Epilepsy research · 2021
Deep learning-based diagnosis of temporal lobe epilepsy associated with hippocampal sclerosis: An MRI study.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
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.
Ito Yosuke, Fukuda Masafumi, Matsuzawa Hitoshi, Masuda Hiroshi, Kobayashi Yu, Hasegawa Naoya, Kitaura Hiroki, Kakita Akiyoshi, Fujii Yukihiko
Аннотация
PURPOSE: The currently available indicators-sensitivity and specificity of expert radiological evaluation of MRIs-to identify mesial temporal lobe epilepsy (MTLE) associated with hippocampal sclerosis (HS) are deficient, as they cannot be easily assessed. We developed and investigated the use of a novel convolutional neural network trained on preoperative MRIs to aid diagnosis of these conditions.
SUBJECTS AND METHODS: We enrolled 141 individuals: 85 with clinically diagnosed mesial temporal lobe epilepsy (MTLE) and hippocampal sclerosis International League Against Epilepsy (HS ILAE) type 1 who had undergone anterior temporal lobe hippocampectomy were assigned to the MTLE-HS group, and 56 epilepsy clinic outpatients diagnosed as nonepileptic were assigned to the normal group. We fine-tuned a modified CNN (mCNN) to classify the fully connected layers of ImageNet-pretrained VGG16 network models into the MTLE-HS and control groups. MTLE-HS was diagnosed using MRI both by the fine-tuned mCNN and epilepsy specialists. Their performances were compared.
RESULTS: The fine-tuned mCNN achieved excellent diagnostic performance, including 91.1% [85%, 96%] mean sensitivity and 83.5% [75%, 91%] mean specificity. The area under the resulting receiver operating characteristic curve was 0.94 [0.90, 0.98] (DeLong's method). Expert interpretation of the same image data achieved a mean sensitivity of 73.1% [65%, 82%] and specificity of 66.3% [50%, 82%]. These confidence intervals were located entirely under the receiver operating characteristic curve of the fine-tuned mCNN.
CONCLUSIONS: Deep learning-based diagnosis of MTLE-HS from preoperative MR images using our fine-tuned mCNN achieved a performance superior to the visual interpretation by epilepsy specialists. Our model could serve as a useful preoperative diagnostic tool for ascertaining hippocampal atrophy in patients with MTLE.
SUBJECTS AND METHODS: We enrolled 141 individuals: 85 with clinically diagnosed mesial temporal lobe epilepsy (MTLE) and hippocampal sclerosis International League Against Epilepsy (HS ILAE) type 1 who had undergone anterior temporal lobe hippocampectomy were assigned to the MTLE-HS group, and 56 epilepsy clinic outpatients diagnosed as nonepileptic were assigned to the normal group. We fine-tuned a modified CNN (mCNN) to classify the fully connected layers of ImageNet-pretrained VGG16 network models into the MTLE-HS and control groups. MTLE-HS was diagnosed using MRI both by the fine-tuned mCNN and epilepsy specialists. Their performances were compared.
RESULTS: The fine-tuned mCNN achieved excellent diagnostic performance, including 91.1% [85%, 96%] mean sensitivity and 83.5% [75%, 91%] mean specificity. The area under the resulting receiver operating characteristic curve was 0.94 [0.90, 0.98] (DeLong's method). Expert interpretation of the same image data achieved a mean sensitivity of 73.1% [65%, 82%] and specificity of 66.3% [50%, 82%]. These confidence intervals were located entirely under the receiver operating characteristic curve of the fine-tuned mCNN.
CONCLUSIONS: Deep learning-based diagnosis of MTLE-HS from preoperative MR images using our fine-tuned mCNN achieved a performance superior to the visual interpretation by epilepsy specialists. Our model could serve as a useful preoperative diagnostic tool for ascertaining hippocampal atrophy in patients with MTLE.