Development of a neural network model to establish a surgical paradigm for trigeminal neuralgia patients difficult to classify

Development of a neural network model to establish a surgical paradigm for trigeminal neuralgia patients difficult to classify | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Development of a neural network model to establish a surgical paradigm for trigeminal neuralgia patients difficult to classify Ying Wang, Hao Chen, Xiaofeng Jiang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3899748/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Backgrounds: Trigeminal neuralgia (TN) is a serious, intense and recurring pain in the sensory distribution of the trigeminal nerve in the face that is associated with decreased quality of life and increased risk of emotional disorders and physical health problems. Theoretically, TN can be divided into the classic type if vascular compression is found upon the trigeminus or the idiopathic type if vascular compression is not found upon any part of the trigeminus. Microvascular decompression (MVD) and internal neurolysis (IN) surgery are usually performed for classic or idiopathic TN, respectively, with satisfactory treatment effects. However, in clinical practice, there are patients with slight vascular contact with the trigeminus, and this is a dilemma when planning surgery because pain might be caused by this contact, which is usually insufficient to cause demyelination of the trigeminal nerve. Therefore, MVD is probably not effective and requires a second surgery, while IN is generally chosen blindly because of the high success rate along with some side effects and injury. Achieving a model with a clearer classifying boundary, especially for these patients, offers better opportunities for improved treatment efficacy. Methods: Toward this goal, in the present study, an SVM model was constructed with resting-state fMRI data from 70 definite CTN and ITN patients. Specifically, these 70 data points were randomly assigned to the training dataset and test dataset. The linear kernel function and 2-fold cross-validation modes of SVM and feature selection were used, and the process was repeated 10 times. Features maintained in all 10 random allocations were defined as final features of the SVM model. Results: We found that four ROI-pair connectivities were robustly effective in classification. With this model, another 16 patients with slight vascular contact who had received IN without model guidance were reclassified; 13 of these patients were classified as CTN and were likely to be appropriate for MVD. Conclusions: Taken together, the results of the present study could guide future clinical work, and TN patients who are difficult to classify could be labeled and returned to the model for improved classification accuracy in clinical use. trigeminal neuralgia vascular illusion classification neural network model Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Trigeminal neuralgia (TN) is known as the most serious pain in the world. Intense and recurring pain in the sensory distribution of the trigeminal nerve in the face and surgery are the main treatments for severe patients[ 1 ]. The prevalence of the disease varies in different regions, with an average of 12 to 29 people per 100,000 people globally and as high as 56 people per 100,000 people according to a survey of six major cities in China. TN greatly affects patients’ quality of life and often leads to emotional disorders and physical health problems [ 2 – 4 ]. Theoretically, TN can be divided into classic types if vascular is found on the trigeminus, and immediate analgesia can be expected after vascular decompression surgery[ 5 ]. Bendtsen reported on 5149 patients with classic TN who underwent microvascular decompression (MVD) surgery and were followed for 3 to 11 years; the authors found that the pain cure rate reached 62 to 89% [ 6 , 7 ]. Alternatively, if there is no vascular compression upon any part of the trigeminus, TN is usually classified as [ 8 – 10 ], and internal neurolysis (IN) surgery is usually performed with satisfactory treatment efficacy. Internal neurolysis (IN) involves longitudinal incision of the intracranial nerve root to achieve nerve micro destruction and is highly effective at curing ITN and significantly reducing the incidence of facial sensory disorders and the degree of numbness [ 11 , 12 ]; therefore, IN has been gradually accepted by an increasing number of scholars [ 13 – 15 ]. Our research team also used IN to treat ITN and came to similar conclusions [ 16 ]. The different characteristics and treatment choices of these diseases are due to their different pathophysiological bases and pathogenesis. At present, the pathophysiological basis of CTN is generally believed to be related mainly to the pressure of blood vessels on the intracranial segment of the trigeminal nerve, which leads to secondary demyelination of the nerve, resulting in abnormal and oversensitive nerve transmission through the production of “ignition”, “short circuit”, “sympathetic nerve excitation” and other effects, thus causing pain attacks [ 1 , 5 ]. In contrast, the pathophysiological basis and pathogenesis of the ITN are not fully understood and may involve changes in neurons and synapses as well as the involvement of inflammatory and immune responses, among other factors. However, the truth is not simple in clinical practice (i.e., severe vascular compression or no vascular space). Clinicians are usually puzzled when they find slight vascular contact because this contact seems unlikely to cause secondary demyelination of the nerve, and a second IN surgery would be inevitable if MVD surgery failed. This condition looks like CTN, but in fact, the blood vessels are most likely just an illusion (“vascular illusion”) and are actually the ITN. It looks like a dilemma. If IN surgery is performed directly, this may mean unnecessary nerve damage, and any nerve damage has its own side effects, such as numbness and chewing dysfunction [ 16 , 17 ]. If vascular decompression is used, this may mean that the operation fails and a second operation is needed, which is also an injury. It is difficult to accurately distinguish between the ITN and CTN because of this slight vascular contact with currently available preoperative neuroimaging methods. Therefore, surgical procedures are often performed without any reference but rather according to the doctor’s experience, which may lead to ineffective or unnecessary iatrogenic damage after surgery. In fact, single vascular decompression surgery is usually ineffective, and to avoid the need for a second operation due to ineffective decompression and to improve the success rate of the first operation, IN surgery alone [ 13 ] or combined with vascular decompression [ 15 ] is often used in clinical treatment. This is clearly due to a lack of helplessness, and more precise methods need to be further developed. Therefore, it is important to establish a novel preoperative predictive evaluation method to accurately distinguish TN patients who have slight vascular contact with the CTN from those with slight vascular contact with the ITN with “vascular illusion”. This approach would be quite helpful in improving the cure rate while ensuring the reduction of iatrogenic injury. To address this, we need clearly labeled training data, including a group of patients who have been clearly defined as CTN and ITN, as well as poorly classified data (i.e., those with “vascular illusion”). However, as mentioned above, to improve the first success rate and to avoid a second surgery, in previous clinical work, these patients were almost exclusively treated with IN surgery because IN is a kind of radiculotomy. This approach undoubtedly hides a subset of CTN conditions, leaving these patients unlabeled. To solve this problem, a two-step scheme might be used, and the present study is the first step of the whole process. Here, preoperative MR images that were clearly labeled CTN and ITN were collected to construct a classification model. This approach allows us to reclassify poorly classified patients who have undergone IN, although training data for this poorly classified patient population are lacking in this model. Furthermore, the model provides us with a reference for choosing a relatively more appropriate surgical paradigm for poorly classified surgical candidates. In the future, second, poorly classified patients were labeled “classic” if MVD surgery was successful or “idiopathic” if MVD surgery failed according to the model instructions. Afterwards, as these poorly classified patients were labeled and added to the training dataset of the model, we could constantly improve the model’s classification accuracy. Methods Participants Fifty-one classic (28 female, 55.4 ± 9.8 years old) and 19 idiopathic (14 female, 59 ± 8.1 years old) TN patients were recruited from the Department of Neurosurgery of the First Affiliated Hospital of the University of Science and Technology of China (USTC) to construct a classification model. We subsequently tested the distribution of another 16 unclassified patients (12 female, 58.4 ± 4.9 years old) with this model. All patients met the following criteria: had typical clinical manifestations of TN; had a normal MRI or CT evidence of no secondary factors, such as intracranial space-occupying lesions; had a poor response to medication or intolerable side effects; and had no serious complications or substantial damage to the heart, lung, kidney or other important organs. Patients were excluded from the study for any of the following reasons: secondary TN, other types of head and face pain, serious physical or mental illness, drug and alcohol dependence, or received microvascular decompression treatment. The demographic and clinical characteristics of the patient groups are presented in Table 1. Informed consent was obtained from all participants, and the study was approved by the Research Ethics Committee of the First Affiliated Hospital of the USTC and conformed to the tenets of the Declaration of Helsinki. MRI data acquisition and preprocessing Gradient echo-planar magnetic resonance imaging (MRI) data were obtained during the whole procedure on a 3.0 T GE MR750w (United States) at the First Affiliated Hospital of the University of Science and Technology of China. Before entering the MRI scanner, all participants were told to keep their heads steady during all scans. A circularly polarized head coil was used, with foam padding to restrict head motion. Resting-state MRI data with 242 frames were acquired with a T2*-weighted echo-planar imaging sequence (TE = 30 ms, TR = 2000 ms, FOV = 240 mm, matrix = 64 × 64, flip angle = 85°), with 33 axial slices (no gaps, voxel size = 3.6 × 3.6 × 3.6 mm3) covering the whole brain. Corresponding high-resolution T1-weighted three-dimensional gradient-echo (for stereotaxic transformation) images were also collected (TR = 1900 ms; TE = 2.26 ms; TI = 900 ms; 1 mm isotropic voxel; 250 mm field of view (FOV)). The results included in this manuscript were generated from preprocessing performed using fMRIPrep 23.0.0 (full details are provided in the Supplementary Materials). We first used preprocessed fMRI data to quantify head movements and other confounding factors and then used these estimates in the Xcp-d to denoise the fMRI signal and estimate functional connectivity. After preprocessing, we extracted the time series of resting-state fMRI data from 100 brain regions of interest (ROI) of the Schaefer Atlas. Next, we calculated the Pearson correlation coefficient between the time series of each pair of brain regions. With all the correlation coefficients of the resting-state fMRI data, we performed 10 random cross-validations using a support vector machine (SVM) model to effectively classify the patients (Fig. 1 ). Construction of an SVM model for classifying the CTN and ITN Given that random allocation of the 70 data points might introduce sampling bias, 10 random allocations of the 70 data points were conducted, and the intersection of the selected features in all 10 models was determined as the final feature of the SVM model (Fig. 2 ). For each random allocation, the 70 patients were assigned to the training dataset (80%, 56 patients) or test dataset (20%, 14 patients). For the training dataset (56 subjects × 4950 ROI pairs), we adopted the linear kernel function and 2-fold cross-validation mode of SVM. The group label (classical or idiopathic) was the dependent variable (that is, Y value), and the 4950 coefficient values of the ROI pairs were potential features of the SVM (that is, X 1 , …, X 4950 ). After the first SVM modeling with all 4950 features, the features were sorted in ascending order according to the absolute values of the coefficients. The feature with the minimum coefficient was removed, and the remaining features were subjected to the next SVM modeling. Similarly, we removed features one by one. The minimum number of ROI pairs with 100% prediction accuracy within the training dataset was obtained. As shown in Fig. 3 , the horizontal coordinate is 4950 pairwise ROI pairs, and the vertical coordinate is the prediction accuracy. From left to right, we discarded the ROI pairs with the lowest weights in the model one by one to evaluate their impact on the model’s classification accuracy. With the discarding of redundant ROI pairs, the classification accuracy continues to increase, and the core ROI pairs gradually appear. However, as the core ROI pairs were further discarded, the classification accuracy rapidly decreased. Similarly, after 10 random allocations, ten groups of ROI pairs were obtained, and the intersection of the ten groups of ROI pairs was taken for the SVM model for classifying typical and idiopathic patients. With the refined ROI pairs, the 70 data points were again randomly allocated into the training dataset and test dataset. An SVM model was constructed with the training dataset and the refined ROI pairs, and its classification accuracy was evaluated for the test dataset; hence, an averaged classification accuracy was assessed. The 16 poorly classified patients were classified with the model. Results Demographic and clinical comparisons Age, sex, course of disease, medication, etc., may all be modifiers of the neural network. Therefore, we first tested whether there were differences in these variables between the two groups of patients. The three groups of TN patients did not differ in age, sex, duration, or type of drug intake (Table 1). ROI pairs in all ten random SVM models After ten random allocations and feature selection procedures were used to obtain the regions most representative of group differences between the CTN and the ITN when SVM was used with the training dataset, we found that the coefficients of four pairs of ROIs were associated with all ten models, including brain regions distributed in the default mode system, dorsal attention system, limbic system, somatic movement, etc., in both hemispheres. The ROI pairs are shown in Table 2 and are visually displayed in Fig. 4 . Classification accuracy of these ROIs for an independent test dataset After obtaining the four refined ROI pairs, we constructed an SVM model and randomly allocated the data into a test dataset ten times. The average classification accuracy was 90%. Afterwards, with the model, the 16 poorly classified patients were classified into two groups, namely, 13 classical TN patients and 3 idiopathic TN patients. Discussion Surgical planning for patients with slight vascular contact to the trigeminus in clinical practice is difficult because of the available classification criteria and related theory. A model with a clearer classification boundary, especially for patients with slight vascular contact to the trigeminus, is needed to improve treatment efficacy. Toward this goal, the present study constructed a model with resting-state fMRI data from definite CTN and ITN patients and found that four ROI-pair connectivities were robustly effective in classification. With this model, 16 patients who had slight vascular contact and who received IN surgery without model guidance were reclassified. Thirteen of these patients were classified as CTN and were likely to be appropriate for MVD. From this point of view, these patients may receive good treatment via IN surgery, but there may be overtreatment. To prevent this kind of situation from occurring, we conducted this study. In previous clinical practice, there was no good way to judge patients with slight vascular contact. Moreover, our clinicians often adopt IN surgery in these patients. This may mean unnecessary nerve damage because any nerve damage has its own side effects, such as numbness and hypesthesia [ 17 ]. However, compared to the possibility of a second operation if MVD fails, this injury caused by IN surgery is a more difficult option because of the lower volume of the two evils. However, the balance between two possible harms is not minimal harm or the best choice, as the most suitable treatment is the best. Benefiting from the emerging computational tools and conceptual frameworks of network neuroscience, the present study took the first step toward identifying the most suitable treatment rather than being a blind choice. It is desirable to clearly distinguish the different types of trigeminal neuralgia and to construct an accurate classification model. However, in the real world, some causes of trigeminal neuralgia may not be unique; for example, vascular compression, which is one of the causes of trigeminal neuralgia but is not the only cause, may be combined with idiopathic trigeminal neuralgia, such as inflammation. The main purpose of the model is to classify difficult-to-classify TN patients; however, the training dataset of the model in this study did not cover the target. This is the main limitation of this study. Nevertheless, this is an inevitable question in clinical practice. Without any guidance, a one-size-fits-all solution would be the only choice to ensure a high success rate of pain relief, but this would also obscure the labels of these patients. In addition, these poorly classified TN patients can be divided into CTN or ITN patients. Therefore, a model whose training dataset included a definite CTN and ITN could help us find similarities between these poorly classified TN patients and CTN/ITN patients, which might constitute the path to achieving the final goal. In future work, we should adopt a surgical paradigm under the guidance of the model. Patients should be accurately labeled after surgery, and their data should be returned to the model to improve the classification accuracy of the model and to determine its role in clinical practice. Conclusion In summary, for the first time, we present the existence of an unclear classification system for patients with trigeminal neuralgia based on clinical practice and whether the intraoperatively responsible vessel is the cause of pain. Therefore, a classification model was constructed based on clearly labeled data to provide a basis for subsequent preoperative classification of patients, which may help improve treatment efficacy and reduce excessive medical treatment. Abbreviations TN trigeminal neuralgia MVD microvascular decompression IN internal neurolysis MRI magnetic resonance imaging USTC University of Science and Technology of China FOV field of view SVM support vector machine ROI brain regions of interest Declarations Ethics approval and consent to participate Informed consent was obtained from all participants, and the study was approved by the Research Ethics Committee of the First Affiliated Hospital of the USTC and conformed to the tenets of the Declaration of Helsinki. Consent for publication Not applicable Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request Competing interests The authors declare that they have no competing interests. Funding This work was supported by Talent Introduction Plan Project (RC2021004). Authors' contributions Y.W., and X.F.J. designed the study, analyzed the data, interpreted the data, and contributed to article writing. H.C. contributed to data collection. Acknowledgements We thank all the study participants for their support to our research. References Love S, Coakham HB (2001) Trigeminal neuralgia: pathology and pathogenesis. Brain 124(Pt 12):2347–2360 Baghaei S et al (2023) Evaluation of anxiety disorder in patients with trigeminal neuralgia. Surg Neurol Int 14:266 Chang B, Zhu W, Li S (2019) Effects of Depression and Anxiety on Microvascular Decompression Outcome for Trigeminal Neuralgia Patients. World Neurosurg 128:e556–e561 Puskar KR, Droppa M (2015) Trigeminal neuralgia: pain, pricks, and anxiety. J Gerontol Nurs 41(3):8–12 Maarbjerg S et al (2015) Significance of neurovascular contact in classical trigeminal neuralgia. Brain 138(Pt 2):311–319 Bendtsen L et al (2019) European Academy of Neurology guideline on trigeminal neuralgia. Eur J Neurol 26(6):831–849 Bendtsen L et al (2019) European Academy of Neurology guideline on trigeminal neuralgia. Eur J Neurol 26(6):831–849 Badel T et al (2022) Evaulation of Blink Reflex between Patients with Idiopathic Trigeminal Neuralgia and Healthy Volunteers. Acta Clin Croat 61(Suppl 2):121–128 Jay GW, Barkin RL (2022) Trigeminal neuralgia and persistent idiopathic facial pain (atypical facial pain). Dis Mon 68(6):101302 Katusic S et al (1991) Epidemiology and clinical features of idiopathic trigeminal neuralgia and glossopharyngeal neuralgia: similarities and differences, Rochester, Minnesota, 1945–1984. Neuroepidemiology 10(5–6):276–281 Xu R et al (2022) Internal neurolysis versus intraoperative glycerin rhizotomy for trigeminal neuralgia . J Neurosurg, : p. 1–6 Xu R et al (2023) Internal neurolysis versus intraoperative glycerin rhizotomy for trigeminal neuralgia. J Neurosurg 138(1):270–275 Li MW, Jiang XF, Niu C (2021) Efficacy of Internal Neurolysis for Trigeminal Neuralgia without Vascular Compression. J Neurol Surg A Cent Eur Neurosurg 82(4):364–368 Zakrzewska JM (2015) Letter to the Editor: Internal neurolysis for trigeminal neuralgia. J Neurosurg 123(6):1612–1613 Zheng W et al (2021) Long Time Efficacy and Safety of Microvascular Decompression Combined with Internal Neurolysis for Recurrent Trigeminal Neuralgia. J Korean Neurosurg Soc 64(6):966–974 Wu M et al (2018) Outcome of Internal Neurolysis for Trigeminal Neuralgia without Neurovascular Compression and Its Relationship with Intraoperative Trigeminocardiac Reflex. Stereotact Funct Neurosurg 96(5):305–310 Ko AL et al (2015) Long-term efficacy and safety of internal neurolysis for trigeminal neuralgia without neurovascular compression. J Neurosurg 122(5):1048–1057 Tables Table 1 Sample demographic and clinical characteristics CTN ITN Difficult-to-classify F/χ 2 p Age 55.4±9.8 59±8.1 58.4±4.9 1.52 0.2 Gender 28/51 14/19 12/16 3.3 0.2 Duration of TN 5.2±5.4 7.6±7.4 6.5±4.4 0.91 0.4 Drugs 1.4±1.0 1.2±0.5 1.6±1.2 1.35 0.3 Table 2 Four pairs of ROIs whose coefficients were significant in all ten models. ROI 1 ROI 2 '7Networks_LH_SomMot_4' '7Networks_LH_DorsAttn_FEF_1' '7Networks_LH_Limbic_TempPole_1' '7Networks_RH_Default_pCunPCC_2' '7Networks_LH_Cont_pCun_1' '7Networks_RH_DorsAttn_Post_3' '7Networks_LH_Default_PFC_7' '7Networks_RH_DorsAttn_PrCv_1' Additional Declarations No competing interests reported. Supplementary Files SOM.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3899748","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":269892057,"identity":"b2eb3e56-8a0d-4891-89e7-a0ec58f02d69","order_by":0,"name":"Ying Wang","email":"","orcid":"","institution":"The First Affiliated Hospital of USTC, University of Science and Technology of China","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Ying","middleName":"","lastName":"Wang","suffix":""},{"id":269892058,"identity":"034b1372-10e7-4738-ac17-d2d91ad168e7","order_by":1,"name":"Hao Chen","email":"","orcid":"","institution":"The First Affiliated Hospital of USTC, University of Science and Technology of China","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Hao","middleName":"","lastName":"Chen","suffix":""},{"id":269892059,"identity":"147af6fd-c1bb-4be0-84fc-c0743b3bb1ce","order_by":2,"name":"Xiaofeng Jiang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA30lEQVRIie3RPQrCMBiA4UAhXRK7SYs/vUKLa8GrpLjWvYNKQcnmrngJx45fKdQlN3BRnIWKk6BgWnFtMwrmHRIC30MIQUin+8UMYCA3gg0DoIxVCGY16XdMHmYboUbqLRgSMcrpSkF4BTrDI50TbEcl0AS5VheaicMRy9biIMl0D06K/O2ONRPLTRhQXnyILxDzji0EY3nLqybRCUKuQCxJcspnBBOBIFMh1VvyAQeCTe5libDb3+IVZHK78sXYXRqX+zMOXKvXQuS/VxP592S3jVeZIJeFyqROp9P9a28pC0j6V8sBRwAAAABJRU5ErkJggg==","orcid":"","institution":"The First Affiliated Hospital of USTC, University of Science and Technology of China","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Xiaofeng","middleName":"","lastName":"Jiang","suffix":""}],"badges":[],"createdAt":"2024-01-26 10:59:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3899748/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3899748/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":50515769,"identity":"753036d5-dfce-47cd-b364-b8022b98fd96","added_by":"auto","created_at":"2024-02-01 16:49:17","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":79344,"visible":true,"origin":"","legend":"\u003cp\u003eThe flowchart of the study design.\u003c/p\u003e","description":"","filename":"Online1.png","url":"https://assets-eu.researchsquare.com/files/rs-3899748/v1/9f68322c06c93ba77d0175a0.png"},{"id":50515774,"identity":"bc5b592d-c4ff-4d45-baab-f44766025b04","added_by":"auto","created_at":"2024-02-01 16:49:18","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":807417,"visible":true,"origin":"","legend":"\u003cp\u003eSchematic overview of the fMRI data analysis.\u003c/p\u003e","description":"","filename":"Online2.png","url":"https://assets-eu.researchsquare.com/files/rs-3899748/v1/cce6733e005137856f4b2558.png"},{"id":50515773,"identity":"ded54888-b8ff-4bbb-bad5-41b736b56d8d","added_by":"auto","created_at":"2024-02-01 16:49:18","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":478019,"visible":true,"origin":"","legend":"\u003cp\u003eFeature selection of the SVM model. After each random allocation, for the training dataset (66 subjects × 4950 ROI pairs), we adopted the linear kernel function and 2-fold cross-validation mode of SVM. The group label (typical or idiopathic) was the dependent variable (that is, Y value), and the 4950 coefficient values of the ROI pairs were potential features of the SVM (that is, X_1, …, X_4950). LASSO regression was performed with 4950 eigenvalues (correlation coefficients between pairwise and pairwise brain regions) for each patient, which had two categorical dependent variables, typical or idiopathic. After the first SVM modeling with all 4950 features, the features were sorted in ascending order according to the absolute values of the coefficients. We discard the eigenvalues one by one according to the regression coefficients corresponding to these coefficients in the regression and observe the prediction accuracy of the variables corresponding to the remaining eigenvalues. The feature with the minimum coefficient was removed, and the remaining features were subjected to the next SVM modeling. Similarly, we removed features one by one. The horizontal coordinate is the number of dropped eigenvalues, and the vertical coordinate is the classification accuracy of the model. With the discarding of redundant eigenvalues, the classification accuracy of the model gradually improves, and the eigenvalues with 100% classification accuracy are screened out. With further discarding, the classification accuracy of the model begins to decrease again. Therefore, the remaining eigenvalues after the classification accuracy of the model reaches 100% are the important eigenvalues retained by the random grouping data.\u003c/p\u003e","description":"","filename":"Online3.png","url":"https://assets-eu.researchsquare.com/files/rs-3899748/v1/2770b5edf22436e4696a7000.png"},{"id":50515771,"identity":"3e6a9901-479f-49ea-ad85-fa9213254919","added_by":"auto","created_at":"2024-02-01 16:49:18","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":80105,"visible":true,"origin":"","legend":"\u003cp\u003eFour ROI-pair connectivities robustly effective in classification.\u003c/p\u003e","description":"","filename":"Online4.png","url":"https://assets-eu.researchsquare.com/files/rs-3899748/v1/6f328cc842ec65f6bde3c896.png"},{"id":50537658,"identity":"9341ec10-59bd-4bdb-bec1-d7830dde1ce2","added_by":"auto","created_at":"2024-02-02 05:38:23","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1034843,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3899748/v1/d6250033-0647-4f5c-9bab-744242573410.pdf"},{"id":50515770,"identity":"38cef05f-55dc-4a68-b3cf-7a60e9a0bb3f","added_by":"auto","created_at":"2024-02-01 16:49:18","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":22238,"visible":true,"origin":"","legend":"","description":"","filename":"SOM.docx","url":"https://assets-eu.researchsquare.com/files/rs-3899748/v1/10db13eb9c8121a60edfbd64.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Development of a neural network model to establish a surgical paradigm for trigeminal neuralgia patients difficult to classify","fulltext":[{"header":"Introduction","content":"\u003cp\u003eTrigeminal neuralgia (TN) is known as the most serious pain in the world. Intense and recurring pain in the sensory distribution of the trigeminal nerve in the face and surgery are the main treatments for severe patients[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. The prevalence of the disease varies in different regions, with an average of 12 to 29 people per 100,000 people globally and as high as 56 people per 100,000 people according to a survey of six major cities in China. TN greatly affects patients\u0026rsquo; quality of life and often leads to emotional disorders and physical health problems [\u003cspan additionalcitationids=\"CR3\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eTheoretically, TN can be divided into classic types if vascular is found on the trigeminus, and immediate analgesia can be expected after vascular decompression surgery[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Bendtsen reported on 5149 patients with classic TN who underwent microvascular decompression (MVD) surgery and were followed for 3 to 11 years; the authors found that the pain cure rate reached 62 to 89% [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Alternatively, if there is no vascular compression upon any part of the trigeminus, TN is usually classified as [\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e], and internal neurolysis (IN) surgery is usually performed with satisfactory treatment efficacy. Internal neurolysis (IN) involves longitudinal incision of the intracranial nerve root to achieve nerve micro destruction and is highly effective at curing ITN and significantly reducing the incidence of facial sensory disorders and the degree of numbness [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]; therefore, IN has been gradually accepted by an increasing number of scholars [\u003cspan additionalcitationids=\"CR14\" citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Our research team also used IN to treat ITN and came to similar conclusions [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. The different characteristics and treatment choices of these diseases are due to their different pathophysiological bases and pathogenesis. At present, the pathophysiological basis of CTN is generally believed to be related mainly to the pressure of blood vessels on the intracranial segment of the trigeminal nerve, which leads to secondary demyelination of the nerve, resulting in abnormal and oversensitive nerve transmission through the production of \u0026ldquo;ignition\u0026rdquo;, \u0026ldquo;short circuit\u0026rdquo;, \u0026ldquo;sympathetic nerve excitation\u0026rdquo; and other effects, thus causing pain attacks [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. In contrast, the pathophysiological basis and pathogenesis of the ITN are not fully understood and may involve changes in neurons and synapses as well as the involvement of inflammatory and immune responses, among other factors.\u003c/p\u003e \u003cp\u003eHowever, the truth is not simple in clinical practice (i.e., severe vascular compression or no vascular space). Clinicians are usually puzzled when they find slight vascular contact because this contact seems unlikely to cause secondary demyelination of the nerve, and a second IN surgery would be inevitable if MVD surgery failed. This condition looks like CTN, but in fact, the blood vessels are most likely just an illusion (\u0026ldquo;vascular illusion\u0026rdquo;) and are actually the ITN. It looks like a dilemma. If IN surgery is performed directly, this may mean unnecessary nerve damage, and any nerve damage has its own side effects, such as numbness and chewing dysfunction [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. If vascular decompression is used, this may mean that the operation fails and a second operation is needed, which is also an injury.\u003c/p\u003e \u003cp\u003eIt is difficult to accurately distinguish between the ITN and CTN because of this slight vascular contact with currently available preoperative neuroimaging methods. Therefore, surgical procedures are often performed without any reference but rather according to the doctor\u0026rsquo;s experience, which may lead to ineffective or unnecessary iatrogenic damage after surgery. In fact, single vascular decompression surgery is usually ineffective, and to avoid the need for a second operation due to ineffective decompression and to improve the success rate of the first operation, IN surgery alone [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e] or combined with vascular decompression [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] is often used in clinical treatment. This is clearly due to a lack of helplessness, and more precise methods need to be further developed. Therefore, it is important to establish a novel preoperative predictive evaluation method to accurately distinguish TN patients who have slight vascular contact with the CTN from those with slight vascular contact with the ITN with \u0026ldquo;vascular illusion\u0026rdquo;. This approach would be quite helpful in improving the cure rate while ensuring the reduction of iatrogenic injury.\u003c/p\u003e \u003cp\u003eTo address this, we need clearly labeled training data, including a group of patients who have been clearly defined as CTN and ITN, as well as poorly classified data (i.e., those with \u0026ldquo;vascular illusion\u0026rdquo;). However, as mentioned above, to improve the first success rate and to avoid a second surgery, in previous clinical work, these patients were almost exclusively treated with IN surgery because IN is a kind of radiculotomy. This approach undoubtedly hides a subset of CTN conditions, leaving these patients unlabeled.\u003c/p\u003e \u003cp\u003eTo solve this problem, a two-step scheme might be used, and the present study is the first step of the whole process. Here, preoperative MR images that were clearly labeled CTN and ITN were collected to construct a classification model. This approach allows us to reclassify poorly classified patients who have undergone IN, although training data for this poorly classified patient population are lacking in this model. Furthermore, the model provides us with a reference for choosing a relatively more appropriate surgical paradigm for poorly classified surgical candidates. In the future, second, poorly classified patients were labeled \u0026ldquo;classic\u0026rdquo; if MVD surgery was successful or \u0026ldquo;idiopathic\u0026rdquo; if MVD surgery failed according to the model instructions. Afterwards, as these poorly classified patients were labeled and added to the training dataset of the model, we could constantly improve the model\u0026rsquo;s classification accuracy.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eParticipants\u003c/h2\u003e \u003cp\u003eFifty-one classic (28 female, 55.4\u0026thinsp;\u0026plusmn;\u0026thinsp;9.8 years old) and 19 idiopathic (14 female, 59\u0026thinsp;\u0026plusmn;\u0026thinsp;8.1 years old) TN patients were recruited from the Department of Neurosurgery of the First Affiliated Hospital of the University of Science and Technology of China (USTC) to construct a classification model. We subsequently tested the distribution of another 16 unclassified patients (12 female, 58.4\u0026thinsp;\u0026plusmn;\u0026thinsp;4.9 years old) with this model. All patients met the following criteria: had typical clinical manifestations of TN; had a normal MRI or CT evidence of no secondary factors, such as intracranial space-occupying lesions; had a poor response to medication or intolerable side effects; and had no serious complications or substantial damage to the heart, lung, kidney or other important organs. Patients were excluded from the study for any of the following reasons: secondary TN, other types of head and face pain, serious physical or mental illness, drug and alcohol dependence, or received microvascular decompression treatment. The demographic and clinical characteristics of the patient groups are presented in Table\u0026nbsp;1. Informed consent was obtained from all participants, and the study was approved by the Research Ethics Committee of the First Affiliated Hospital of the USTC and conformed to the tenets of the Declaration of Helsinki.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eMRI data acquisition and preprocessing\u003c/h2\u003e \u003cp\u003e Gradient echo-planar magnetic resonance imaging (MRI) data were obtained during the whole procedure on a 3.0 T GE MR750w (United States) at the First Affiliated Hospital of the University of Science and Technology of China. Before entering the MRI scanner, all participants were told to keep their heads steady during all scans. A circularly polarized head coil was used, with foam padding to restrict head motion. Resting-state MRI data with 242 frames were acquired with a T2*-weighted echo-planar imaging sequence (TE\u0026thinsp;=\u0026thinsp;30 ms, TR\u0026thinsp;=\u0026thinsp;2000 ms, FOV\u0026thinsp;=\u0026thinsp;240 mm, matrix\u0026thinsp;=\u0026thinsp;64 \u0026times; 64, flip angle\u0026thinsp;=\u0026thinsp;85\u0026deg;), with 33 axial slices (no gaps, voxel size\u0026thinsp;=\u0026thinsp;3.6 \u0026times; 3.6 \u0026times; 3.6 mm3) covering the whole brain. Corresponding high-resolution T1-weighted three-dimensional gradient-echo (for stereotaxic transformation) images were also collected (TR\u0026thinsp;=\u0026thinsp;1900 ms; TE\u0026thinsp;=\u0026thinsp;2.26 ms; TI\u0026thinsp;=\u0026thinsp;900 ms; 1 mm isotropic voxel; 250 mm field of view (FOV)).\u003c/p\u003e \u003cp\u003eThe results included in this manuscript were generated from preprocessing performed using fMRIPrep 23.0.0 (full details are provided in the Supplementary Materials). We first used preprocessed fMRI data to quantify head movements and other confounding factors and then used these estimates in the Xcp-d to denoise the fMRI signal and estimate functional connectivity. After preprocessing, we extracted the time series of resting-state fMRI data from 100 brain regions of interest (ROI) of the Schaefer Atlas. Next, we calculated the Pearson correlation coefficient between the time series of each pair of brain regions. With all the correlation coefficients of the resting-state fMRI data, we performed 10 random cross-validations using a support vector machine (SVM) model to effectively classify the patients (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eConstruction of an SVM model for classifying the CTN and ITN\u003c/h2\u003e \u003cp\u003eGiven that random allocation of the 70 data points might introduce sampling bias, 10 random allocations of the 70 data points were conducted, and the intersection of the selected features in all 10 models was determined as the final feature of the SVM model (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFor each random allocation, the 70 patients were assigned to the training dataset (80%, 56 patients) or test dataset (20%, 14 patients). For the training dataset (56 subjects \u0026times; 4950 ROI pairs), we adopted the linear kernel function and 2-fold cross-validation mode of SVM. The group label (classical or idiopathic) was the dependent variable (that is, Y value), and the 4950 coefficient values of the ROI pairs were potential features of the SVM (that is, X\u003csub\u003e1\u003c/sub\u003e, \u0026hellip;, X\u003csub\u003e4950\u003c/sub\u003e). After the first SVM modeling with all 4950 features, the features were sorted in ascending order according to the absolute values of the coefficients. The feature with the minimum coefficient was removed, and the remaining features were subjected to the next SVM modeling. Similarly, we removed features one by one. The minimum number of ROI pairs with 100% prediction accuracy within the training dataset was obtained. As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, the horizontal coordinate is 4950 pairwise ROI pairs, and the vertical coordinate is the prediction accuracy. From left to right, we discarded the ROI pairs with the lowest weights in the model one by one to evaluate their impact on the model\u0026rsquo;s classification accuracy. With the discarding of redundant ROI pairs, the classification accuracy continues to increase, and the core ROI pairs gradually appear. However, as the core ROI pairs were further discarded, the classification accuracy rapidly decreased. Similarly, after 10 random allocations, ten groups of ROI pairs were obtained, and the intersection of the ten groups of ROI pairs was taken for the SVM model for classifying typical and idiopathic patients. With the refined ROI pairs, the 70 data points were again randomly allocated into the training dataset and test dataset. An SVM model was constructed with the training dataset and the refined ROI pairs, and its classification accuracy was evaluated for the test dataset; hence, an averaged classification accuracy was assessed. The 16 poorly classified patients were classified with the model.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eDemographic and clinical comparisons\u003c/h2\u003e \u003cp\u003eAge, sex, course of disease, medication, etc., may all be modifiers of the neural network. Therefore, we first tested whether there were differences in these variables between the two groups of patients. The three groups of TN patients did not differ in age, sex, duration, or type of drug intake (Table\u0026nbsp;1).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eROI pairs in all ten random SVM models\u003c/h2\u003e \u003cp\u003eAfter ten random allocations and feature selection procedures were used to obtain the regions most representative of group differences between the CTN and the ITN when SVM was used with the training dataset, we found that the coefficients of four pairs of ROIs were associated with all ten models, including brain regions distributed in the default mode system, dorsal attention system, limbic system, somatic movement, etc., in both hemispheres. The ROI pairs are shown in Table\u0026nbsp;2 and are visually displayed in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eClassification accuracy of these ROIs for an independent test dataset\u003c/h2\u003e \u003cp\u003eAfter obtaining the four refined ROI pairs, we constructed an SVM model and randomly allocated the data into a test dataset ten times. The average classification accuracy was 90%. Afterwards, with the model, the 16 poorly classified patients were classified into two groups, namely, 13 classical TN patients and 3 idiopathic TN patients.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eSurgical planning for patients with slight vascular contact to the trigeminus in clinical practice is difficult because of the available classification criteria and related theory. A model with a clearer classification boundary, especially for patients with slight vascular contact to the trigeminus, is needed to improve treatment efficacy. Toward this goal, the present study constructed a model with resting-state fMRI data from definite CTN and ITN patients and found that four ROI-pair connectivities were robustly effective in classification. With this model, 16 patients who had slight vascular contact and who received IN surgery without model guidance were reclassified. Thirteen of these patients were classified as CTN and were likely to be appropriate for MVD. From this point of view, these patients may receive good treatment via IN surgery, but there may be overtreatment. To prevent this kind of situation from occurring, we conducted this study.\u003c/p\u003e \u003cp\u003eIn previous clinical practice, there was no good way to judge patients with slight vascular contact. Moreover, our clinicians often adopt IN surgery in these patients. This may mean unnecessary nerve damage because any nerve damage has its own side effects, such as numbness and hypesthesia [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. However, compared to the possibility of a second operation if MVD fails, this injury caused by IN surgery is a more difficult option because of the lower volume of the two evils. However, the balance between two possible harms is not minimal harm or the best choice, as the most suitable treatment is the best. Benefiting from the emerging computational tools and conceptual frameworks of network neuroscience, the present study took the first step toward identifying the most suitable treatment rather than being a blind choice.\u003c/p\u003e \u003cp\u003eIt is desirable to clearly distinguish the different types of trigeminal neuralgia and to construct an accurate classification model. However, in the real world, some causes of trigeminal neuralgia may not be unique; for example, vascular compression, which is one of the causes of trigeminal neuralgia but is not the only cause, may be combined with idiopathic trigeminal neuralgia, such as inflammation.\u003c/p\u003e \u003cp\u003eThe main purpose of the model is to classify difficult-to-classify TN patients; however, the training dataset of the model in this study did not cover the target. This is the main limitation of this study. Nevertheless, this is an inevitable question in clinical practice. Without any guidance, a one-size-fits-all solution would be the only choice to ensure a high success rate of pain relief, but this would also obscure the labels of these patients. In addition, these poorly classified TN patients can be divided into CTN or ITN patients. Therefore, a model whose training dataset included a definite CTN and ITN could help us find similarities between these poorly classified TN patients and CTN/ITN patients, which might constitute the path to achieving the final goal. In future work, we should adopt a surgical paradigm under the guidance of the model. Patients should be accurately labeled after surgery, and their data should be returned to the model to improve the classification accuracy of the model and to determine its role in clinical practice.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn summary, for the first time, we present the existence of an unclear classification system for patients with trigeminal neuralgia based on clinical practice and whether the intraoperatively responsible vessel is the cause of pain. Therefore, a classification model was constructed based on clearly labeled data to provide a basis for subsequent preoperative classification of patients, which may help improve treatment efficacy and reduce excessive medical treatment.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTN\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003etrigeminal neuralgia\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMVD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003emicrovascular decompression\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIN\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003einternal neurolysis\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMRI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003emagnetic resonance imaging\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eUSTC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eUniversity of Science and Technology of China\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eFOV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003efield of view\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSVM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003esupport vector machine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eROI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ebrain regions of interest\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003ch3\u003eEthics approval and consent to participate\u003c/h3\u003e\n\u003cp\u003eInformed consent was obtained from all participants, and the study was approved by the Research Ethics Committee of the First Affiliated Hospital of the USTC and conformed to the tenets of the Declaration of Helsinki.\u003c/p\u003e\n\u003ch3\u003eConsent for publication\u003c/h3\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003ch3\u003eAvailability of data and materials\u003c/h3\u003e\n\u003cp\u003eThe datasets used and/or analysed during the current study are available from the corresponding author on reasonable request\u003c/p\u003e\n\u003ch3\u003eCompeting interests\u003c/h3\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003ch3\u003eFunding\u003c/h3\u003e\n\u003cp\u003eThis work was supported by Talent Introduction Plan Project (RC2021004).\u003c/p\u003e\n\u003ch3\u003eAuthors\u0026apos; contributions\u003c/h3\u003e\n\u003cp\u003eY.W., and X.F.J. designed the study, analyzed the data, interpreted the data, and contributed to article writing. H.C. contributed to data collection.\u003c/p\u003e\n\u003ch3\u003eAcknowledgements\u003c/h3\u003e\n\u003cp\u003eWe thank all the study participants for their support to our research.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eLove S, Coakham HB (2001) Trigeminal neuralgia: pathology and pathogenesis. Brain 124(Pt 12):2347\u0026ndash;2360\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaghaei S et al (2023) Evaluation of anxiety disorder in patients with trigeminal neuralgia. Surg Neurol Int 14:266\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChang B, Zhu W, Li S (2019) Effects of Depression and Anxiety on Microvascular Decompression Outcome for Trigeminal Neuralgia Patients. World Neurosurg 128:e556\u0026ndash;e561\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePuskar KR, Droppa M (2015) Trigeminal neuralgia: pain, pricks, and anxiety. J Gerontol Nurs 41(3):8\u0026ndash;12\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMaarbjerg S et al (2015) Significance of neurovascular contact in classical trigeminal neuralgia. Brain 138(Pt 2):311\u0026ndash;319\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBendtsen L et al (2019) European Academy of Neurology guideline on trigeminal neuralgia. Eur J Neurol 26(6):831\u0026ndash;849\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBendtsen L et al (2019) European Academy of Neurology guideline on trigeminal neuralgia. Eur J Neurol 26(6):831\u0026ndash;849\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBadel T et al (2022) Evaulation of Blink Reflex between Patients with Idiopathic Trigeminal Neuralgia and Healthy Volunteers. Acta Clin Croat 61(Suppl 2):121\u0026ndash;128\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJay GW, Barkin RL (2022) Trigeminal neuralgia and persistent idiopathic facial pain (atypical facial pain). Dis Mon 68(6):101302\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKatusic S et al (1991) Epidemiology and clinical features of idiopathic trigeminal neuralgia and glossopharyngeal neuralgia: similarities and differences, Rochester, Minnesota, 1945\u0026ndash;1984. Neuroepidemiology 10(5\u0026ndash;6):276\u0026ndash;281\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu R et al (2022) \u003cem\u003eInternal neurolysis versus intraoperative glycerin rhizotomy for trigeminal neuralgia\u003c/em\u003e. J Neurosurg, : p. 1\u0026ndash;6\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu R et al (2023) Internal neurolysis versus intraoperative glycerin rhizotomy for trigeminal neuralgia. J Neurosurg 138(1):270\u0026ndash;275\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi MW, Jiang XF, Niu C (2021) Efficacy of Internal Neurolysis for Trigeminal Neuralgia without Vascular Compression. J Neurol Surg A Cent Eur Neurosurg 82(4):364\u0026ndash;368\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZakrzewska JM (2015) Letter to the Editor: Internal neurolysis for trigeminal neuralgia. J Neurosurg 123(6):1612\u0026ndash;1613\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZheng W et al (2021) Long Time Efficacy and Safety of Microvascular Decompression Combined with Internal Neurolysis for Recurrent Trigeminal Neuralgia. J Korean Neurosurg Soc 64(6):966\u0026ndash;974\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu M et al (2018) Outcome of Internal Neurolysis for Trigeminal Neuralgia without Neurovascular Compression and Its Relationship with Intraoperative Trigeminocardiac Reflex. Stereotact Funct Neurosurg 96(5):305\u0026ndash;310\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKo AL et al (2015) Long-term efficacy and safety of internal neurolysis for trigeminal neuralgia without neurovascular compression. J Neurosurg 122(5):1048\u0026ndash;1057\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTable\u0026nbsp;1\u0026nbsp;Sample demographic and clinical characteristics\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"18.840579710144926%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.130434782608695%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eCTN\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"17.028985507246375%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eITN\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.840579710144926%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eDifficult-to-classify\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.579710144927537%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eF/\u0026chi;\u003c/em\u003e\u003cem\u003e\u003csup\u003e2\u003c/sup\u003e\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.579710144927537%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"18.840579710144926%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eAge\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.130434782608695%\" valign=\"top\"\u003e\n \u003cp\u003e55.4\u0026plusmn;9.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"17.028985507246375%\" valign=\"top\"\u003e\n \u003cp\u003e59\u0026plusmn;8.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.840579710144926%\" valign=\"top\"\u003e\n \u003cp\u003e58.4\u0026plusmn;4.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.579710144927537%\" valign=\"top\"\u003e\n \u003cp\u003e1.52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.579710144927537%\" valign=\"top\"\u003e\n \u003cp\u003e0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"18.840579710144926%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eGender\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.130434782608695%\" valign=\"top\"\u003e\n \u003cp\u003e28/51\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"17.028985507246375%\" valign=\"top\"\u003e\n \u003cp\u003e14/19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.840579710144926%\" valign=\"top\"\u003e\n \u003cp\u003e12/16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.579710144927537%\" valign=\"top\"\u003e\n \u003cp\u003e3.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.579710144927537%\" valign=\"top\"\u003e\n \u003cp\u003e0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"18.840579710144926%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eDuration of TN\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.130434782608695%\" valign=\"top\"\u003e\n \u003cp\u003e5.2\u0026plusmn;5.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"17.028985507246375%\" valign=\"top\"\u003e\n \u003cp\u003e7.6\u0026plusmn;7.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.840579710144926%\" valign=\"top\"\u003e\n \u003cp\u003e6.5\u0026plusmn;4.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.579710144927537%\" valign=\"top\"\u003e\n \u003cp\u003e0.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.579710144927537%\" valign=\"top\"\u003e\n \u003cp\u003e0.4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"18.840579710144926%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003eDrugs\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.130434782608695%\" valign=\"top\"\u003e\n \u003cp\u003e1.4\u0026plusmn;1.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"17.028985507246375%\" valign=\"top\"\u003e\n \u003cp\u003e1.2\u0026plusmn;0.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.840579710144926%\" valign=\"top\"\u003e\n \u003cp\u003e1.6\u0026plusmn;1.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.579710144927537%\" valign=\"top\"\u003e\n \u003cp\u003e1.35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.579710144927537%\" valign=\"top\"\u003e\n \u003cp\u003e0.3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;Four pairs of ROIs whose coefficients were significant in all ten models.\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"615\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"54.146341463414636%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eROI 1\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"45.853658536585364%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eROI 2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"54.146341463414636%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026apos;7Networks_LH_SomMot_4\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"45.853658536585364%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026apos;7Networks_LH_DorsAttn_FEF_1\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"54.146341463414636%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026apos;7Networks_LH_Limbic_TempPole_1\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"45.853658536585364%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026apos;7Networks_RH_Default_pCunPCC_2\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"54.146341463414636%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026apos;7Networks_LH_Cont_pCun_1\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"45.853658536585364%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026apos;7Networks_RH_DorsAttn_Post_3\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"54.146341463414636%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026apos;7Networks_LH_Default_PFC_7\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"45.853658536585364%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026apos;7Networks_RH_DorsAttn_PrCv_1\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"trigeminal neuralgia, vascular illusion, classification, neural network model","lastPublishedDoi":"10.21203/rs.3.rs-3899748/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3899748/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackgrounds:\u003c/strong\u003e Trigeminal neuralgia (TN) is a serious, intense and recurring pain in the sensory distribution of the trigeminal nerve in the face that is associated with decreased quality of life and increased risk of emotional disorders and physical health problems. Theoretically, TN can be divided into the classic type if vascular compression is found upon the trigeminus or the idiopathic type if vascular compression is not found upon any part of the trigeminus. Microvascular decompression (MVD) and internal neurolysis (IN) surgery are usually performed for classic or idiopathic TN, respectively, with satisfactory treatment effects. However, in clinical practice, there are patients with slight vascular contact with the trigeminus, and this is a dilemma when planning surgery because pain might be caused by this contact, which is usually insufficient to cause demyelination of the trigeminal nerve. Therefore, MVD is probably not effective and requires a second surgery, while IN is generally chosen blindly because of the high success rate along with some side effects and injury. Achieving a model with a clearer classifying boundary, especially for these patients, offers better opportunities for improved treatment efficacy.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods: \u003c/strong\u003eToward this goal, in the present study, an SVM model was constructed with resting-state fMRI data from 70 definite CTN and ITN patients. Specifically, these 70 data points were randomly assigned to the training dataset and test dataset. The linear kernel function and 2-fold cross-validation modes of SVM and feature selection were used, and the process was repeated 10 times. Features maintained in all 10 random allocations were defined as final features of the SVM model.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults:\u003c/strong\u003eWe found that four ROI-pair connectivities were robustly effective in classification. With this model, another 16 patients with slight vascular contact who had received IN without model guidance were reclassified; 13 of these patients were classified as CTN and were likely to be appropriate for MVD.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions:\u003c/strong\u003eTaken together, the results of the present study could guide future clinical work, and TN patients who are difficult to classify could be labeled and returned to the model for improved classification accuracy in clinical use.\u003c/p\u003e","manuscriptTitle":"Development of a neural network model to establish a surgical paradigm for trigeminal neuralgia patients difficult to classify","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-02-01 16:49:13","doi":"10.21203/rs.3.rs-3899748/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"05825d8f-037c-40f0-a160-90fc0d1d444f","owner":[],"postedDate":"February 1st, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-02-02T05:30:16+00:00","versionOfRecord":[],"versionCreatedAt":"2024-02-01 16:49:13","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-3899748","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3899748","identity":"rs-3899748","version":["v1"]},"buildId":"7rjqhiLT3MXkJMwkYKINL","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}