Neuroflux Is Published In The International Journal of High School Research
- Neuroflux Team
- Jul 6
- 2 min read
Updated: Sep 17
We are proud to announce that Neuroflux has been officially published in the June 2025 issue of the International Journal of High School Research! It marks a significant milestone in our mission to advance explainable AI in neuro-oncology imaging.
Rigorous Peer Review Process
Across our 7-month publishing process, our research underwent an extensive external peer review process by 3 independent reviewers who assessed the technical rigour and clinical relevance of Neuroflux.
We are grateful to have received overwhelmingly positive evaluation results across the following criteria:
Topic appropriateness rated as "Highly Appropriate"
Conceptual adequacy received "Outstanding" from all reviewers
Technical adequacy scored "Outstanding" and "Good"
Contribution to the field rated "Outstanding" and "Good"
What the Reviewers Said:
The reviewers highlighted the presentation quality as "clear and well-structured, making it easy to follow along". We were also commended on the sophistication of our work by demonstrating "a college-level and above understanding of how machine learning models are built and tested for human diseases, particularly for neuroscience."
Research Highlights
The published paper comprehensively details our multimodal U-Net architecture and its applications in tumor segmentation. Neuroflux's key contributions include:
Novel hybrid loss function combining Categorical Cross-Entropy with Dice Loss for improved segmentation accuracy
Our real-time Grad-CAM integration aims to build clinical trust through transparent decision-making
Holistic multimodal MRI analysis using T1ce and FLAIR sequences for comprehensive tumor characterisation
Lightweight implementation and high computational efficiency to deploy across diverse clinical environments
International Collaboration and Global Accessibility
This publication represents a truly international collaborative effort, across Romania, Singapore and the United States. Since September 2024, we worked to build Neuroflux with the shared goal of advancing equitable healthcare through ethical AI.
Our work also emphasised Neuroflux's potential to democratize access to expert-level brain tumor analysis, particularly in regions with limited neuroradiology expertise. Neuroflux's open-source approach under the Apache 2.0 license reflects our vision for ensuring our technology remains accessible and adaptable for helathcare systems worldwide.
Access the Full Research
Our full 7-page research paper is now available for download and citation at https://doi.org/10.36838/v7i6.64
It provides a comprehensive overview of Neuroflux's technical details, our validation procedures, and the performance metrics that supports NeuroFlux's potential for clinical implementation.
Publication Details
Title: Neuroflux: An Interpretable Multimodal U-Net Model for Glioblastoma Multiforme Segmentation Through Grad-CAM
Authors: Eve Ang Yee Leng, Darius Mihai Oltei, Pranav Vedula, Shriya Viswanathan
Journal: International Journal of High School Research Volume 7, Issue 6
Pages: 64-70
DOI: 10.36838/v7i6.64
Citation Information
Ang, E. Y. L., Oltei, D. M., Vedula, P., & Viswanathan, S. (2025).
Neuroflux: An Interpretable Multimodal U-Net Model for Glioblastoma
Multiforme Segmentation Through Grad-CAM. International Journal of
High School Research, 7(6), 64-70. https://doi.org/10.36838/v7i6.64Looking Forward
With our research now formally published and peer-reviewed, we're excited to continue improving Neuroflux's capabilities in diverse clinical environments. The publication serves as a foundation for future work in explainable AI for medical imaging.
Get Involved
Interested in implementing Neuroflux at your institution or collaborating on further research? Visit our GitHub repository for the complete codebase and documentation, or contact us here!
