Blood Coagulation Curve Analysis Using Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing blood coagulation test methods struggle to accurately predict the cause of prolonged coagulation time due to limitations in input variables, leading to difficulties in distinguishing between similar changes caused by various prolongation factors.
Innovation Solution
A blood specimen analysis method utilizing a deep learning algorithm that processes blood coagulation curves and their derivatives to identify the cause of prolonged coagulation time, incorporating a device with a measurement unit and controller to prepare samples and input data into the algorithm for accurate prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a neural network is used with limited predictor variables, then the prediction process can be simplified, but prediction accuracy deteriorates when changes in predictor variables with respect to multiple prolongation causes are similar
Solution Approach 1:
The patent transforms the input data from limited discrete predictor variables to a comprehensive continuous optical profile dataset. By utilizing the entire optical profile curve and its derivatives (first derivative, second derivative) as input parameters, the system captures subtle variations in coagulation behavior that discrete variables miss, thereby improving prediction accuracy while maintaining computational feasibility through deep learning algorithms.
Solution Approach 2:
The patent adds temporal and derivative dimensions to the analysis by incorporating the optical profile across multiple time points and its mathematical derivatives. This transforms a static limited-variable approach into a dynamic multi-dimensional analysis, enabling the system to distinguish between similar prolongation causes through their unique temporal patterns and rate of change characteristics.
2Ease of operation
If specific predictor variables are selected for neural network input, then data processing becomes more manageable, but the ability to distinguish between similar changes caused by various prolongation factors is reduced
Solution Approach 1:
The patent performs preliminary data preparation by calculating the optical profile and its first and second derivatives before inputting to the deep learning model. This preprocessing step organizes the complex temporal data into meaningful features that highlight different aspects of coagulation behavior, making the subsequent analysis more effective while maintaining data processing manageability through systematic preparation.
Solution Approach 2:
The patent introduces mathematical derivatives as intermediary representations between the raw optical profile data and the final prediction. The first derivative captures the rate of change, while the second derivative captures the acceleration of change, serving as intermediate features that enhance the model's ability to distinguish between similar prolongation causes without directly processing the entire raw dataset.
Data Source
AI summary
Disclosed is an analysis method for a blood specimen, including: obtaining a data group including a plurality of data forming a blood coagulation curve or a differential curve thereof; inputting the data group into a deep learning algorithm; and outputting, on the basis of a result obtained from the deep learning algorithm, information regarding a cause of prolongation of blood coagulation time of the blood specimen.