Ji Wu ^1* Chengjian Qin ² Yixuan Zhou ^1* Xuanlei Wei ^2* Deling Qin ^2* Keyu Chen ¹ Yuankun Cai ¹ Lei Shen ¹ Jingyi Yang ^1* Dongyuan Xu ¹ Songshan Chai ^1* NANXIANG XIONG ^1*

• ¹ Wuhan University, Wuhan, Hubei Province, China
• ² Affiliated Hospital of Youjiang Medical University for Nationalities, Baise, Guangx, China

Background：Classical Trigeminal neuralgia (TN) is defined as spontaneous pain in the region of the trigeminal nerve, which seriously affects patients' quality of life. Percutaneous balloon compression of the ganglion is a reproducible surgical procedure that reduces the incidence of TN, but the postoperative ineffectiveness or recurrence exists in some patients.It is proposed to establish a machine learning-based clinical-imaging Nomogram to predict the prognosis of recurrence of trigeminal neuralgia treated with percutaneous balloon compression (PBC). Methods:Clinical data of 117 Classical TN patients treated with PBC at Zhongnan Hospital of Wuhan University from January 2017 to August 2023 were retrospectively included. All imaging histologic morphological features were extracted from the intraoperative X-ray balloon region using 3D slicer. Clinical one-way analysis, LASSO (Least absolute shrinkage and selection operator) and random forest were used to predict the relationship between postoperative outcomes of TN PBC, and a predictive model was constructed using ROC curves, the impact of imaging histology on TN prognosis was assessed using Decision Curve Analysis (DCA), and finally, the model was validated using ROC curves. Results:Sixteen morphology-related imaging histological features were finally extracted for analysis. After a one-way logistic regression analysis, the original_shape_Elongation, original_shape_Maximum2DDiameterRow,original_shape_MinorAxisLength,original_s hape_SurfaceArea,original_shape_MeshVolume,original_shape_VoxelVolume, original_shape_SurfaceVolumeRatio,r,original_shape_MajorAxisLength,original_shape _Maximum3DDiameter and original_shape_Maximum2DDiameterSlice with morphology-related imaging histologic features and Affected side,Therapeutic effect of drugs,Trigeminal division and other clinical features, a total of 14 predictors. Screening was performed using random forest tree (RF) and performing lasso regression analysis, and finally the imaging features such as original_shape_Maximum2D DiameterRow, original_shape_Elongation were screened to predict the prognostic prediction of TN for PBC treatment with the areas under the roc curves being. 0.812, 0.874; and the predictive Nomogram was constructed, and the area under the ROC curve of the model was 0.872, suggesting that the model has good predictive ability.The DCA and calibration curves showed that the Nomogram has high applicability in clinical practice. Conclusion: Machine learning combined with clinical imaging and histologic screening of clinical information has a good predictive potential for the efficacy of PBC in the treatment of TN, which is suitable for clinical application in TN patients after PBC.