtra, India 

This study explores the integration of advanced AI techniques in digital dental imaging to enhance the accuracy of dental diagnoses, particularly in assessing root canal treatment (RCT) complications and cavity severity. The proposed approach effectively repairs cracks and stretches identified in X-ray images by leveraging sophisticated image processing methods such as contrast stretching and active shape modelling. Multilinear subspace learning, including multilinear principal component analysis (MPCA) and multilinear discriminant analysis (MLDA), further refines the diagnostic process, offering superior accuracy compared to traditional methods. This innovative methodology not only improves diagnostic precision but also serves as a reliable tool for providing secondary opinions, thereby advancing the field of dentistry.

Performance Analysis of Smart System with Algorithmic Optimization for Cavities Detection 

Authors

• Shashikant Patil, EXTC Department, MPSTME, SVKMs NMIMS, Shirpur, Maharashtra, India
• Smita Nirkhi, AI Department, GHRIET, Nagpur, Maharashtra, India
• Suresh Kurumbanshi, EXTC Department, MPSTME, SVKMs NMIMS, Shirpur, Maharashtra, India
• Mayank Kothari, EXTC Department, MPSTME, SVKMs NMIMS, Shirpur, Maharashtra, India  
• Sachin Sonawane, EXTC Department, MPSTME, SVKMs NMIMS, Shirpur, Maharashtra, India

This study introduces a novel AI-driven methodology for the early detection and analysis of dental cavities, addressing the limitations of existing approaches in identifying carious lesions at their initial stages. The proposed approach comprises three key segments: pre-processing, feature extraction, and classification. Image quality is first enhanced using advanced pre-processing techniques, including contrast improvement, grey thresholding, and active contouring. Features are then extracted using Multilinear Principal Component Analysis (MPCA).

Finally, classification is performed with a neural network (NN) optimized by the Adaptive Dragonfly Algorithm (DA). The integration of MPCA with adaptive DA (MNP-ADA) demonstrates superior accuracy and classification performance compared to conventional methods, offering a promising tool for enhancing diagnostic efficacy in dental care.

Survey of Fusion Techniques for the Design of Efficient Multimodal Systems

Authors

• S.A. Chhabria, G.H. Raisoni College of Engg., Research Scholar, Nagpur, India
• R.V. Dharaskar, Integrated Campus, Nanded
• V.M. Thakare, Amravati University

This paper provides a comprehensive survey of fusion techniques employed in designing efficient multimodal systems, particularly in the context of human gesture recognition for intelligent human-computer interaction (HCI). Human gestures encompass a range of visual actions, including hand movements, facial expressions, torso movements, eye motions, and speech, all of which convey meaning and facilitate interaction. Previous research has focused on the manual components of gestures, such as hand movements, and has yet to integrate other modalities fully.

The paper introduces a multimodal gesture recognition framework that synthesizes various feature groups, notably hand movement features and facial expression features, to enhance the accuracy and robustness of gesture recognition systems. The proposed approach aims to create a more holistic and effective system for understanding and interpreting human gestures by combining inputs from different modalities.

Hybrid fusion combines early and late fusion elements to balance the trade-offs between data integration and processing complexity. The survey highlights the importance of selecting appropriate fusion techniques based on the specific application and the nature of the modalities involved. It also underscores the need for robust algorithms to handle the variability and noise inherent in multimodal data, ensuring accurate gesture recognition across diverse contexts.

A Comparative Analysis of Deep Learning Models and Conventional Approaches for Osteoporosis Detection in Hip X-Ray Images

Authors

• Virja Kawade, Department of Artificial Intelligence, G H Raisoni Institute of Engineering and Technology, Nagpur, India
• Vedant Naikwade, Department of Artificial Intelligence, G H Raisoni Institute of Engineering and Technology, Nagpur, India
• Vibha Bora, G H Raisoni College of Engineering, BETiC, Nagpur, India
• Sharda Chhabria, Department of Artificial Intelligence, G H Raisoni Institute of Engineering and Technology, Nagpur, India

This study explores the potential of deep learning techniques as an alternative to the costly DEXA scan for osteoporosis detection. Osteoporosis, characterized by reduced bone mineral density (BMD) and an increased risk of fractures, is traditionally diagnosed using DEXA, the current gold standard. However, due to the high costs associated with DEXA, it is often inaccessible to individuals from lower socioeconomic backgrounds. The study proposes applying deep learning algorithms on X-ray images to offer a more affordable diagnostic option.

The research evaluates and compares four deep learning models—ResNet-50, Inception Net, YOLOv7, and Ultralytics YOLO v8—on an augmented X-ray dataset of 117 hip images. Among these, the Ultralytics YOLO v8 model demonstrated the highest accuracy, indicating its potential as a viable alternative for osteoporosis detection. This approach could democratize access to accurate osteoporosis diagnostics, making it more affordable and accessible to a broader population.