Enhanced Systems and Methods for Phase Unwrapping in Dense MRI using Deep Learning

The present inventive concept relates to improved systems and methods for phase unwrapping in dense MRI using deep learning, addressing the limitations of existing techniques in regions with high displacement gradients and improving the precision of strain analysis.

The original patent, 'Systems and methods for phase unwrapping for dense MRI using deep learning', introduced a novel approach for myocardial strain imaging. However, the existing method may struggle with phase unwrapping errors in regions with high displacement gradients, leading to reduced accuracy in strain analysis. The present inventive concept addresses this limitation by introducing novel multi-scale approaches, hybrid optimization techniques, and robust training methods to improve the accuracy and efficiency of phase unwrapping.

The present inventive concept comprises a system for phase unwrapping in dense MRI using deep learning, which utilizes a novel multi-scale approach to improve the accuracy of phase unwrapping in regions with high displacement gradients. This is achieved by incorporating a deep learning-based algorithm that corrects for phase wrapping errors, thereby improving the precision of strain analysis. Additionally, the system employs a hybrid approach combining deep learning and traditional optimization techniques to reduce computational time and improve accuracy. The inventive concept also includes a novel regularization technique to reduce noise and improve image quality in post-processing. Furthermore, a method for training a deep learning model for phase unwrapping in dense MRI is disclosed, comprising generating a large dataset of synthetic phase images with varying levels of noise and displacement, and using the dataset to train the model to be more robust to real-world imaging conditions.

The present inventive concept introduces a novel multi-scale approach, hybrid optimization techniques, and robust training methods that are not disclosed in the original patent. These innovations provide an inventive step over the existing art, enabling more accurate and efficient phase unwrapping in dense MRI.

Alternative embodiments of the inventive concept may include using different deep learning architectures, such as convolutional neural networks (CNNs) or recurrent neural networks (RNNs), to improve phase unwrapping accuracy. Additionally, the inventive concept may be adapted for use in other medical imaging modalities, such as computed tomography (CT) or ultrasound.

The present inventive concept has significant commercial potential in the medical imaging industry, particularly in the field of cardiac strain imaging. The improved accuracy and efficiency of phase unwrapping enabled by the inventive concept may lead to increased adoption of dense MRI in clinical practice, resulting in improved patient outcomes and reduced healthcare costs.