Rapid localization of large-vessel occlusion (LVO) is critical for effective treatment of acute ischaemic stroke. Doppler ultrasound provides a portable and non-invasive means of assessing cerebral hemodynamics, but its clinical use is limited by operator dependence, signal noise, and reduced robustness under real-world acquisition conditions, particularly in emergency and prehospital settings. This work presents an interpretable, physics-informed framework for cerebral arterial occlusion localization based on discriminant analysis of Doppler spectral features. Rather than targeting fine-grained vessel identification [1], the approach focuses on clinically relevant arterial territories (anterior, middle, posterior cerebral arteries and vertebral arteries), consistent with acute stroke triage requirements. A large, labeled database of cerebral blood flow waveforms is generated using 1D hemodynamic simulations (Nektar 1D) [2], in which occlusion locations are explicitly prescribed. Doppler velocity signals derived from carotid arteries are transformed into the frequency domain, and discriminative spectral features are identified through a systematic frequency-band selection strategy. Supervised classification is performed using Linear Discriminant Analysis (LDA), chosen for its robustness, data efficiency, and interpretability. A key contribution is the explicit modeling of realistic Doppler noise sources, including additive Gaussian noise, signal-dependent Poisson noise from blood scatterer statistics, and impulsive motion-related artifacts. Results show that low-frequency spectral components (2–12 Hz) carry the most discriminative information, yielding high localization accuracy even under severe noise contamination. Notably, reliable occlusion localization is achieved using single-sided carotid measurements, suggesting that diagnostically relevant information propagate across hemispheres. The proposed approach is computationally lightweight and suitable for real-time deployment, supporting its potential translation to point-of-care and prehospital stroke assessment. REFERENCES [1] A. Sen, L. Navarro, S. Avril, M. Aguirre, “A data-driven computational methodology towards a pre-hospital acute ischaemic stroke screening tool using haemodynamics waveforms”, Computer Methods and Programs in Biomedicine 244 (2024) 107982. [2] J. Alastruey, K. Parker, J. Peiró, S. Byrd, S. Sherwin, “Modelling the circle of Willis to assess the effects of anatomical variations a