Frontiers in neuroscience · 2026
Evaluating the detection of small brain lesions in magnetic resonance using deep learning.
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.
Nogales Alberto, Sicilia Miguel Ángel, de la Pinta Carolina, García-Barriocanal Elena, García-Tejedor Álvaro J, Guadalupe Diego
Аннотация
INTRODUCTION: Detecting brain lesions using Artificial Intelligence methods has been a focus of prior research, with numerous datasets supporting this task to improve clinical results. However, evaluation metrics and reported results often summarise overall performance without considering variations in lesion size. In clinical practice, the detection of small tumours is particularly critical for early diagnosis and treatment effectiveness.
METHODS: This study evaluates the performance of Artificial Intelligence models on established datasets with a specific focus on lesion identification and lesion size. We introduce a novel Deep Learning model tailored to detect small brain tumours in Magnetic Resonance Imaging, integrating a clinically defined "small tumour" concept into both the training and evaluation processes.
RESULTS: The proposed approach demonstrates robust performance, achieving loss values ranging from 1.5 to 11.1 and Dice Scores between 96.3% and 98.1% across multiple datasets.
DISCUSSION: The main contribution of this work is the incorporation of clinically meaningful lesion-size information into model development and assessment. These findings suggest that explicitly considering small tumours can improve the clinical relevance of Artificial Intelligence systems for brain lesion detection and support earlier diagnosis and more effective treatment planning.
METHODS: This study evaluates the performance of Artificial Intelligence models on established datasets with a specific focus on lesion identification and lesion size. We introduce a novel Deep Learning model tailored to detect small brain tumours in Magnetic Resonance Imaging, integrating a clinically defined "small tumour" concept into both the training and evaluation processes.
RESULTS: The proposed approach demonstrates robust performance, achieving loss values ranging from 1.5 to 11.1 and Dice Scores between 96.3% and 98.1% across multiple datasets.
DISCUSSION: The main contribution of this work is the incorporation of clinically meaningful lesion-size information into model development and assessment. These findings suggest that explicitly considering small tumours can improve the clinical relevance of Artificial Intelligence systems for brain lesion detection and support earlier diagnosis and more effective treatment planning.