Breast cancer remains the most prevalent cancer among women worldwide, underscoring the critical need for effective early detection methods. Thermography presents a non-invasive and radiation-free alternative, offering significant potential in breast cancer screening; however, challenges such as noise, low resolution, and anatomical variability limit its potential. The aim of this project is to propose a method for segmenting the breast in thermograms to analyze thermal variations that might be associated with medical pathologies. To address these issues, we developed a hybrid segmentation approach that combines traditional computer vision techniques with a Dense Multiscale U-Net architecture, leveraging anatomical references extracted directly from thermographic images. Our method integrates the adaptability of deep learning with the domain-specific precision of traditional computer vision techniques, creating a hybrid framework designed to overcome individual limitations and enhance segmentation robustness. Our method demonstrated substantial advancements in segmentation performance, achieving a Dice Similarity Coefficient (DSC) of 0.987 and an Intersection over Union (IoU) of 0.963. These results not only surpassed those of standalone computer vision methods (DSC: 0.983, IoU: 0.934) and neural networks (DSC: 0.972, IoU: 0.922), but also highlighted the robustness, reproducibility, and precision of the hybrid approach in addressing critical challenges. Notably, the method effectively delineates the weak supramammary boundary, a common limitation of other techniques, thereby enhancing its clinical relevance and reliability in diverse imaging scenarios. This approach enhances the precision and robustness of breast segmentation, particularly in challenging regions such as the supramammary boundary. By bridging the gap between traditional and deep learning methods, this hybrid framework represents a significant advancement in thermography-based breast segmentation and lays the foundation for improving diagnostic accuracy in breast cancer screening.
This paper presents a system for enabling real-time synthesis of whole-body locomotion and manipulation policies for real-world legged robots. Motivated by recent advancements in robot simulation, we leverage the efficient parallelization capabilities of the MuJoCo simulator to achieve fast sampling over the robot state and action trajectories. Our results show surprisingly effective real-world locomotion and manipulation capabilities with a very simple control strategy. We demonstrate our approach on several hardware and simulation experiments: robust locomotion over flat and uneven terrains, climbing over a box whose height is comparable to the robot, and pushing a box to a goal position. To our knowledge, this is the first successful deployment of whole-body sampling-based MPC on real-world legged robot hardware. Experiment videos and code can be found at: https://whole-body-mppi.github.io/
Breast cancer remains the most prevalent cancer among women worldwide, emphasizing the demand for accessible and accurate screening technologies. Infrared