Frontiers in endocrinology · 2026
Sequence-specific radiomics for diagnosis of spinal bone loss.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Xue Tingyu, Li Yaguang, Zhao Huayi, Zhang Tianzi, Wang Jiayi, Jiang WenHao, Lei Licun, Wang Yong
Аннотация
OBJECTIVE: To establish a sequence-specific predictive model for spinal bone loss by leveraging conventional lumbar MRI, targeting abnormal bone density or osteoporosis differentiations.
METHODS: A total of 320 MR scans from 160 patients (52 men and 108 women; mean age 61.27 ± 12.72 years) who underwent lumbar MRI and quantitative computed tomography (QCT) examinations were retrospectively enrolled in this study cohort. Radiomic features were extracted from the lumbar spine MR images. With QCT as the reference standard, six radiomic-based machine learning models including K-nearest neighbor (KNN), support vector machine (SVM), Linear Discriminant Analysis (LDA), logistic regression (LR), stochastic gradient descent (SGD), Gaussian NB were developed to predict abnormal bone density and osteoporosis using T1WI alone, T2WI alone, and the combined T1WI+T2WI. The dataset was randomly split into a training/validation set and a testing set in a 7:3 ratio. The performance metrics of the models were calculated and evaluated.
RESULTS: Among the six machine learning models evaluated, T1WI and T2WI each exhibited prominent advantages for predicting osteoporosis and abnormal bone mass, respectively. Take KNN as an example. T1WI achieved the highest AUC (0.821) for predicting osteoporosis on test set (mean of 10 repeated evaluations), significantly higher than T2WI (AUC = 0.782) and the combined T1WI+T2WI approach (AUC = 0.775). In contrast, T2WI demonstrated superior performance for the prediction of abnormal bone density, with an AUC of 0.942 (T1WI and T1WI+T2WI were 0.884 and 0.923, respectively).
CONCLUSION: Our investigation into predicting abnormal bone density and osteoporosis from lumbar spine MRI sequences shows that predictive efficacy is sequence-dependent. T1WI features proved more effective for osteoporosis identification, while T2WI features were better for abnormal bone density prediction, highlighting the importance of sequence selection based on target pathology.
METHODS: A total of 320 MR scans from 160 patients (52 men and 108 women; mean age 61.27 ± 12.72 years) who underwent lumbar MRI and quantitative computed tomography (QCT) examinations were retrospectively enrolled in this study cohort. Radiomic features were extracted from the lumbar spine MR images. With QCT as the reference standard, six radiomic-based machine learning models including K-nearest neighbor (KNN), support vector machine (SVM), Linear Discriminant Analysis (LDA), logistic regression (LR), stochastic gradient descent (SGD), Gaussian NB were developed to predict abnormal bone density and osteoporosis using T1WI alone, T2WI alone, and the combined T1WI+T2WI. The dataset was randomly split into a training/validation set and a testing set in a 7:3 ratio. The performance metrics of the models were calculated and evaluated.
RESULTS: Among the six machine learning models evaluated, T1WI and T2WI each exhibited prominent advantages for predicting osteoporosis and abnormal bone mass, respectively. Take KNN as an example. T1WI achieved the highest AUC (0.821) for predicting osteoporosis on test set (mean of 10 repeated evaluations), significantly higher than T2WI (AUC = 0.782) and the combined T1WI+T2WI approach (AUC = 0.775). In contrast, T2WI demonstrated superior performance for the prediction of abnormal bone density, with an AUC of 0.942 (T1WI and T1WI+T2WI were 0.884 and 0.923, respectively).
CONCLUSION: Our investigation into predicting abnormal bone density and osteoporosis from lumbar spine MRI sequences shows that predictive efficacy is sequence-dependent. T1WI features proved more effective for osteoporosis identification, while T2WI features were better for abnormal bone density prediction, highlighting the importance of sequence selection based on target pathology.