Abstract
Stroke poses an immense public health burden and remains among the primary causes of death and disability worldwide. Emergent therapy is often precluded by late or indeterminate times of onset before initial clinical presentation. Rapid, mobile, safe and low-cost stroke detection technology remains a deeply unmet clinical need. Past studies have explored the use of microwave and other small form-factor strategies for rapid stroke detection; however, widespread clinical adoption remains unrealized. Here, we investigated the use of microwave scattering perturbations from ultra wide-band antenna arrays to learn dielectric signatures of disease. Two deep neural networks (DNNs) were used for: (1) stroke detection (“classification network”), and (2) characterization of the hemorrhage location and size (“discrimination network”). Dielectric signatures were learned on a simulated cohort of 666 hemorrhagic stroke and control subjects using 2D stochastic head models. The classification network yielded a stratified K-fold stroke detection accuracy > 94% with an AUC of 0.996, while the discrimination network resulted in a mean squared error of < 0.004 cm and < 0.02 cm, for the stroke localization and size estimation, respectively. We report a novel approach to intelligent diagnostics using microwave wide-band scattering information thus circumventing conventional image-formation.
Highlights
Stroke poses an immense public health burden and remains among the primary causes of death and disability worldwide
Other imaging approaches have been explored for stroke detection, including measurement of electrical properties (EP) of tissues using transcranial currents; dielectric map reconstruction is limited by the localization of induced currents to the tissue surfaces and the ill-posedness of the reconstruction[17]; deep-seated structures may not be unambiguously resolved with high fidelity, limiting anatomic and physiologic detail and clinical applicability[18]
Upon convergence of the classification network, test set prediction resulted in sensitivity = 0.89, precision = 1.0, specificity = 1.0, accuracy = 0.94, and f1-score = 0.94
Summary
Stroke poses an immense public health burden and remains among the primary causes of death and disability worldwide. Subarachnoid hemorrhage, such as from rupture of cerebral aneurysms, is a potentially devastating injury occurring precipitously and often without warning in otherwise healthy individuals[7,8]; including the sizeable percentage of those succumbing even before hospital arrival, 30-day mortality rates in some series exceed 40%9,10 Such diseases demand fast, accurate, and safe characterization to facilitate diagnosis, prognostication, and treatment selection[7,8]. Magnetic resonance imaging (MRI) devices have been explored in acute-care s ettings[11], but wide scale dissemination to the pre-clinical environment is limited by cost, technical and logistical constraints (e.g. shielding, size, safety, etc.), and access in underserved p opulations[12]. Near-field microwave imaging (MI) systems estimate the spatial distribution of tissue EP by measuring the scattered field from an array of antennas, as initially described by Lin and Clark[18] and subsequently by Semnov19,20, Presson[21], and Abbosh[22]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
