Dynamic Analysis Device Noise Filtering for Lung Imaging

The patent segments the noise removal process into multiple filtering stages operating in different feature quantity spaces. First filtering is applied in spatial frequency space, followed by second filtering in time frequency space, and optionally third filtering in another feature quantity space. This multi-stage segmentation allows progressive noise removal while preserving useful signal components, resolving the contradiction between ease of imaging during breathing and analysis accuracy.

The patent transitions from single-dimension filtering to multi-dimensional filtering by operating in multiple feature quantity spaces (spatial frequency, time frequency, and other feature spaces). This dimensional expansion enables comprehensive noise separation that accounts for both spatial patterns and temporal characteristics of noise, thereby maintaining analysis accuracy even when imaging during breathing without breath holding.

The filtering process is segmented into distinct stages, each operating in a different feature quantity space with specific filtering parameters. First filtering in spatial frequency space removes spatial noise patterns, second filtering in time frequency space removes temporal noise patterns, and optional third filtering in additional feature spaces provides further noise removal. This segmentation enables high noise removal accuracy while keeping each individual filtering stage relatively simple.

The patent introduces multiple feature quantity spaces as intermediary domains between the raw dynamic image data and the final analysis results. Each feature quantity space serves as an intermediary that transforms the data into a representation where specific types of noise can be effectively separated. These intermediary spaces (spatial frequency, time frequency, and other feature spaces) enable accurate noise removal without requiring complex direct processing of the original data.