Projects
Our lab harnesses advanced deep learning and image processing techniques to push the boundaries of medical diagnostics. Through multidisciplinary projects, we develop cutting-edge solutions that enhance image quality, automate feature detection, and improve early disease diagnosis, paving the way for smarter, more accurate clinical decision-making.
Featured
Pre-trained U-Net models with EfficientNetB0, VGG16, and DenseNet121 backbones were fine-tuned on a large dataset of CEM images with separate validation and independent testing sets, and evaluated using IoU and Dice scores over multiple training epochs.
A novel TVHC approach leveraging radiomic skewness masks, morphological operations, and chest wall segmentation on annotated mammograms to detect hotspots, with classification performed using multiple deep learning models and AUROC-based performance comparisons.
More
A novel deep learning framework leveraging multi-scale CNN architectures and adversarial training to reconstruct high-resolution MRI images from low-resolution inputs, with enhanced feature fusion and perceptual loss optimization, and performance validated using quantitative metrics.
A novel deep learning framework leveraging spatio-temporal modeling and optical flow estimation to correct for motion artifacts in contrast-enhanced ultrasound imaging, enhancing image clarity and diagnostic reliability.
A novel multimodal deep learning framework integrating MRI-based neuroimaging with retinal biomarker analysis to facilitate early Alzheimer’s diagnosis through advanced feature extraction and fusion techniques.