Ipris 3D Nodule Features for Lung CT Diagnosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing approaches for distinguishing adenocarcinoma from granuloma on lung CT scans are sub-optimal, as they primarily focus on features within the nodule and are affected by scanner settings, leading to unnecessary surgical procedures due to inability to confidently diagnose benign vs. malignant nodules.
Innovation Solution
The use of intra-perinodular textural transition (Ipris) features, which capture the transitional heterogeneity from the inside to the outside of the nodule, by partitioning the nodule into nested shells and extracting features from 2D slices, to train a support vector machine (SVM) classifier for accurate differentiation between benign and malignant nodules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional radiomic features (shape, texture, intensity) are used for nodule characterization, then the diagnostic process is simple, but the accuracy of distinguishing benign from malignant nodules is insufficient
Solution Approach 1:
The patent segments the nodule into multiple nested shells (core, middle, peripheral regions) and analyzes textural transitions between these shells. This segmentation approach captures the spatial heterogeneity of the nodule, enabling more accurate differentiation between benign and malignant nodules by examining how texture properties change from the core to the periphery, rather than treating the nodule as a homogeneous whole.
Solution Approach 2:
The patent transitions from conventional 2D slice-based analysis to 3D volumetric analysis by computing Ipris features across multiple nested shells. This dimensional expansion allows capture of spatial textural transitions in three dimensions, providing richer diagnostic information about nodule heterogeneity and improving classification accuracy.
2Productivity
If existing texture-based features are extracted from segmented nodules, then the analysis is computationally efficient, but the features are affected by scanner settings and reconstruction parameters
Solution Approach 1:
The patent computes Ipris features by analyzing intensity transitions across nested shells at different radial distances from the nodule center. This parameter-based approach (measuring transition intensity, gradient magnitude, and spatial frequency across shells) creates features that are inherently more robust to scanner variations, as they capture relative textural patterns rather than absolute intensity values that vary with reconstruction settings.
3Device complexity
If margin sharpness is used to detect lymphocytic infiltration, then the analysis is simple, but it only examines the nodule interface and ignores the tumor core
Solution Approach 1:
The patent divides the nodule into nested shells (core, middle, peripheral regions) and computes textural transition features at each shell boundary. This segmentation enables examination of both the nodule interface (peripheral shell) and internal regions (core and middle shells), capturing lymphocytic infiltration patterns throughout the entire nodule volume rather than only at the margin.
Solution Approach 2:
The patent extends the analysis from 2D margin examination to 3D volumetric analysis by computing Ipris features across nested shells in three dimensions. This captures spatial textural transitions throughout the entire nodule volume, including the core region, providing comprehensive information about internal heterogeneity and lymphocytic infiltration patterns.
Data Source
AI summary
Embodiments classify lung nodules by accessing a 3D radiological image of a region of tissue, the 3D image including a plurality of voxels and slices, a slice having a thickness; segmenting the nodule represented in the 3D image across contiguous slices, the nodule having a 3D volume and 3D interface, where the 3D interface includes an interface voxel; partitioning the 3D interface into a plurality of nested shells, a nested shell including a plurality of 2D slices, a 2D slice including a boundary pixel; extracting a set of intra-perinodular textural transition (Ipris) features from the 2D slices based on a normal of a boundary pixel of the 2D slices; providing the Ipris features to a machine learning classifier which computes a probability that the nodule is malignant, based, at least in part, on the set of Ipris features; and generating a classification of the nodule based on the probability.