Stroke is the leading cause of long-term disability and the second leading cause of death worldwide.1 Strokes are either ischemic, which account for over 85% of all strokes and occur due to the interruption of blood flow caused by thrombosis, or hemorrhagic, which occur due to bleeding.2,3 During ischemic stroke, a reduction or complete loss in blood supply causes a starved oxygen environment and leads to neuronal damage within minutes and ultimately to sensorimotor and cognitive impairment.2,3
Functional near-infrared spectroscopy (fNIRS) is a family of noninvasive neuroimaging techniques that estimate concentrations of oxygenated and deoxygenated hemoglobin near the surface of the brain and thus infer relative changes in neural activation.1 There are various technical approaches to NIRS measurements, most prominently the continuous-wave, frequency domain, and time-domain techniques. Each approach provides unique benefits, utilizes different instrumentation, and yields different raw data.1,2
Objective and Impact Statement. Segmentation of blood vessels from two-photon microscopy (2PM) angiograms of brains has important applications in hemodynamic analysis and disease diagnosis. Here, we develop a generalizable deep learning technique for accurate 2PM vascular segmentation of sizable regions in mouse brains acquired from multiple 2PM setups. The technique is computationally efficient, thus ideal for large-scale neurovascular analysis. Introduction. Vascular segmentation from 2PM angiograms is an important first step in hemodynamic modeling of brain vasculature. Existing segmentation methods based on deep learning either lack the ability to generalize to data from different imaging systems or are computationally infeasible for large-scale angiograms. In this work, we overcome both these limitations by a method that is generalizable to various imaging systems and is able to segment large-scale angiograms. Methods. We employ a computationally efficient deep learning framework with a loss function that incorporates a balanced binary-cross-entropy loss and total variation regularization on the network’s output. Its effectiveness is demonstrated on experimentally acquired in vivo angiograms from mouse brains of dimensions up to . Results. To demonstrate the superior generalizability of our framework, we train on data from only one 2PM microscope and demonstrate high-quality segmentation on data from a different microscope without any network tuning. Overall, our method demonstrates 10× faster computation in terms of voxels-segmented-per-second and 3× larger depth compared to the state-of-the-art. Conclusion. Our work provides a generalizable and computationally efficient anatomical modeling framework for brain vasculature, which consists of deep learning-based vascular segmentation followed by graphing. It paves the way for future modeling and analysis of hemodynamic response at much greater scales that were inaccessible before.
Objective and Impact Statement . Segmentation of blood vessels from two-photon microscopy (2PM) angiograms of brains has important applications in hemodynamic analysis and disease diagnosis. Here, we develop a generalizable deep learning technique for accurate 2PM vascular segmentation of sizable regions in mouse brains acquired from multiple 2PM setups. The technique is computationally efficient, thus ideal for large-scale neurovascular analysis. Introduction . Vascular segmentation from 2PM angiograms is an important first step in hemodynamic modeling of brain vasculature. Existing segmentation methods based on deep learning either lack the ability to generalize to data from different imaging systems or are computationally infeasible for large-scale angiograms. In this work, we overcome both these limitations by a method that is generalizable to various imaging systems and is able to segment large-scale angiograms. Methods . We employ a computationally efficient deep learning framework with a loss function that incorporates a balanced binary-cross-entropy loss and total variation regularization on the network’s output. Its effectiveness is demonstrated on experimentally acquired in vivo angiograms from mouse brains of dimensions up to . Results . To demonstrate the superior generalizability of our framework, we train on data from only one 2PM microscope and demonstrate high-quality segmentation on data from a different microscope without any network tuning. Overall, our method demonstrates 10× faster computation in terms of voxels-segmented-per-second and 3× larger depth compared to the state-of-the-art. Conclusion . Our work provides a generalizable and computationally efficient anatomical modeling framework for brain vasculature, which consists of deep learning-based vascular segmentation followed by graphing. It paves the way for future modeling and analysis of hemodynamic response at much greater scales that were inaccessible before.
With the rapidly expanding list of fluorescent probes for neuronal activity and the development of large-scale measurements,1 widefield microscopy has become the measurement modality of choice for mesoscale studies of neuronal circuits and underlying behavior when cellular resolution is not required.2–7 Probes available in distinct spectral variants8 can be spectrally multiplexed for simultaneous imaging of different aspects of neuronal activity.9 Using multiplexed probes for in vivo widefield imaging of head-fixed mice with “crystal skull” cranial windows10 or transcranial imaging, optionally combined with thinned-skull preparations,11 provides a view of the entire dorsal part of the cerebral cortex including the primary sensory, motor, and higher association areas.12,13
The use of fNIRS has grown exponentially since its inception,1,2 expanding our understanding of brain activity in a variety of contexts, such as in psychiatry,3–7 naturalistic environments,8 developmental research,9–11 low-resource contexts and global health,11–17 and for hyperscanning studies.18–21 Benefits of functional near-infrared spectroscopy (fNIRS) over other neuroimaging modalities are well-defined for its comparative motion artifact resistance, temporal resolution, portability and wearability, operational requirements and ease of use, comfort, and general ecological validity.22