Реальный опыт автоматизированного обнаружения множественных склерозных повреждений в клинике клинического обучения.
Обработка статьи · этапы, ошибки и управление
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.
Martins Jarnalo C O, Dieckens D, Attrach H, Broere M, Klein Kranenbarg R A M, Wokke B, van den Bosch J H, van Rosmalen J, de Beukelaar J
Аннотация
Оригинал публикации
Real-world experience with automated multiple sclerosis lesion detection in a clinical teaching hospital.
BACKGROUND: An increase in the number of lesions between two longitudinal MRI scans is an important marker for tracking inflammation in multiple sclerosis. Accurate segmentation of new lesions is essential for assessing inflammatory activity. Quantitative evaluations of software for computer-aided MRI assessment in real-world settings, particularly those performing analysis during the MRI acquisition, are scarce. This study assesses the diagnostic value, feasibility and efficiency of computer-aided MRI assessment software integrated in the routine workflow of a teaching hospital's MS centre.
METHODS: The software visualises differences between two MRI Fluid-Attenuated Inversion Recovery sequences of the brain. Baseline and follow-up scans of 69 randomly selected people with MS (PwMS) were included. 114 MRI scan pairs were assessed by consensus reading of two neuroradiologists, and by the software. Feasibility and efficiency was evaluated with the data of 25 prospectively included PwMS. The reading times of the MRI technician, the radiologist alone, and the radiologist using the software were assessed.
RESULTS: The sensitivity (94 %) and negative predictive value (89 %) of the software were high. The F1-score was 0.6. The software identified ten true positives that were missed by the consensus reading. The specificity and overall accuracy of the software were low, due to many false positives, reducing efficiency. Using the software during follow-up MRI examinations lowered the reading time by 88 s (58 %) on average.
CONCLUSIONS: Despite the unsatisfactory diagnostic value of the software, the reading-time gain in routine practice was promising. It was feasible to implement the software in the workflow, potentially creating a one-stop shop for PwMS.