Machine Learning Statistical Analysis Of Lumbar Disc Degeneration Based On MRS Metabolomics

The lumbar spine is the most sensitive in the human aging process,which can best reflect the changes of human functions.In this paper,we examine the biochemical metabolic data of human lumbar discs based on MRI technology,construct a classifier of lumbar disc degeneration levels using machine learning methods.It also provides a picture of the lumbar intervertebral discs in order to assist physicians in the study of lumbar disc degeneration.Firstly,this paper introduces the basic data processing methods such as Kruskal-Wallis rank sum test and grid search,and details the models used in constructing the lumbar disc degeneration grade classifier including multinomial logistic regression,support vector machine and gradient boosting decision tree,outlining their implementation principles and optimization processes.Secondly,the biometabolic indicators of lumbar discs measured by MRS were selected,and the valid indicators were screened from the available data,and the Kruskal-Wallis rank sum test was applied to discuss the Pfirrmann degeneration grade,age group and lumbar disc degeneration degree.Spearman correlation analysis was used to correlate biochemical metabolic indices with Pfirrmann’s degeneration grade,age and the degree of lumbar disc degeneration.The results showed that all six metabolic indices in the data set were significantly correlated with the degree of lumbar disc degeneration.Then,we used these indicators to construct multinomial logistic regression,support vector machine and gradient boosting decision tree classifier models on the training set,and optimized the training to select the optimal parameters of the corresponding classifiers,and compared the performance of the three LIDD diagnostic classifiers on the test set.Furtherly,drawing on the idea of user profiling for analysing user behaviour,the lumbar disc is profiled.The degenerated lumbar disc data were divided into biochemical metabolic features and patient attribute features,and the data label structure was designed accordingly.The K-means clustering method was used to cluster the degenerated lumbar discs based on biochemical indicators,and the results of the clustering were combined with the personal attributes of the patients corresponding to each lumbar disc to investigate whether the lumbar discs with different metabolic information have distinctive group characteristics.Finally,the results of this paper are summarized and shortcomings are suggested,and the future research on lumbar disc degeneration is expected.