Simulated 3D Radiological Datasets for AI Training

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current AI systems in diagnostic radiology face challenges in detecting small diagnostic features within large datasets, such as in MRI images, which limits their effectiveness in early cancer detection and treatment, and they lack optimization for clinical impact and performance quantification.

Innovation Solution

The method involves generating and using simulated 3D radiological datasets with adjustable tissue structures to train and test AI algorithms, allowing for the creation of realistic volumetric data that can mimic various medical conditions, including cancer, to improve detection accuracy and clinical relevance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If AI systems analyze large medical image datasets, then detection coverage is improved, but detection precision of small features deteriorates

Engineering Contradiction:

Improvedataset coverageVSAvoidsmall feature detection accuracy

Core Design Contradiction:

Area of stationary objectVSMeasurement precision

Solution Approach 1:

The system segments the large medical image dataset into multiple sub-volumes using a 3D cursor, allowing focused analysis of specific regions. This segmentation enables the AI to maintain high detection precision for small features within each sub-volume while still achieving comprehensive coverage of the entire dataset through systematic progression through multiple segments.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system transitions from 2D image analysis to 3D volumetric analysis by generating and analyzing 3D datasets. This dimensional change allows for better spatial context and improved detection of small features within the broader dataset, resolving the contradiction between coverage and precision.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Reliability

If simulated 3D datasets are generated to train AI algorithms, then training effectiveness is improved, but computational resource requirements increase

Engineering Contradiction:

ImproveAI training effectivenessVSAvoidcomputational resource consumption

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system generates simulated 3D datasets in advance to create comprehensive training corpora before actual AI model training. This preliminary action allows for thorough preparation of training data with known ground truth, improving training effectiveness while enabling efficient batch processing that reduces overall computational resource consumption during the training phase.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11238197B1Generating a 3D dataset containing a simulated surgical device

Publication Date: 2022.02.01 RED PACS LLC

AI summary

This patent includes a method and apparatus for the generation of a simulated, realistic medical device, which can be inserted into a 3D radiological dataset from CT, MRI, PET, SPECT or DTS examinations. This simulated dataset can be segmented, filtered, manipulated, used with artificial intelligence algorithms and viewed in conjunction with head display units and geo-registered tools.