Date-Line : 10th October - 20th December

Description : Following the initial screening of the antibacterial activity of crude extracts of the leaves of Ixora coccinea, Dracaena fragrans, and Duranta repens (Pet. ether, CHCL3, and CH3OH), the result showed that a few medicinal bio-compounds with antibacterial activities were promising. However, the phyto- chemicals responsible for these crude extracts' bacterial inhibitory zone are not specifically identified. The prospective compound for the creation of a novel antibiotic medication must be identified using computational approaches. 

Process : Data Filtering 

Methodologies : Collected Authentic Wet & Dry Lab Data , Graph Neural Network , Molecular Graph Implementation , Bacterial Behaviour Analysis

Date-Line : 3rd October - 29th November

Description : This innovative study combines quantum computing techniques with hierarchical federated learning to analyze multi-source knee radiography data, utilizing Quantum Convolutional Neural Networks (QCNN) for enhanced image processing. The research incorporates blockchain technology to ensure secure data transactions across the federated network, potentially revolutionizing both the accuracy and privacy aspects of medical image analysis. 

Process : Model Profiling

Methodologies : Quantum Machine Learning (QML) , Image Pre-Processing , Quantum Data Augmentation , Variational circuits

Findings : The findings of this project revealed that the transformation of 2D ultrasound data into 3D mesh representations, facilitated by an advanced Variational Quantum Neural Network framework, significantly augmented the differentiation of ovarian tumours. This innovative approach not only achieved superior accuracy in tumour classification but also enhanced interpret ability, thereby providing clinicians with a more robust tool for informed decision-making in ontological assessments. 

Contributions : First Author

Writing Process : Completed approximately 60% 

Target Journal: IEEE Transactions on Quantum Engineering

Dateline : 23rd November

Findings : This project appears to develop a novel multi-modal graph neural network for analyzing 3D ovarian ultrasound images using self-supervised contrastive learning. The approach likely incorporates adaptive graph generation and region-of-interest focused feature extraction to improve analysis of ovarian structures from ultrasound mesh data. 

Contributions : First Author

Writing Process : Completed approximately 40% 

Target Journal: IEEE Transactions on Neural Networks and Learning Systems

Dateline : 1st December

Findings : The Brain-X-GNN study found that feature-based Graph Neural Networks (GNNs) offer superior accuracy and computational efficiency in identifying brain tumors compared to traditional models. Additionally, GNNs provide enhanced explainability by highlighting important brain regions and features, making them more interpretable and suitable for clinical applications. 

Role : Mentor

Process : Writing

Target Journal: Journal of Healthcare Informatics Research 

Dateline : 1st August -29th October

Findings : The study found that using self-supervised learning with Graph Neural Networks (GNNs) in a federated learning setup improved arrhythmia detection on a 12-lead ECG dataset. This approach also enhanced model performance by utilizing distributed data without compromising privacy.  

Role : Mentor

Process : Implementing

Target Journal: Springer Nature

Dateline : 25th November - 15th December