Predictive Study Of Smoking Behavior Based On Diffusion Tensor Imaging

Smoking causes many diseases,including cancer,and smoking is one of the most serious public health problems in the world.At present,China has about 300 million smokers and has become the world’s largest tobacco producer and consumer.Numerous studies have shown that smoking can cause changes in brain cognition,structure,and function,and the adaptive effect of this change is the key to the formation of addiction.It is worth noting that the ageing of smoking in China is getting worse.The earlier a smoker starts smoking,the greater the harm to the body,the higher the chance of becoming a lifelong smoker,and the lower the success rate of quitting.Therefore,in-depth study of the neural mechanisms of adolescent smoking addicts is very important for prevention and treatment of adolescent smoking cessation.Most of the current imaging studies based on smoking addiction use data analysis based on group comparisons.However,this method is limited to group-level inference.The results of research are often of limited value in clinical diagnosis and treatment.Compared with this univariate method,the multi-modal analysis method including machine learning is highly sensitive to the differences in the slight spatial patterns of the brain.Using this method for individual prediction can be better applied to clinical diagnosis and treatment.This paper mainly uses diffusion tensor magnetic resonance data as a carrier and uses multivariate mode analysis as a means to carry out a classification study of smoking addiction and smoking relapse at the individual level,discovering objective biomarkers that can accurately characterize it,and further exploring smoking Brain mechanisms of addiction and relapse.Imaging studies on smoking addiction have often been limited to inter-group studies and cannot be used for individual identification and diagnosis.As an effective tool,the multimodal analysis method is widely used in imaging studies.We apply this method in combination with a variety of white matter information extracted from diffuse tensor magnetic resonance imaging to classify adolescents with smoking addiction and healthy control groups to mine white matter neural markers that are closely related to smoking addiction.The study finally obtained a classification accuracy of 88%.The most discriminative features obtained were the sagittal stratum,the external capsule,the superior longitudinal fasciculus,the anterior radiation radiata,and the superior frontal occipital fasciculus.Our findings not only indicate that the above white matter characteristics can be used to predict smoking status,but also the potential ability of machine learning techniques to mine neurobiological information related to smoking.Avoiding "relapse" is the key to quitting smoking.Quitting smoking can reduce the risk of lung cancer morbidity and death,and standardizing smoking cessation treatment can improve the success rate of smoking cessation.However,due to the lack of biomarkers that accurately characterize smoking and relapse,existing smoking cessation treatments have not achieved significant results.As an extension of the previous work,this study also used a multivariate pattern analysis method to explore the differences in white matter neural mechanisms between smoking relapsed individuals and withdrawal individuals.We use different algorithms to build the model and use leave-one-out cross-validation to evaluate the model’s generalization ability.The results show that the highest accuracy rate obtained is 85%.The white matter characteristics that contribute to the classification prediction are mainly fornix,cerebrallar peduncle,corpus callosum,left anterior radiation radiata,right posterior thalamus radiation and right posterior limb of internal capsule.The results of this study identify specific white matter microstructure biomarkers of smoking recurrence risk,which help define the neurobiological recurrence risk characteristics of smoking addiction and provide a basis for better smoking cessation treatment.