Robarts Research Institute, Western University, London, Canada; Department of Medical Biophysics, Western University, London, Canada.
Division of Respirology, Department of Medicine, Western University, London, Canada.
Acad Radiol. 2024 Sep;31(9):3825-3836. doi: 10.1016/j.acra.2024.03.014. Epub 2024 Apr 18.
It remains difficult to predict longitudinal outcomes in long-COVID, even with chest CT and functional MRI. Xe MRI reflects airway dysfunction, measured using ventilation defect percent (VDP) and in long-COVID patients, MRI VDP was abnormal, suggestive of airways disease. While MRI VDP and quality-of-life improved 15-month post-COVID infection, both remained abnormal. To better understand the relationship of airways disease and quality-of-life improvements in patients with long-COVID, we extracted Xe ventilation MRI textures and generated machine-learning models in an effort to predict improved quality-of-life, 15-month post-infection.
Long-COVID patients provided written-informed consent to 3-month and 15-month post-infection visits. Pyradiomics was used to extract Xe ventilation MRI texture features, which were ranked using a Random-Forest classifier. Top-ranking features were used in classification models to dichotomize patients based on St. George's Respiratory Questionnaire (SGRQ) score improvement greater than the minimal-clinically-important-difference (MCID). Classification performance was evaluated using the area under the receiver-operator-characteristic-curve (AUC), sensitivity, and specificity.
120 texture features were extracted from Xe ventilation MRI in 44 long-COVID participants (54 ± 14 years), including 30 (52 ± 12 years) with ΔSGRQ≥MCID and 14 (58 ± 18 years) with ΔSGRQ<MCID. An MRI-texture model (AUC=0.89) outperformed a clinical-measurement model (AUC=0.72) for predicting improved SGRQ, 12 months later. Top-performing textures correlated with MRI VDP (P < .05), central-airways resistance (P < .05), forced-vital-capacity (ρ = .37, P = .01) and diffusing-capacity for carbon-monoxide (ρ = .39, P = .03).
A machine learning model exclusively trained on Xe MRI ventilation textures explained improved SGRQ-scores 12 months later, and outperformed clinical models. Their unique spatial-intensity information helps build our understanding about long-COVID airway dysfunction.
即使使用胸部 CT 和功能磁共振成像（MRI），预测长新冠的纵向结局仍然具有挑战性。氙气 MRI 反映气道功能障碍，通过通气缺陷百分比（VDP）进行测量，在长新冠患者中，MRI VDP 异常，提示气道疾病。尽管 MRI VDP 和生活质量在感染新冠后 15 个月时有所改善，但两者仍存在异常。为了更好地理解长新冠患者气道疾病与生活质量改善之间的关系，我们提取了氙气通气 MRI 纹理并生成了机器学习模型，以预测感染后 15 个月生活质量的改善。
长新冠患者签署了知情同意书，同意在感染后 3 个月和 15 个月进行随访。Pyradiomics 用于提取氙气通气 MRI 纹理特征，使用随机森林分类器对其进行排名。使用排名最高的特征，基于圣乔治呼吸问卷（SGRQ）评分改善大于最小临床重要差异（MCID），对患者进行分类模型分析。使用受试者工作特征曲线下面积（AUC）、敏感性和特异性来评估分类性能。
在 44 名长新冠参与者（54±14 岁）的氙气通气 MRI 中提取了 120 个纹理特征，其中 30 名（52±12 岁）患者的 SGRQ 改善值大于 MCID，14 名（58±18 岁）患者的 SGRQ 改善值小于 MCID。MRI 纹理模型（AUC=0.89）预测 12 个月后 SGRQ 改善的性能优于临床测量模型（AUC=0.72）。表现最佳的纹理与 MRI VDP（P<.05）、中央气道阻力（P<.05）、用力肺活量（ρ=0.37，P=0.01）和一氧化碳弥散量（ρ=0.39，P=0.03）相关。
仅基于氙气 MRI 通气纹理训练的机器学习模型可以解释 12 个月后 SGRQ 评分的改善，并且优于临床模型。它们独特的空间强度信息有助于我们了解长新冠的气道功能障碍。
