Sequential Medical Image Analysis Using Difference Images for Diagnosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining clinical parameters from sequential medical images, such as MRI and CT scans, are inefficient and lack accuracy in diagnosing conditions like soft-tissue injuries and bone fractures, particularly in orthopedic imaging.
Innovation Solution
A deep learning model utilizing convolutional neural networks (CNNs) and long short-term memory networks (LSTM) processes difference images generated from sequential medical images to improve the diagnosis of conditions by analyzing changes between successive images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional methods are used to determine clinical parameters from sequential medical images, then the process is simpler, but the diagnostic accuracy is lower (86.6%)
Solution Approach 1:
The patent segments the sequential medical images by generating difference images that highlight changes between adjacent frames. This segmentation transforms the original image sequence into a set of difference images that emphasize temporal changes, enabling the CNN to focus on relevant diagnostic features while improving accuracy to 93.3%
Solution Approach 2:
The patent applies preliminary action by pre-processing the sequential images to generate difference images before feeding them to the predictive model. This preliminary transformation of the data into a more informative representation (difference images) enhances the input quality for the CNN, thereby improving diagnostic accuracy without requiring changes to the core model architecture
2Measurement precision
If difference images are generated and processed through CNN, then diagnostic accuracy improves to 93.3%, but the processing time and computational complexity increase
Solution Approach 1:
The patent extracts the essential diagnostic information by generating difference images that isolate changes between sequential frames. This extraction process removes redundant static information and highlights only the dynamic changes relevant to diagnosis, improving accuracy while potentially reducing the computational burden by focusing on fewer, more informative features
3Reliability
If multiple convolutional layers are used in the predictive model, then feature extraction capability is enhanced, but the device complexity increases
Solution Approach 1:
The patent applies local quality by using CNNs with multiple convolutional layers that can detect local patterns and features at different levels of abstraction. The convolutional layers progressively extract more complex features from the difference images, with early layers detecting simple edges and textures, and deeper layers identifying more complex diagnostic patterns, thereby enhancing diagnosis reliability
Data Source
AI summary
Systems and methods are provided for determining a clinical parameter via evaluation of sequential medical images. A sequence of at least three medical images for a patient are captured at a scanner. A set of at least two difference images are generated from the sequence of at least three medical images. Each difference image represents a difference in content between two adjacent images in the sequence of at least three medical images. The set of at least two difference images are provided to a predictive model. The predictive model includes an artificial neural network having at least one convolutional layer. A clinical parameter for the patient is determined at the predictive model from at least the set of at least two difference images.