Down's syndrome dermatoglyph auxiliary screening constructed on basis of machine learning algorithm
A Down syndrome and screening technology, applied in the field of medical diagnosis, can solve problems such as inability to calculate accuracy and inability to reflect
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0122] Embodiment 1, dividing training set and independent verification set
[0123] Randomly screen 5% (about 11 cases) of Down syndrome patients and 5% (about 40 cases) of normal control samples to form an independent verification data set (Test set) for machine learning, and the remaining samples (about 1005 cases) are used as model construction The training data set (Trainng set). When randomly sampling the sample, the sex ratio of the sample was taken into account.
Embodiment 2
[0124] Embodiment 2, training set feature variable screening
[0125] The 56 skin texture features of 1005 samples were screened for important feature variables, following the figure 1 Middle (in the dotted box) important feature screening process:
[0126] 1. Use Chi-square test (categorical variable) and mean t test (quantitative variable) to carry out the difference test in the Down syndrome case group (case group) and normal control (control), and there is no significant difference between the case group and the control group feature item;
[0127] 2. Use the XGBoost machine learning method to rank the importance of the feature items with significant differences between the case group and the control group obtained in the first step, and retain the feature items with a cumulative importance of more than 99%;
[0128] 3. Retain the correlation between the feature items in step 2, and remove the items with lower importance ranking among the feature items with higher corr...
Embodiment 3
[0130] Embodiment 3, independent verification set verification
[0131] Using XGBoost to construct the optimal model for the training set that can distinguish between case samples and control samples, apply it to 51 independent verification set test sets, and calculate the average of the optimal combination of feature items for screening Down syndrome samples in the independent verification set Accuracy (Accuracy), True Positive Rate (TPR, True Positive Rate), False Negative Rate (FNR, FalseNegative Rate) and other indicators.
[0132] In addition, support vector machine (SVM, supportvector machine) (Suykens and Vandewalle, 1999) and linear discriminant analysis (LDA, Linear Discriminant Analysis) (Mika et al., 1999) were also used to verify the robustness of the model (Robust). The effectiveness of the combination of characteristic variables in independent samples in screening for Down syndrome was measured.
[0133] result
[0134] 1. The first comprehensive map of the d...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More