A5027255310- Advanced Neural Network Applications
- Medical Image Segmentation Techniques
- Brain Tumor Detection and Classification
- COVID-19 diagnosis using AI
- Radiomics and Machine Learning in Medical Imaging
- Domain Adaptation and Few-Shot Learning
- Advanced Image and Video Retrieval Techniques
- AI in cancer detection
- Robotic Path Planning Algorithms
- Medical Imaging and Analysis
- Infrastructure Maintenance and Monitoring
- Robotics and Automated Systems
- Medical Imaging Techniques and Applications
- Generative Adversarial Networks and Image Synthesis
Monash University
2021-2024
Precise, and automated segmentation of construction demolition waste (CDW) is crucial for recognizing the composition mixed streams facilitating automatic sorting. Training a neural network image challenging due to time resource-intensive nature annotating large-scale datasets, particularly domain-specific recognition in cluttered environments. In this paper, we propose semi-supervised multi-class approach recognize CDW real-world settings, utilizing an adversarial dual-view framework. doing...
Segmentation of images is a long-standing challenge in medical AI. This mainly due to the fact that training neural network perform image segmentation requires significant number pixel-level annotated data, which often unavailable. To address this issue, we propose semi-supervised technique based on concept multi-view learning. In contrast previous art, introduce an adversarial form dual-view and employ critic formulate learning problem as min-max problem. Thorough quantitative qualitative...
Robots are closer than ever to leap from the defined world navigation undefined navigation. For that it is crucial have flexible and alternative platforms enhance mobile self-made robots next level. In this project a vision based robot developed using Microsoft Windows software as an most common Linux Robot Operating System (ROS). Currently almost all with advanced functionalities use ROS. But what lacks in system flexibility for beginners. Developing on platform was not possible few years...
Motion artifacts in Magnetic Resonance Imaging (MRI) are one of the frequently occurring due to patient movements during scanning. is estimated be present approximately 30% clinical MRI scans; however, motion has not been explicitly modeled within deep learning image reconstruction models. Deep (DL) algorithms have demonstrated effective for both task and correction task, but two tasks considered separately. The involves removing undersampling such as noise aliasing artifacts, whereas...
Accurate hippocampus segmentation in brain MRI is critical for studying cognitive and memory functions diagnosing neurodevelopmental disorders. While high-field MRIs provide detailed imaging, low-field are more accessible cost-effective, which eliminates the need sedation children, though they often suffer from lower image quality. In this paper, we present a novel deep-learning approach automatic of bilateral hippocampi MRIs. Extending recent advancements infant to underserved communities...
As intensities of MRI volumes are inconsistent across institutes, it is essential to extract universal features multi-modal MRIs precisely segment brain tumors. In this concept, we propose a volumetric vision transformer that follows two windowing strategies in attention for extracting fine and local distributional smoothness (LDS) during model training inspired by virtual adversarial (VAT) make the robust. We trained evaluated network architecture on FeTS Challenge 2022 dataset. Our...
We propose a Transformer architecture for volumetric segmentation, challenging task that requires keeping complex balance in encoding local and global spatial cues, preserving information along all axes of the volume. Encoder proposed design benefits from self-attention mechanism to simultaneously encode while decoder employs parallel self cross attention formulation capture fine details boundary refinement. Empirically, we show choices result computationally efficient model, with...
This paper proposes an adversarial learning based training approach for brain tumor segmentation task. In this concept, the 3D network learns from dual reciprocal approaches. To enhance generalization across predictions and to make robust, we adhere Virtual Adversarial Training by generating more examples via adding some noise on original patient data. By incorporating a critic that acts as quantitative subjective referee, uncertainty information associated with results. We trained evaluated...