StimIQ

Inspiration

On average, 80% of the population knows someone with Parkinson's disease (PD). As the fastest-growing neurodegenerative disease globally, the already profound impact of PD is only accelerating. Our team has personal connections to people living with PD, some of whom rely on advanced, later-stage care formats like deep brain stimulation (DBS).

Despite its potential, the exact mechanisms of DBS remain partially unclear, and perceived benefits vary drastically from patient to patient. Currently, DBS requires an arduous, iterative loop of refining electrode parameters. This process is often based on insufficient data and limited clinical time, leading to the sub-optimal use of a treatment that possesses the potential to grant 10 to 15 years of significantly higher quality of life.

The challenge was set: How can we reduce the time it takes to optimize DBS for individual patients? How can we lower the technical barrier for doctors performing these tunings to make the treatment more accessible? And how can we calibrate parameters more accurately by factoring in a wider array of variables?

To ensure our approach was grounded in clinical reality, we consulted Valtteri Kaasinen (MD, PhD, Professor of Neurology at the University of Turku) and Kai Lehtimäki (MD, PhD, Associate Professor at Tampere University Hospital). Their deep, experiential feedback on the problem space and our proposed solutions provided the validation and encouragement we needed to chase this incredibly technical and ambitious goal.

What Is the Solution

StimIQ is an intelligent clinical tool designed to help doctors optimally tune DBS systems for individual patients. The platform introduces a novel personalized optimization cycle that continuously tracks patient data—combining subjective, diary-based Patient Reported Outcomes (PROMs) with objective IMU sensor data—to update the Bayesian posterior of the DBS parameter setup.

Crucially, StimIQ shifts the treatment paradigm from a single, rushed 6-month checkup to a continuous, proactive 3-step process: gathering pre-appointment data, facilitating guided in-clinic tuning, and monitoring the post-appointment follow-up phase.

To ensure clinical trust, the tool provides a text-based explanation for why the algorithm suggests specific parameter adjustments. It cross-references these suggestions against a FAISS vector database containing a vast library of clinical guidelines. By presenting previous settings, comprehensive patient data, proposed changes, and plain-text explanations in a streamlined UI, StimIQ actively prevents the "data tsunami" that our interviewed experts highlighted as a major burden. Ultimately, it makes DBS calibration faster, safer, and highly effective.

How we built it

We tackled the problem by fusing advanced machine learning with a highly user-centric clinical interface. At the core, we utilized Bayesian optimization to manage the dynamic, highly individualized nature of DBS parameter tuning. To process complex clinical guidelines and generate readable, actionable insights, we implemented a Retrieval-Augmented Generation (RAG) architecture powered by a FAISS vector database.

The most critical—and complex—component of our build was data integration. We designed the architecture to ingest and harmonize completely divergent data streams, merging rigid, objective IMU sensor readings with highly subjective patient diary entries (tracking nuances like overall "feeling" and quality of life). Finally, we wrapped this dense algorithmic logic into a clean, minimalist frontend, ensuring the tool remains intuitive for doctors operating under severe time constraints.

Challenges we ran into

Explainability in Healthcare: Finding a reliable way to make complex algorithmic parameter decisions transparent and explainable to medical professionals so they feel confident applying them.
Hardware Simulation: Developing a way to accurately simulate IMU sensor responses based on specific parameter adjustments sent to a DBS device.
Merging the Subjective and Objective: Creating a mathematical balance that successfully combines qualitative data (patient mood, perceived quality of life) with quantitative data (sensor metrics).

Accomplishments that we're proud of

Expert Validation: Securing validation and incredibly detailed pain-point analysis from two leading neurology professors, grounding our technical hack with profound clinical relevance.
Holistic Optimization: Designing a system that moves beyond just monitoring beta oscillations (which can dangerously oversimplify the problem). Instead, StimIQ respects multiple indicators, rightfully placing Patient Reported Outcomes (PROMs) at the center of the tuning process.
A Doctor-First UI: Successfully abstracting complex Bayesian optimization and vector database queries into a simple, low-friction interface that allows doctors to safely tune DBS devices without needing to be specialized technical experts.

What we learned

Validation is Everything: Refining our problem space through rigorous user and professional interviews yielded massive, immediate benefits to our product architecture.
Data Quality > Data Quantity: We learned from both the interviews and by testing, that simply making more data visible does not equal a better system; it often just overwhelms the user. Curation and actionable insights are key.
The Power of Simulation: We discovered that inevitable gaps in data flows can be successfully and relatively accurately patched using simulation models for demonstration purposes.

What's next for StimIQ

Our immediate goal is to solve the glaring clinical bottlenecks in current DBS therapies, but our long-term vision extends far beyond invasive procedures. The true value of StimIQ lies in its human-in-the-loop optimization engine. Moving forward, we are focusing on three scalable pillars:

Hardware-Agnostic Care: Currently, proprietary software from manufacturers like Boston Scientific, Medtronic, and Abbott creates immense clinical friction. By developing a single, unified dashboard, we can significantly reduce cognitive load and training time for doctors, ensuring seamless and standardized patient care regardless of the underlying hardware brand.
Democratizing Clinical Expertise: DBS management is currently gatekept by specialized university hospitals, leaving regional clinics and ERs unequipped. By providing algorithmically validated parameter estimates to non-specialized doctors, StimIQ eliminates clinical guesswork and the fear of malpractice, ensuring safe, equitable care no matter where a patient is treated or when an emergency strikes.
Scaling Beyond Invasive Treatments: The ultimate hope for neurology over the next 10-15 years is to phase out invasive surgeries in favor of non-invasive therapies. Because StimIQ’s human-in-the-loop optimization engine is fundamentally treatment-agnostic, it is perfectly positioned to seamlessly adapt to and optimize next-generation therapies, such as advanced non-invasive neuromodulation or highly personalized pharmacology.