Automatic Tomography Slice Selection for Radiology
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Radiologists face difficulties in efficiently reviewing and segmenting regions of interest across a large stack of 2D cross-sectional slice images, as existing tools require tedious scrolling and manual interaction to find the most suitable slice for segmentation.
Innovation Solution
An apparatus comprising a graphical user interface controller, a segmenter unit, and a goodness of segmentation evaluator that allows users to specify a segmentation seed point interactively, automatically segmenting multiple slices, evaluating the goodness of segmentation, and displaying the best slice with the highest segmentation value, thereby reducing user interaction and ambiguity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual scrolling through multiple slices is used to find the best slice for segmentation, then the radiologist can review all slices, but the process becomes tedious and time-consuming
Solution Approach 1:
The system performs automatic slice selection and segmentation evaluation without requiring continuous manual intervention. The radiologist provides initial input (seed point or region), and the system autonomously evaluates multiple slices, segments them, and presents the optimal result, making the system serve itself in the evaluation process
Solution Approach 2:
The system pre-evaluates multiple slices by performing segmentation on each one before the radiologist needs to select the best slice. This preliminary segmentation and evaluation of all candidate slices eliminates the need for manual scrolling and allows direct presentation of the optimal slice
2Measurement precision
If segmentation is performed on multiple slices manually, then the best slice can be identified, but the complexity of handling segmentation tools increases
Solution Approach 1:
The system merges the segmentation function with the slice evaluation and selection process. Instead of using separate segmentation tools for each slice, the segmentation capability is integrated into the automated workflow that evaluates multiple slices, combining these functions into a unified system
Solution Approach 2:
The system uses temporary, automated segmentation instances for each slice evaluation that are discarded after evaluation. Rather than maintaining complex persistent segmentation tools, the system creates and destroys segmentation objects automatically for each slice, simplifying the overall tool complexity
3Measurement precision
If all segmented slices are displayed to the radiologist, then the best slice can be selected, but information overload occurs
Solution Approach 1:
The system extracts only the most important information (the single best slice with highest segmentation quality) from the set of all segmented slices and presents it to the radiologist. Instead of displaying all slices, the system separates and highlights only the optimal result, reducing information overload while maintaining evaluation accuracy
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An apparatus for automatic selection of optimal tomography slices by executing a number of tentative segmentations using the same interactively provided in-slice seed point on some or all available tomography slices. The appropriate segmentation boundaries are then marked and the slice with the best segmentation goodness value (figure of merit) is presented to the user via a viewer. The steps are repeated when the user changes the seed point. The optimal segmentation boundary is displayed on top of a single simulated mammography image, fused from all tomography slice images.