Spatio-Temporal Image Analysis for Accurate Endoscopy Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing computer-implemented systems for medical image analysis, such as endoscopy and capsule endoscopy, struggle to accurately detect features of interest and efficiently analyze image sequences, leading to inefficient and potentially erroneous medical diagnoses.

Innovation Solution

A computer-implemented system utilizing convolutional neural networks (CNN) for intelligent image analysis, including local and global spatio-temporal processing modules, and a timeseries analysis module to identify features of interest in medical images, providing real-time feedback and post-procedure reports.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If manual image examination is performed by physicians, then diagnostic accuracy can be maintained through human judgment, but the process becomes time-consuming and inefficient

Engineering Contradiction:
Improvediagnostic accuracyVSAvoidexamination time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces the mechanical system of manual image examination with an automated computer vision system using deep learning algorithms. The system processes medical images through neural networks that automatically detect and classify features of interest, eliminating the need for time-consuming manual review while maintaining diagnostic accuracy through sophisticated pattern recognition capabilities.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent creates a virtual copy of the human diagnostic process through trained neural networks that replicate expert physician judgment. The system learns from labeled training datasets containing numerous medical images with expert annotations, effectively copying human diagnostic reasoning into an automated algorithm that can process images rapidly without sacrificing accuracy.

Inventive Principle:
Principle #26Copying

2Loss of information

If a large number of images are collected during medical procedures, then comprehensive diagnostic information is obtained, but the volume of data makes manual review impractical

Engineering Contradiction:
Improvediagnostic information completenessVSAvoidimage processing efficiency
Core Design Contradiction:
Loss of informationVSProductivity

Solution Approach 1:

The patent extracts only the most relevant diagnostic information from large volumes of medical images using automated detection algorithms. The system identifies and extracts features of interest such as lesions, polyps, or abnormalities, separating critical diagnostic data from redundant normal tissue images, thereby maintaining information completeness while dramatically improving processing efficiency.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments the large dataset of medical images into meaningful groups based on detected features and characteristics. The system divides images into categories such as normal, abnormal, or requiring further review, and processes them in organized sequences, enabling efficient handling of large volumes while ensuring comprehensive diagnostic coverage through systematic analysis.

Inventive Principle:
Principle #1Segmentation

3Extent of automation

If extant computer-implemented systems are used for image analysis, then automation is provided, but the systems lack accuracy in detecting features of interest

Engineering Contradiction:
Improveanalysis automationVSAvoidfeature detection accuracy
Core Design Contradiction:
Extent of automationVSMeasurement precision

Solution Approach 1:

The patent improves feature detection accuracy by optimizing multiple parameters including neural network architecture, learning rates, batch sizes, and data augmentation strategies. The system adjusts these parameters based on the specific medical imaging modality and diagnostic task, enabling high-precision automated detection while maintaining full automation. Training on diverse datasets with varied parameters enhances the system's ability to accurately detect subtle features of interest.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4552076B1Computer-implemented systems and methods for intelligent image analysis using spatio-temporal information
Publication Date: 2026.03.11 COSMO ARTIFICIAL INTELLIGENCE AI LTD
  • EP4552076B1 patent drawingFigure 1A
  • EP4552076B1 patent drawingFigure 1B
  • EP4552076B1 patent drawingFigure 2

AI summary

A computer-implemented method is provided for detecting at least one feature of interest in images captured with an imaging device. The method includes receiving an ordered set of images from the captured images, the ordered set of images being temporally ordered and analyzing one or more subsets of the ordered set of images using a local spatio-temporal processing module, the local spatio-temporal processing module being configured to determine the presence of characteristics related to the at least one feature of interest in each image of each subset of images and to annotate the subset of images based on the determined characteristics in each image of each subset of images. The method further includes processing a set of feature vectors of the ordered set of images using a global spatio-temporal processing module, the global spatio-temporal processing module being configured to refine the determined characteristics associated with each subset of images, and calculating one or more values for each image using a timeseries analysis module, the numerical value being representative of the at least one feature of interest and calculated using the refined characteristics associated each subset of images and spatio-temporal information. Still further, the method may include generating a report, a data or electronic file, integration into another reporting system or electronic medical records, and/or generating an electronic display on the at least one feature of interest using the multiple values associated with each image of each subset of the ordered set of images.