Ghost Cytometry Classification Model Using Positive Waveform Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional flow cytometry methods using the ghost cytometry (GC) method struggle to create effective classification models for identifying cells with specific morphological characteristics due to the difficulty in preparing training data that reflects the diversity of negative cells, which often have varying and unspecified morphological characteristics.
Innovation Solution
A classification model generation method that uses first waveform data from particles with specific morphological characteristics and second waveform data from a mixture of unspecified particles, along with a positive rate indicating the proportion of particles with specific morphological characteristics, to train a classification model that can identify whether a particle has the specific morphological characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised machine learning is used to create a classification model with waveform data from both positive and negative cells, then classification accuracy can be improved, but it becomes impractical to prepare training data reflecting the diversity of negative cells
Solution Approach 1:
The patent extracts only the positive cell waveform data from the training set, removing the need to collect and prepare diverse negative cell data. The classification model is trained using solely positive cell waveforms, which are then applied to classify both positive and negative cells in the test sample, thereby simplifying the data preparation process while maintaining classification capability
Solution Approach 2:
The patent creates a classification model that learns the characteristics of positive cells from their waveform data, and uses this learned model to identify positive cells in new samples. The model copies the morphological patterns of positive cells during training and applies this knowledge to classify unknown cells, eliminating the need for negative cell training data
2Adaptability or versatility
If a mixture of unspecified cells is created as training sample to reflect morphological diversity, then training coverage can be improved, but it becomes difficult to perform training due to impracticality of sample creation
Solution Approach 1:
The patent extracts and uses only the positive cell subset from the mixed cell population for training purposes. By focusing exclusively on positive cell waveform data, the method achieves comprehensive training coverage for the target phenotype without requiring the complex preparation of diverse negative cell samples
Solution Approach 2:
The classification model serves itself by learning from positive cell characteristics alone and using this self-acquired knowledge to identify positive cells in test samples. The model does not require external negative cell data to define what it is looking for, as it learns the positive phenotype patterns independently
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This method enables the generation of a classification model that can accurately identify particles with specific morphological characteristics even when training data from negative cells with diverse morphological characteristics is limited or impractical to obtain.
Implementation Method 1
irradiating light to each cell moving through the channel, and measuring scattered light or fluorescence from the cells irradiated with light
Implementation Method 2
measuring scattered light or fluorescence from the cells irradiated with light
Data Source
AI summary
A classification model that outputs identification information indicating whether or not a particle has specific morphological characteristics when waveform data is input is generated by training using training data including first waveform data that is obtained by irradiating light to particles contained in a first sample and having specific morphological characteristics and indicates morphological characteristics of the particles, information indicating that the first waveform data has been obtained from particles contained in the first sample, second waveform data indicating morphological characteristics of unspecified particles contained in a second sample, information indicating that the second waveform data has been obtained from particles contained in the second sample, and a positive rate that is the proportion of particles having the specific morphological characteristics.