Machine Learning Models of Voxel-Level [18F] Fluorodeoxyglucose Positron Emission Tomography Data Excel at Predicting Progressive Supranuclear Palsy Pathology

IF 7.7 1区医学 Q1 CLINICAL NEUROLOGY

Annals of Neurology Pub Date : 2025-05-30 DOI:10.1002/ana.27265

Addison S. Braun BS, Ryota Satoh PhD, Nha Trang Thu Pham BS, Neha Singh-Reilly PhD, Farwa Ali MD, Dennis W. Dickson MD, Val J. Lowe MD, Jennifer L. Whitwell PhD, Keith A. Josephs MD, MST, MSc

{"title":"Machine Learning Models of Voxel-Level [18F] Fluorodeoxyglucose Positron Emission Tomography Data Excel at Predicting Progressive Supranuclear Palsy Pathology","authors":"Addison S. Braun BS, Ryota Satoh PhD, Nha Trang Thu Pham BS, Neha Singh-Reilly PhD, Farwa Ali MD, Dennis W. Dickson MD, Val J. Lowe MD, Jennifer L. Whitwell PhD, Keith A. Josephs MD, MST, MSc","doi":"10.1002/ana.27265","DOIUrl":null,"url":null,"abstract":"<div>\n \n <section>\n \n <h3> Objective</h3>\n \n <p>To determine whether a machine learning model of voxel level [<sup>18</sup>f]fluorodeoxyglucose positron emission tomography (PET) data could predict progressive supranuclear palsy (PSP) pathology, as well as outperform currently available biomarkers.</p>\n </section>\n \n <section>\n \n <h3> Methods</h3>\n \n <p>One hundred and thirty-seven autopsied patients with PSP (n = 42) and other neurodegenerative diseases (n = 95) who underwent antemortem [<sup>18</sup>f]fluorodeoxyglucose PET and 3.0 Tesla magnetic resonance imaging (MRI) scans were analyzed. A linear support vector machine was applied to differentiate pathological groups with sensitivity analyses performed to assess the influence of voxel size and region removal. A radial basis function was also prepared to create a secondary model using the most important voxels. The models were optimized on the main dataset (n = 104), and their performance was compared with the magnetic resonance parkinsonism index measured on MRI in the independent test dataset (n = 33).</p>\n </section>\n \n <section>\n \n <h3> Results</h3>\n \n <p>The model had the highest accuracy (0.91) and F-score (0.86) when voxel size was 6mm. In this optimized model, important voxels for differentiating the groups were observed in the thalamus, midbrain, and cerebellar dentate. The secondary models found the combination of thalamus and dentate to have the highest accuracy (0.89) and F-score (0.81). The optimized secondary model showed the highest accuracy (0.91) and F-scores (0.86) in the test dataset and outperformed the magnetic resonance parkinsonism index (0.81 and 0.70, respectively).</p>\n </section>\n \n <section>\n \n <h3> Interpretation</h3>\n \n <p>The results suggest that glucose hypometabolism in the thalamus and cerebellar dentate have the highest potential for predicting PSP pathology. Our optimized machine learning model outperformed the best currently available biomarker to predict PSP pathology. ANN NEUROL 2025;98:410–420</p>\n </section>\n </div>","PeriodicalId":127,"journal":{"name":"Annals of Neurology","volume":"98 2","pages":"410-420"},"PeriodicalIF":7.7000,"publicationDate":"2025-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Annals of Neurology","FirstCategoryId":"3","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1002/ana.27265","RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"CLINICAL NEUROLOGY","Score":null,"Total":0}

引用次数: 0

Abstract

Objective

To determine whether a machine learning model of voxel level [¹⁸f]fluorodeoxyglucose positron emission tomography (PET) data could predict progressive supranuclear palsy (PSP) pathology, as well as outperform currently available biomarkers.

Methods

One hundred and thirty-seven autopsied patients with PSP (n = 42) and other neurodegenerative diseases (n = 95) who underwent antemortem [¹⁸f]fluorodeoxyglucose PET and 3.0 Tesla magnetic resonance imaging (MRI) scans were analyzed. A linear support vector machine was applied to differentiate pathological groups with sensitivity analyses performed to assess the influence of voxel size and region removal. A radial basis function was also prepared to create a secondary model using the most important voxels. The models were optimized on the main dataset (n = 104), and their performance was compared with the magnetic resonance parkinsonism index measured on MRI in the independent test dataset (n = 33).

Results

The model had the highest accuracy (0.91) and F-score (0.86) when voxel size was 6mm. In this optimized model, important voxels for differentiating the groups were observed in the thalamus, midbrain, and cerebellar dentate. The secondary models found the combination of thalamus and dentate to have the highest accuracy (0.89) and F-score (0.81). The optimized secondary model showed the highest accuracy (0.91) and F-scores (0.86) in the test dataset and outperformed the magnetic resonance parkinsonism index (0.81 and 0.70, respectively).

Interpretation

The results suggest that glucose hypometabolism in the thalamus and cerebellar dentate have the highest potential for predicting PSP pathology. Our optimized machine learning model outperformed the best currently available biomarker to predict PSP pathology. ANN NEUROL 2025;98:410–420

查看原文本刊更多论文

体素水平的机器学习模型[18F]氟脱氧葡萄糖正电子发射断层扫描数据在预测进行性核上性麻痹病理方面表现出色。

目的：确定体素水平[18f]氟脱氧葡萄糖正电子发射断层扫描（PET）数据的机器学习模型是否可以预测进行性核上性麻痹（PSP）病理，并优于目前可用的生物标志物。方法：对137例PSP及其他神经退行性疾病的尸检患者（n = 42）进行死前[18f]氟脱氧葡萄糖PET和3.0特斯拉磁共振成像（MRI）扫描分析。采用线性支持向量机区分病理组，并进行敏感性分析，以评估体素大小和区域去除的影响。还准备了径向基函数来使用最重要的体素创建二级模型。在主数据集（n = 104）上对模型进行优化，并将其性能与独立测试数据集（n = 33）的MRI测量的磁共振帕金森指数进行比较。结果：当体素大小为6mm时，该模型具有最高的准确率（0.91）和f值（0.86）。在优化后的模型中，在丘脑、中脑和齿状小脑中观察到区分各组的重要体素。二级模型发现丘脑和齿状体的组合具有最高的准确性（0.89）和f分数（0.81）。优化后的二次模型在测试数据集中的准确率（0.91）和f分数（0.86）最高，优于磁共振帕金森指数（0.81和0.70）。解释：结果表明，丘脑和齿状小脑的葡萄糖代谢低下具有预测PSP病理的最高潜力。我们优化的机器学习模型在预测PSP病理方面优于目前可用的最佳生物标志物。Ann neurol 2025。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Annals of Neurology 医学-临床神经学

CiteScore

18.00

自引率

1.80%

发文量

270

审稿时长

3-8 weeks

期刊介绍： Annals of Neurology publishes original articles with potential for high impact in understanding the pathogenesis, clinical and laboratory features, diagnosis, treatment, outcomes and science underlying diseases of the human nervous system. Articles should ideally be of broad interest to the academic neurological community rather than solely to subspecialists in a particular field. Studies involving experimental model system, including those in cell and organ cultures and animals, of direct translational relevance to the understanding of neurological disease are also encouraged.