The Grand Challenges of Non-Invasive Brain Sensing & Diagnostic Biomarkers

I. The Measurement Problem: Why Is the Brain So Hard to See?

Every organ in the body can be imaged — hearts with echocardiography, livers with CT, bones with plain X-ray. But the brain resists easy measurement. Understanding why is the first step toward engineering better solutions.

• The Skull as a Low-Pass Filter The cranium presents a formidable barrier to every sensing modality.
  • Bone attenuates electromagnetic signals by 10–20 dB, acting as a spatial low-pass filter that smears cortical source activity across centimeters at the scalp surface.
  • Ultrasound waves suffer near-total reflection at the skull–brain interface (~99% impedance mismatch), limiting transcranial sonography to narrow temporal bone windows.
  • Near-infrared photons undergo extreme scattering in bone and tissue (mean free path ~0.1 mm), restricting fNIRS sensitivity to the outer ~15 mm of cortex.
  • Only strong static and RF magnetic fields (MRI) penetrate the skull without significant distortion — but at enormous cost in hardware and infrastructure.
• Signal Scale: Microvolts in a Noisy World Neural signals are extraordinarily small relative to environmental and physiological noise.
  • Scalp EEG signals range from 1–100 μV, while 60 Hz power-line noise can exceed 1 mV — a 10,000:1 interference ratio.
  • The BOLD fMRI signal change during neural activation is only 1–5% of the baseline signal, requiring extensive statistical averaging.
  • Cardiac pulsation, respiratory motion, and spontaneous vasomotor oscillations introduce physiological noise that can exceed the neural signal of interest.
• The Inverse Problem Is Ill-Posed Recovering the 3D source distribution from surface measurements is fundamentally underdetermined.
  • In EEG and MEG, infinitely many source configurations inside the brain can produce identical patterns on the scalp (non-uniqueness of the electromagnetic inverse problem).
  • Regularization assumptions (minimum norm, beamformers, dipole fitting) impose prior beliefs that may bias clinical interpretation.
  • This is not merely a computational inconvenience — it is a fundamental physical constraint (Helmholtz, 1853).
• Temporal vs. Spatial Resolution: The Fundamental Trade-Off No single modality captures neural dynamics at both millimeter and millisecond scales simultaneously.
  • Electrophysiology (EEG, MEG) offers millisecond-level temporal resolution but centimeter-scale spatial blurring.
  • Hemodynamic imaging (fMRI, fNIRS) offers millimeter spatial resolution but is delayed 4–6 seconds by the hemodynamic response function (HRF).
  • This duality drives the engineering imperative for multimodal integration — combining modalities to compensate for each other's blind spots.

II. The Modality Landscape: Tools We Have Today

Each non-invasive modality exploits a different physical phenomenon to infer neural activity. None is universally "best" — each has a distinct niche defined by its physics.

III. What Is a Diagnostic Biomarker?

A biomarker is a measurable indicator of a biological state. But not all biomarkers are diagnostically useful. The FDA and NIH BEST (Biomarkers, EndpointS, and other Tools) framework defines a hierarchy:

• Susceptibility / Risk Biomarkers Indicate predisposition to developing a disease before symptoms appear.
  • Example: Amyloid PET positivity in pre-symptomatic Alzheimer's disease.
  • Example: White matter hyperintensity volume on MRI as a vascular dementia risk factor.
• Diagnostic Biomarkers Confirm the presence of a disease or stratify patients into subtypes.
  • Example: Interictal epileptiform discharges on EEG for epilepsy diagnosis.
  • Example: Dopamine transporter SPECT (DaTscan) for Parkinson's vs. essential tremor.
  • Diagnostic biomarkers must demonstrate clinical sensitivity (true positive rate) and specificity (true negative rate) exceeding clinically meaningful thresholds.
• Monitoring Biomarkers Track disease progression or treatment response over time.
  • Example: Brain volumetric changes on serial MRI in multiple sclerosis.
  • Example: EEG spectral power ratios tracking response to anesthesia (BIS index).
• Predictive & Pharmacodynamic Biomarkers Predict response to a specific therapy or demonstrate target engagement.
  • Example: fMRI activation patterns predicting antidepressant response.
  • Example: TMS-evoked EEG potentials measuring cortical excitability changes from drug administration.
• Surrogate Endpoints vs. Clinical Endpoints A surrogate endpoint substitutes for a clinical outcome — but this substitution must be rigorously validated.
  • Many imaging biomarkers have been proposed as surrogates (e.g., hippocampal volume for cognitive decline) but few have achieved full FDA qualification.
  • The gap between "statistically associated with disease" and "FDA-qualified surrogate endpoint" is one of the largest bottlenecks in neuroimaging translation.

IV. The Six Grand Challenges

These are the defining unsolved problems at the intersection of neuroscience, imaging physics, and clinical medicine. Your semester project should address at least one.

V. What Would "Solved" Look Like? Design Targets

For your semester project, a viable diagnostic neuroimaging system should aspire to meet quantitative performance targets. Use these as starting benchmarks when framing your clinical needs specification:

VI. The Semester Project: Your Mission

Over the next 14 weeks, your team will design a multimodal neuroimaging or diagnostic biomarker system that addresses one of the grand challenges above. The project trajectory follows five phases:

• Phase 1: Discovery & Clinical Framing (Weeks 1–3) Select your target disorder, interview clinical stakeholders, visit the UAB Research MRI Core, and draft your Clinical Needs Specification.
• Phase 2: Biophysics & Sensors (Weeks 4–6) Model your contrast mechanism, design your sensor array architecture, and participate in the live fNIRS demonstration.
• Phase 3: Architecture & Midterm Review (Weeks 7–9) Present your preliminary design at the mid-semester pin-up, then build acquisition sequences and reconstruction pipelines.
• Phase 4: Safety & Regulation (Weeks 10–12) Validate artifact robustness, visit the Neuroengineering EEG Core, and formulate your FDA regulatory strategy.
• Phase 5: Verification & Grand Pitch (Weeks 13–14) Build your virtual prototype simulation, draft an IRB-ready clinical protocol, and deliver your final pitch at the Grand Symposium.

See the complete 14-week interactive timeline →

VII. Assigned & Recommended Reading

Required for Week 1

• Biomarkers, EndpointS, and other Tools (BEST) Resource Required
  FDA-NIH Biomarker Working Group
  NCBI Bookshelf, 2016 (updated 2024). Chapters 1–3.
• Magnetoencephalography for brain electrophysiology and imaging Required
  Baillet, S.
  Nature Neuroscience, 20(3), 327–339 (2017).

Recommended Deep Dives

• NIA-AA Research Framework: Toward a biological definition of Alzheimer's disease
  Jack, C. R. et al.
  Alzheimer's & Dementia, 14(4), 535–562 (2018).
• Best practices for fNIRS publications
  Yücel, M. A. et al.
  Neurophotonics, 8(1), 012101 (2021).
• EEG source imaging: a practical review of the analysis steps
  Michel, C. M. & Brunet, D.
  Frontiers in Neurology, 10, 325 (2019).
• A hitchhiker's guide to functional magnetic resonance imaging
  Soares, J. M. et al.
  Frontiers in Neuroscience, 10, 515 (2016).