Professor Helen Dang is a new faculty member in the Math & Computer Science Department at Molloy University. She brings with her a wealth of knowledge and expertise in the field of Artificial Intelligence (AI) research, machine learning, and its practical applications. Her commitment to fostering a deep understanding of these subjects is evident through the array of courses she teaches. From foundational Data Structures and Algorithms to cutting-edge AI, Machine Learning and Data Science. Professor Dang's courses are designed to provide students with a comprehensive understanding of key concepts as well as a grasp of the core knowledge and especially their applications in the real world.
Professor Dang uses a hands-on, project-based approach to education that is exemplified through the Capstone course. This course not only serves as a culmination of students' academic journey but also offers them the opportunity to collaborate with her on research projects.
Recently, an article written by Professor Dang and her team was published in the esteemed journal Scientific Reports. In it, Professor Dang and her team proposed a new model that is effective in disease diagnosis using image data with advanced techniques such as AI and machine learning.
The research focused on a subject close to her heart: the detection of meningioma tumors. Fully aware of the challenges posed by a tumor’s lower pixel intensity, she believed that a more sophisticated approach was needed. Modern medical platforms, she argued, demanded a fully automated system for meningioma detection.
The professor’s idea was to create a novel and highly efficient Hybrid Convolutional Neural Network (HCNN) classifier. The team worked on building a cutting-edge system designed to distinguish meningioma brain images from their non-meningioma counterparts.
The HCNN classification technique, as described in the article, is a complex interplay of the Ridgelet transform, feature computations, a classifier module, and a segmentation algorithm. Professor Dang's keen insight led her and her team to incorporate the Ridgelet transform to enhance pixel stability during the decomposition process. The features crucial for classification were derived from the coefficients of the Ridgelet, creating a robust foundation for the HCNN.
The classifier module, a core component of the creation, meticulously analyzed these features to differentiate meningioma from non-meningioma images. The segmentation algorithm, another critical element, identified tumor pixels with remarkable precision, further elevating the diagnostic capabilities of the system.
To validate the creation, Professor Dang and her team subjected the HCNN-based meningioma detection system to rigorous testing using well-established datasets, including BRATS 2019, Nanfang, and the recently released BRATS 2022.
For the BRATS 2019 dataset, the proposed HCNN achieved an impressive 99.31% sensitivity, 99.37% specificity, and 99.24% segmentation accuracy. The Nanfang dataset witnessed similarly remarkable results, with 99.35% sensitivity, 99.22% specificity, and 99.04% segmentation accuracy on brain Magnetic Resonance Imaging (MRI). The results were nothing short of extraordinary, showcasing the team’s commitment to pushing the boundaries of technology in medicine.
The pinnacle of their success lay in the system's performance on the cutting-edge BRATS 2022 dataset. The HCNN technique boasted an unparalleled 99.81% classification accuracy, 99.2% sensitivity, 99.7% specificity, and 99.8% segmentation accuracy. Finally, the team diligently compared the HCNN algorithm with state-of-the-art meningioma detection algorithms, firmly establishing their creation as the new benchmark in the field.
According to the professor, conducting this research was a collaborative effort that involved a dedicated team working on the forefront of medical imaging technology. She explained, “The successful development and publication of the novel HCNN method, integrating the Ridgelet transform, were the result of rigorous experimentation and analysis on diverse datasets, such as BRATS 2019, Nanfang, and BRATS 2022. Moving forward, our aim is to further refine and extend this approach, contributing to the ongoing efforts to automate meningioma detection and potentially reshape healthcare practices through advanced and accurate diagnostic tools.”
Professor Dang's contribution not only advanced the field of medical imaging but also reflected the potential of artificial intelligence to reshape the future of healthcare.
In conclusion she stated, "The new model proposed in this paper, not only exhibited exceptional accuracy in meningioma detection but also showcased significant benefits for automated systems on modern medical platforms. With outstanding sensitivity, specificity, and segmentation accuracy across diverse datasets, the HCNN approach presented in this study holds promise for advancing the efficiency and reliability of meningioma tumor detection, offering a valuable contribution to the field of medical image analysis."