Vascular Imaging Decision Transparency via Feature Importance

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current vascular imaging technologies, such as IVUS and OCT, face challenges in accurately evaluating coronary plaque and calcium burden due to unrecognized image artefacts from motion, contrast agent variations, and existing implants, which can lead to erroneous decision-making in data-driven systems.

Innovation Solution

A system that determines the relative importance of each image feature in vascular medical images by using a medical image database, rule generating unit, diagnostic metric computation unit, and decision propagation unit to compute an overall diagnostic metric, allowing users to validate and adjust the impact of image features, and generate revised diagnostic rules based on user input.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If data-driven decision support systems are used to automatically evaluate vascular images, then diagnostic efficiency is improved, but transparency and understandability of the decision-making process deteriorates

Engineering Contradiction:

Improvediagnostic efficiencyVSAvoidtransparency of decision-making

Core Design Contradiction:

ProductivityVSLoss of information

Solution Approach 1:

The system implements feedback mechanisms by visualizing the impact of individual image features on the overall diagnostic metric through saliency maps and feature importance rankings. This allows clinicians to see how the system arrived at its conclusions and verify the reasoning, thereby maintaining transparency while preserving automated diagnostic efficiency.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The diagnostic evaluation is segmented into individual image features (e.g., plaque characteristics, calcium burden, stenosis measurements) rather than treating the image as a whole. Each feature is independently analyzed and its contribution to the overall diagnosis is quantified and displayed, providing transparency about which features drove the decision while maintaining automated processing speed.

Inventive Principle:

Principle #1Segmentation

2Stability of the object's composition

If automatically generated diagnostic rules are applied to vascular images, then diagnostic consistency is improved, but reliability deteriorates due to unrecognized image artefacts

Engineering Contradiction:

Improvediagnostic consistencyVSAvoidaccuracy of diagnostic metric

Core Design Contradiction:

Stability of the object's compositionVSReliability

Solution Approach 1:

The system provides feedback to clinicians by highlighting potentially problematic image regions and displaying the relative importance of different features. This enables clinicians to review the automated analysis and correct errors caused by unrecognized artefacts, thereby maintaining reliability while preserving consistent automated diagnostic rules.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system allows dynamic adjustment of diagnostic parameters and feature weights based on clinical context. Clinicians can modify the relative importance of different image features or adjust diagnostic thresholds when artefacts are detected, enabling the system to adapt to specific patient conditions while maintaining overall diagnostic consistency through standardized rules.

Inventive Principle:

Principle #35Parameter changes

3Device complexity

If the relative importance of image features is not disclosed, then system complexity is reduced, but ease of operation deteriorates due to inability to validate decisions

Engineering Contradiction:

Improvesystem complexityVSAvoidability to validate decisions

Core Design Contradiction:

Device complexityVSEase of operation

Solution Approach 1:

The system segments the diagnostic output into individual feature contributions, displaying the relative importance of each image feature separately. This segmentation provides operational transparency and validation capability without requiring complete system complexity to be disclosed or understood by the user, thereby maintaining ease of operation while managing complexity through structured presentation.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS11721439B2Resolving and steering decision foci in machine learning-based vascular imaging

Publication Date: 2023.08.08 KONINKLIJKE PHILIPS NV

AI summary

A system (SY) for determining a relative importance of each of a plurality of image features (Fn) of a vascular medical image impacting an overall diagnostic metric computed for the image from an automatically-generated diagnostic rule. A medical kin image database (MIDB) includes a plurality of vascular medical images (M1 . . . k). A rule generating unit (RGU) analyzes the plurality of C vascular medical images and automatically generates at least one diagnostic rule corresponding to a common diagnosis of a subset of the plurality of vascular medical images based on a plurality of image features common to the subset of vascular medical images. An image providing unit (IPU) provides a current vascular medical image (CVMI) including the plurality of image features. A diagnostic metric computation unit (DMCU) computes an overall diagnostic metric for the current vascular medical image by applying the at least one automatically-generated diagnostic rule to the current vascular medical image. A decision propagation unit (DPU) identifies, in the current vascular medical image, the relative importance of each of the plurality of image features on the computed overall diagnostic metric.