Study On Diagnostic Markers Of Male Infertility And BPNN Diagnosis Model

Objective: This study aimed to analyze the correlation between diagnostic markers for male infertility and total testicular volume,semen analysis parameters,to explore the application value of markers alone and combined detection for evaluating male spermatogenic ability and diagnosing male infertility,also to build BPNN-based infertility evaluation and diagnosis model and evaluate its classification efficiency.In order to provide a scientific basis for the more rational use of relevant markers and clinical decision-making.Methods: The study included male infertility patients(n = 199)and healthy fertility males(n = 99)as the research objects,who were continuously admitted to the six centers in China from September 2018 to March 2019.Chemiluminescence immunoassay(CLIA)was used to detect the level of seven diagnostic markers of the research subjects,including serum inhibin B(INHB),anti-Müllerian hormone(AMH),follicle stimulating hormone(FSH),luteinizing hormone(LH),Estrogen(E2),Testosterone(T)and Prolactin(PRL),meanwhile the total testicular volume(TV)and the average concentration of sperm was measured.Spearman r was calculated to analyze the correlation between markers,logistic regression was adopted to build joint diagnosis model.According to ROC curve to evaluate the value of the markers alone or combined detection of infertility and evaluate male spermatogenesis ability,evaluation indicators included AUC and its 95% CI,sensitivity,specificity and likelihood ratio.Finally,an infertility classification evaluation and diagnosis model was constructed based on BP neural network,with marker detection results as model input and diagnosis results as model output,the data was divided into training and test sets(the ratio was 7:3),according to the model prediction accuracy and AUC to evaluate model prediction value.Results:(1)This study shows that serum marker levels in OA patients are basically normal.The reproductive centers can effectively distinguish OA from other infertility diseases based on semen parameters and serum markers.The marker PRL has no significant difference between the normal group and the case group(P>0.05),so there is no significance for the diagnosis of male infertility.However,other markers are of reference value for male spermatogenic function evaluation,among them,E2,T,AMH and INHB levels are positively correlated with spermatogenesis,while LH and FSH levels are negatively correlated with spermatogenesis.Comprehensive testing of these markers can accurately assess spermatogenic function and diagnose the type of infertility.(2)When the marker is detected separately,the diagnostic value of INHB for male infertility is better than FSH,and FSH is better than AMH.INHB is significantly better than FSH not only to distinguish between oligozoospermia and azoospermia,but also to distinguish OA from NOA(P<0.05).INHB can be the primary marker for the analysis and diagnosis of the cause of male infertility.AMH performs poorly in the classification of infertility diagnosis,it is less effective than FSH in the assessment of oligozoospermia and diagnosis of azoospermia,and has no diagnostic value for oligozoospermia and azoospermia.Therefore,alone testing of AMH has very limited value in the diagnosis of male infertility.In general,Serological markers have a higher clinical value for the diagnosis of male infertility compared with the total testicular volume,which is a traditional spermatogenic function evaluation indicator.(3)Joint diagnosis model based on logistic multi-factor regression indicates that the combined detection of INHB and FSH in the diagnosis of oligozoospermia was significantly more effective than the single detection of INHB.The Specificity increases from 63% to 81.3%,and AUC increases from 0.730 to 0.819(P<0.05),the sensitivity is almost unchanged.However,in the diagnosis and identification of other infertility,the combined diagnosis model of any combined detection does not show obvious advantages compared with the detection of the marker alone.(4)Using BP neural network to construct infertility diagnosis and evaluation model,it is found that the differential diagnosis model of azoospermia shows extremely high prediction accuracy,the prediction accuracy rates on the training set and the test set are97.1% and 100.0%,respectively,AUC=0.967.The prediction accuracy of the spermatogenic function grading evaluation model on the training set and test set is81.5% and 79.3%,compared with the normal group and the azoospermia group,the model’s ability to recognize oligozoospermia is relatively low.The prediction accuracy of the separately constructed oligozoospermia grading assessment model on the training set and test set is 61.8% and 65.2%,respectively,the AUC is only 0.716.Hence the classification effect is not good.Conclusions: Individual and joint detection of male infertility diagnostic markers INHB and FSH has high application value.The traditional index TV as an auxiliary index can effectively improve the diagnosis and grading efficiency combined with INHB and FSH for male infertility.BPNN-based infertility diagnosis and grading evaluation model can accurately predict azoospermia categories and evaluate spermatogenesis function.Therefore,it is an effective auxiliary diagnostic tool for male infertility.