Welcome to the Stroke Innovation (& Machine Learning) Lab. Our domain is clinical neurology with an emphasis on neurovascular care; our method is quality improvement, increasingly powered by sound as a biomarker. PI: Dr. Houman Khosravani, Division of Neurology, University of Toronto.
We pioneer machine-learning applications that elevate quality improvement in acute stroke care. Our flagship, Project MASA (Machine Learning Assisted Swallowing Assessment), analyzes voice biomarkers to rethink swallowing assessment — Sonic Diagnosis.
The lab is part of the Neurology Quality and Innovations Lab (NQIL) at the University of Toronto, founded with Drs. Sara Mitchell and Charles Kassardjian, and works with support from T-CAIREM and Sunnybrook Health Sciences Centre, leveraging bedside physiologic recordings and deep learning for audio signals. See also ML4QI at U of T Neurology.
How MASA works: spectral analysis of the human voice during swallowing assessment, classified against TOR-BSST labelling. Our 68-patient proof-of-concept at a comprehensive stroke centre trained DenseNet and ConvNeXt models (plus an ensemble) on Mel-spectrogram images of vowel and sentence recordings (Frontiers in Neuroscience, 2023; open-source code), and Tuning In (CHIL 2024; arXiv:2402.10100) benchmarked audio classifiers with limited clinical data across two first-of-their-kind datasets — NIHSS speech and sustained vowels. Visual Transformers are next.
Typically 1 T-CAIREM summer student per year, occasionally SRI / CREMS. Interested in QI or clinical ML in neurology? Contact the lab and say so.
A training-free pipeline that turns an image, GIF, or video into 5.1 spatial audio — segmenting visual sound sources, estimating 3D position from monocular depth, synthesizing mono audio per source, and rendering with room acoustics. Led by Rishit Dagli; accepted as a SIGGRAPH 2025 poster. Clinical horizon: richer auditory experience for patients with vision loss, including bilateral occipital stroke. Project site · arXiv:2406.06612 · ACM record.
The next phase of AI4QI: ML voice analysis for dysphagia screening beyond stroke, including palliative-care settings. Announced January 2025 with AFP MD funding; data recording is complete, with active analysis through 2026. Luke Zhang joined as the Summer 2026 T-CAIREM student, working alongside Dr. Emmanuel Adegunna (PGY5 Neurology) as his senior resident research project.
A project in development aiming to elevate stroke-care quality across institutions globally, from a QI perspective — including benchmarking against standards in acute stroke care. More details forthcoming. (Dr. B. Tilley contributes to 1stroke.org alongside LLM-in-EHR work.)
AI for Promotion EXcellence — with Dr. Brian Wong, a principled framework for AI in U of T senior academic promotion: transparent, equitable synthesis of research, teaching, and service portfolios that augments — never replaces — human judgment in high-stakes decisions.
| Member | Base | Work |
|---|---|---|
| Rishit Dagli | Computer Science, U of T · T-CAIREM 2024 award | Project MASA; pioneered SEE2SOUND (now 1-year studentship at NVIDIA) |
| Dr. A. Balachandar | Neurology, U of T | Project MASA — now Independent Investigator / Alumni Collaborator; Movement Disorders Fellow & PhD candidate (clinician-scientist program), intracranial electrophysiology & DBS |
| Rami Saab | BEng McMaster · MASc U of T · Temerty Medicine | Project MASA (vocal features) — now Neurology Resident, U of T |
| Eptehal Nashnoush, MSc | U of T Data Science · T-CAIREM | Project MASA; Google CSRMP; co-founded Toronto Health Datathon (2023) |
| H. Mahdi | Western University, Medicine | Project MASA |
| Dr. B. Tilley | Neurology PGME, U of T | LLMs in EHRs; 1stroke.org |
| Dr. E. Adegunna | Neurology PGME, U of T | Palliative care in stroke; Pal-MASA |
| Diya Ahmed | UGME, U of T Medicine | QI in Neurology; LLMs in Neurology |
| Rebecca Wu | UGME, U of T Medicine | LLMs in Neurology |
| Sofia Berrada | Undergraduate, U of T | ML4QI in stroke |
| Jake Lance | Undergraduate, U of T | ML4QI in stroke |
Dr. A. Balachandar (Neurology PGY4/5) · R. Saab · E. Nashnoush, MSc (T-CAIREM 2023 award) · H. Mahdi · Dr. B. Sivanandan (Palliative Care in Stroke) · Dr. M. Mahendiran (Family Medicine; palliative care research) · Georges Khalaf (CREMS Summer 2024) · Dr. E. Adegunna (PGY2) · Dr. B. Tilley (PGY1) · Diya Ahmed.
Dr. A. Balachandar · R. Saab (T-CAIREM 2022 award) · M. Panchal (CREMS).
Luke Zhang (U of T) — T-CAIREM Summer 2026, Pal-MASA, with Dr. E. Adegunna (PGY5 Neurology senior resident project). Hamza Mahdi (Schulich Medicine graduate; BEng TMU, MEng Waterloo) contributed to Tuning In and is pursuing Orthopedic Surgery at U of T. Full bios on the SiLab team page.
A code stroke is a resuscitation. SiLab introduced neurovascular resuscitation — applying medical and trauma resuscitation principles to hyperacute stroke — and built a simulation program to teach it through Crisis Resource Management: situational awareness, role clarity, closed-loop communication, cognitive-load mitigation, and structured debriefing.
From 2019 in situ ED simulation, the work has grown into interprofessional streams across code stroke nursing, PGY1 Stroke Boot Camp, SRC-TO/FU/Ottawa, B4 unit skills days, EVT sedation, and CSRN-led Door-to-CT improvement. With NICE, it now includes neurovascular EVT simulation.
Simulation is used as a QI instrument: alongside PDSA redesign, Sunnybrook median door-to-needle time fell 25%, from 38.5 to 29 minutes. Explore every stream, outcome, course, acknowledgement, and reference on the dedicated page.
Despite major advances, many patients still face substantial morbidity and mortality after stroke. We champion evidence-based protocols that seamlessly integrate palliative medicine into acute stroke treatment — defining best practice so patients receive optimal, patient-centred outcomes across the whole care spectrum.
Members of the lab also contribute to Stroke FM, the official podcast of the Canadian Stroke Consortium.
