Pulmonary Nodule Segmentation via Distance Map Region Growing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current segmentation algorithms for pulmonary nodules in medical images struggle with accuracy, particularly for non-solid and part-solid nodules, which are frequent and potentially malignant, due to their subtle appearance and irregular intensity variations, and often require iterative or scale-space representations, making them computationally inefficient and limited in handling convex nodules.
Innovation Solution
A semi-automated segmentation algorithm that computes a distance map and uses a two-step region growing process to differentiate between nodule and non-nodule regions, employing a sphericity measure to locate the nodule core and grow regions while being insensitive to attached structures, thus handling solid, non-solid, and part-solid nodules without estimating nodule size, and refining segmentation through a competition process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing segmentation algorithms are used for pulmonary nodules, then solid nodules can be segmented, but non-solid and part-solid nodules cannot be accurately segmented due to their subtle appearance and irregular intensity variations
Solution Approach 1:
The patent transforms the original image intensity values into a distance map representation, where each pixel value represents the distance to the nearest background point. This parameter transformation makes the segmentation process insensitive to the original intensity variations that cause difficulties with non-solid and part-solid nodules, enabling accurate segmentation across all nodule types.
Solution Approach 2:
The patent introduces a distance map as an intermediary representation between the original image and the final segmentation. This intermediate structure captures the geometric relationship between foreground and background, serving as a mediator that enables the region growing algorithm to successfully segment various nodule types regardless of their intensity characteristics.
2Measurement precision
If iterative methods or scale-space representations are used to improve segmentation accuracy, then various density types can be handled, but computational efficiency decreases
Solution Approach 1:
The patent performs a preliminary transformation of the image into a distance map before segmentation. This pre-processing step captures all necessary geometric information in a single computation, eliminating the need for iterative refinement or multi-scale analysis, thereby achieving both high accuracy and computational efficiency.
Solution Approach 2:
The patent extracts the essential geometric information from the original image by computing the distance map, separating the segmentation task from the intensity variations. This extraction allows the algorithm to focus solely on boundary detection without being distracted by intensity heterogeneity, improving both accuracy and efficiency.
3Reliability
If region growing is used to segment nodules, then the nodule can be extracted from attached structures, but the process requires size estimation which increases computational complexity
Solution Approach 1:
The patent modifies the region growing algorithm to automatically detect the nodule boundary by monitoring when the growing region first touches the background. This self-terminating mechanism eliminates the need for external size estimation parameters, allowing the algorithm to adapt to nodules of any size without increasing complexity.
Data Source
AI summary
A method for differentiating pulmonary nodules in digitized medical images includes identifying an object of interest from a digital image of the lungs, computing a first distance map of each point of the object of interest, determining a seed point from the first distance map, starting from the seed point, growing a first region by adding successive adjacent layers of points until a background point is reached, and partitioning the first region into a nodule region and a non-nodule region.