IDNE 701 Β· Introduction to Neuroengineering Β· Week 1
Why is the human brain the hardest organ to measure? Physical, biological, and engineering barriers defining the diagnostic frontier.
What you will be able to do by the end of this lecture:
Section I
Every organ in the body can be imaged β but the brain resists easy non-invasive measurement.
Bone attenuates electrical potentials by 10β20 dB. Current spreads laterally through resistive cranium, blurring point dipole sources across centimeters.
Skull reflects ~99% of ultrasound acoustic energy at normal incidence. Transcranial ultrasound requires narrow acoustic bone windows (transtemporal).
Photons undergo multiple scattering in scalp and bone (mean free path ~0.1 mm). Penetration is physically capped at outer ~15 mm of cortex.
Static & RF magnetic fields pass through bone without distortion β but require multi-ton superconducting magnets and RF cages ($1Mβ7M).
Infinitely many distinct 3D current distributions inside a volume conductor can generate the exact same potential distribution on the outer boundary surface.
No single non-invasive modality captures both millimeter spatial and millisecond temporal dynamics.
Temporal: ~1 millisecond (direct post-synaptic potentials)
Spatial: ~5β10 mm (diffuse, inverse-constrained)
Directness: Direct neural electrical/magnetic readout
Temporal: ~1β5 seconds (delayed by neurovascular coupling HRF)
Spatial: ~1 mm (high 3D volumetric localization)
Directness: Indirect metabolic / blood oxygenation proxy
Section II
Comparison across physical transduction mechanisms, spatiotemporal bounds, and deployment footprint:
| Modality | Physical Transduction | Spatial Res. | Temporal Res. | Portability | Approx. Cost |
|---|---|---|---|---|---|
| EEG | Scalp electric potentials (EPSPs) | ~10 mm | ~1 ms | β β β β β Wearable | $5Kβ50K |
| MEG | Neuromagnetic fields (SQUID/OPM) | ~5 mm | ~1 ms | β ββββ Shielded Room | $2Mβ4M |
| fMRI | BOLD magnetic susceptibility | ~1 mm | ~1 s (HRF ~5s) | β ββββ Fixed Scanner | $1Mβ7M |
| fNIRS | NIR oxy/deoxy-Hb absorption | ~10 mm | ~10 ms | β β β β β Mobile Cart | $20Kβ200K |
| DTI | Water diffusion along tracts | ~1 mm | Minutes (Structural) | β ββββ MRI Hardware | (uses MRI) |
| fUS | Ultrafast Doppler microvasculature | ~0.1 mm | ~10 ms | β β β ββ Cart System | $50Kβ300K |
Section III
FDA-NIH BEST (Biomarkers, EndpointS, and other Tools) Taxonomy:
Identifies likelihood of developing condition before symptom onset (e.g., Pre-symptomatic Amyloid PET positivity in Alzheimer's).
Confirms disease presence or stratifies subtype (e.g., Interictal epileptiform discharges on EEG; DaTscan in Parkinson's).
Tracks disease progression or therapeutic response longitudinally (e.g., Serial MRI brain volumetry in MS).
Forecasts response to specific intervention or validates target engagement (e.g., TMS-EEG cortical excitability).
Why 95% of published neuroimaging biomarkers fail to reach the clinic:
Statistically significant group difference in small cohort.
Test-retest reliability, repeatability, phantom calibration.
Sensitivity, specificity, ROC AUC in blinded multi-center trials.
Demonstrated impact on patient outcomes & therapeutic choices.
Formally qualified Medical Device Development Tool (MDDT).
Section IV
Your semester design project will address one of these open clinical frontiers:
50β80% of vulnerable neurons are already lost at clinical diagnosis. We need sensing years before cognitive decline.
30% of epilepsy is drug-resistant. Non-invasive mapping of the epileptogenic zone without surgical intracranial grids.
MDD, PTSD, and schizophrenia are diagnosed solely by questionnaires. Zero FDA-cleared imaging biomarkers exist.
Point-of-care distinction between ischemic and hemorrhagic stroke in ambulances and remote triage clinics.
Closing the loop on TMS, DBS, and focused ultrasound with real-time target engagement imaging feedback.
Continuous bedside crib monitoring for preterm hypoxia, hemorrhage, and cerebral autoregulation disruption.
Section V
Setting engineering specifications for a viable clinical diagnostic tool:
| Parameter | Minimum Viable (MVP) | Aspirational Goal | Clinical Significance |
|---|---|---|---|
| Sensitivity (TPR) | β₯ 80% | β₯ 95% | Avoid missing true disease cases |
| Specificity (TNR) | β₯ 80% | β₯ 90% | Prevent invasive or harmful false-positive workups |
| Time to Result | < 60 minutes | < 10 minutes | Emergency stroke/TBI therapeutic windows |
| Operator Skill | Certified Technologist | Paramedic / Bedside Nurse | Deployment outside tertiary academic hospitals |
| System Cost | < $500K | < $50K | Global health access & outpatient clinic adoption |
Section VI
From clinical problem definition to virtual prototype simulation across 14 weeks:
Weeks 1β3
Clinical Need, Stakeholder Protocol, MRI Core Tour
Weeks 4β6
Biophysical Physics, Sensor Array, fNIRS Demo
Weeks 7β9
β
Midterm PDR, Inverse Solvers & Signal Chain
Weeks 10β12
Safety/SAR, EEG Core Visit, FDA Regulatory 510(k)
Weeks 13β14
Simulation Codebase, β
Grand Pitch Symposium
1. The "Magic Wand" Question: If you could have one perfect, unconstrained measurement of the brain, which neurological condition would you solve first?
2. Value of Portability: A $500 EEG headset vs. a $3M 7T MRI scanner. Under what clinical setting is the $500 tool undeniably superior?
3. The Correlation vs. Mechanism Trap: Why do so many ML-derived neuroimaging biomarkers fail validation when tested on independent clinical cohorts?
FDA-NIH Biomarker Working Group
Chapters 1β3: Taxonomy & Validation
NCBI Bookshelf (NBK326791) β
Baillet, S. (2017)
Nature Neuroscience, 20(3), 327β339
DOI: 10.1038/nn.4504 β
Next: Neuropathology, Clinical Stakeholders & Biomarker Definition