The Value Of Mr Characteristics In Comparative Analysis Of Pyogenic Spondylitis And Tuberculous Spondylitis In Adult: Based On Artificial Neural Network

Objective:To investigate the value of MRI in differential diagnosis of pyogenic spondylitis and tuberculous spondylitis using the artificial neural network.Methods:MRI features of 87 cases of spinal infection(47 cases of pyogenic spondylitis and 40 cases of tuberculous spondylitis)confirmed by pathology or clinical diagnosis were analyzed retrospectively.The general clinical data and MRI features of pyogenic spondylitis and tuberculous spondylitis were compared by t test,rank sum test,chi-square test and Logistics regression analysis method.The diagnostic boundary value of MRI parameter scores of the two groups were analyzed by the receiver operating characteristic curve(ROC).Artificial neural network method was used to further analyze the weight of each parameter and obtain the differential diagnosis scales of both two.Results:Among the 87 cases,the average age of the patients was 55.53 years,and 53 were male.228 spinal segments(including 20 accessory segments)involved,thoracic and lumbar segments accounted for 91%.The most important parameters of pyogenic spondylitis were slight bony destruction and severe disc destruction,easily forming paravertebral granuloma with unclear boundary,limited subligmentous spread scope,and involved vertebral body diffusely and homogeneously enhanced.While in tuberculous spondylitis,the bone destruction were heavy and disc retained more,vertebral bodies collapsed often and vertebral intraosseous abscess and paraspinal spreading abscess with clear boundary easily forming.The ROC analysis showed that the scores of MRI parameters is for the diagnosis of spinal infection.The MRI score of PS and TS was analyzed by ROC,and the area under the curve was 0.986.When the cutoff value was 0.5,the diagnostic sensitivity is 0.929 and the specificity is 1.According to the results of artificial neural network model,combined with different MR signs weight,according to the score scale to score the MR signs of the case,using the tuberculous spondylitis score standard,if the total score is more than 48 points,tuberculosis spondylitis can be diagnosed;if the total score is less than 48,try to use the pyogenic spondylitis score standard,if the total score is more than 66,the diagnosis of pyogenic spondylitis,the total score is less than 66,then it can be neither pyogenic spondylitis nor tuberculous spondylitis,may be brucellosis,fungal spondylitis or other types of infection or even non-infectious lesions.Conclusion:Distinctive MRI features,especially after contrast-enhancement,may facilitate the differential diagnosis of pyogenic spondylitis and tuberculous spondylitis.Artificial neural network is helpful for the differential diagnosis of spinal infection and the formulation of diagostic scoring scale.