Neurophotonics · 2026
Physics-informed neural network for mapping vascular and tissue dynamics using laser speckle contrast imaging.
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.
Li Shuying, Tang Rockwell, Liu Bingxue, Krepulec Victoria, Donovan Will, Boas David, Cheng Xiaojun, Tian Lei
Аннотация
SIGNIFICANCE: Quantitatively mapping both cerebral blood flow and tissue dynamics from laser speckle contrast imaging (LSCI) is powerful for studying cerebral blood flow and neurovascular coupling, particularly in the context of stroke. Conventional multi-exposure fitting is slow and difficult to scale. Efficient, physically grounded methods are needed to extract both vascular and tissue dynamic biomarkers from LSCI data.
AIM: We aim to develop and validate a physics-informed neural network (PINN) that quantitatively estimates fast (vascular) and slow (tissue-related) speckle decorrelation parameters directly from LSCI measurements without requiring ground-truth labels.
APPROACH: We developed a PINN to estimate fast (vascular) and slow (tissue-related) speckle decorrelation parameters directly from multi-exposure LSCI data without requiring ground-truth labels. The analytical LSCI model is embedded in the network loss function, enforcing consistency with speckle physics during training. The model operates in a self-supervised manner and performs pixel-wise inference across full-field images. The framework was evaluated using in vivo mouse stroke LSCI datasets.
RESULTS: The PINN accurately recovered fast decorrelation rates associated with cerebral blood flow and slower dynamics linked to tissue and cellular motion. The parameter maps closely match those from traditional nonlinear fitting, but at orders-of-magnitude higher speed, reducing analysis to under 1 s per image. It also generalizes to unseen subjects and remains robust under noise.
CONCLUSIONS: Our approach establishes physics-informed learning as a practical framework for near-real-time extraction of vascular and cellular biomarkers from LSCI, enabling longitudinal monitoring of stroke progression and potentially facilitating clinical translation.
AIM: We aim to develop and validate a physics-informed neural network (PINN) that quantitatively estimates fast (vascular) and slow (tissue-related) speckle decorrelation parameters directly from LSCI measurements without requiring ground-truth labels.
APPROACH: We developed a PINN to estimate fast (vascular) and slow (tissue-related) speckle decorrelation parameters directly from multi-exposure LSCI data without requiring ground-truth labels. The analytical LSCI model is embedded in the network loss function, enforcing consistency with speckle physics during training. The model operates in a self-supervised manner and performs pixel-wise inference across full-field images. The framework was evaluated using in vivo mouse stroke LSCI datasets.
RESULTS: The PINN accurately recovered fast decorrelation rates associated with cerebral blood flow and slower dynamics linked to tissue and cellular motion. The parameter maps closely match those from traditional nonlinear fitting, but at orders-of-magnitude higher speed, reducing analysis to under 1 s per image. It also generalizes to unseen subjects and remains robust under noise.
CONCLUSIONS: Our approach establishes physics-informed learning as a practical framework for near-real-time extraction of vascular and cellular biomarkers from LSCI, enabling longitudinal monitoring of stroke progression and potentially facilitating clinical translation.