SYSTEMS AND METHODS FOR SYSTEM AGNOSTIC AUTOMATED DETECTION OF CARDIOVASCULAR ANOMALIES AND/OR OTHER FEATURES

Smart overview of the Invention

The patent application describes systems and methods that utilize artificial intelligence and machine learning to process medical images from various imaging systems, such as ultrasounds or echocardiograms, to detect congenital heart defects (CHDs) and other cardiovascular anomalies. The technology is system-agnostic, meaning it can analyze data irrespective of the imaging system's type, software, or hardware. This approach aims to overcome biases that may exist in current models due to variations in image styles from different imaging systems.

Technical Field

The technology pertains to image processing systems that incorporate artificial intelligence to identify cardiovascular anomalies. It addresses the limitations of current medical imaging techniques, which, despite their capability to visualize internal anatomy, often miss anomalies due to human error or insufficient training. The described system seeks to enhance the accuracy of anomaly detection by processing images in a way that is not influenced by the specific characteristics of the imaging system used.

Background

Medical imaging, such as ultrasound, allows healthcare providers to examine internal anatomy and identify abnormalities without surgery. However, congenital heart defects, particularly in prenatal imaging, are frequently undetected due to subtle visual cues and practitioner limitations. Existing AI systems face challenges like bias from training data that may favor certain imaging systems, leading to inaccurate results. This new approach aims to mitigate such biases by using styled images for analysis.

Methodology

The method involves analyzing medical images styled to incorporate data from multiple imaging systems, thus reducing bias in anomaly detection. A style transfer generator processes the input images and generates styled versions that reflect various imaging styles. These styled images are then analyzed to detect cardiovascular anomalies. The system can also be trained using styled images that represent a standard style, enhancing its ability to recognize anomalies consistently across different imaging systems.

System and Implementation

The system comprises memory and processors configured to execute instructions for processing image data from different imaging systems. It determines style groups and representative images, processes new image data using a style transfer generator, and analyzes the styled images to assess the likelihood of cardiovascular anomalies. This system allows for the detection of anomalies in a manner that is not dependent on the specific imaging system, thus providing a more reliable diagnostic tool for healthcare providers.