Introduction to Brain-Computer Interfaces/Module 6: Clinical Applications

Module 6: Clinical Applications

Introduction

For a person with locked-in syndrome—fully conscious but completely paralyzed—a brain-computer interface is not a convenience or novelty. It is a lifeline. The ability to select letters, express needs, or simply answer "yes" or "no" can mean the difference between isolation and connection, between despair and dignity.

Clinical applications represent the most compelling and impactful uses of BCI technology. While consumer devices might help someone meditate or play games, clinical BCIs restore fundamental human capabilities: communication, movement, independence. These applications drive much of the research funding, inspire the most dedicated researchers, and demonstrate the profound potential of neurotechnology.

This module explores how BCIs are used in clinical settings—from communication aids that give voice to the voiceless, to motor systems that restore movement to paralyzed limbs, from seizure prediction that could prevent dangerous events to neurofeedback that may treat neurological and psychiatric conditions. We'll examine both the remarkable successes and the ongoing challenges, providing a realistic picture of where clinical BCIs stand today and where they're heading.

Learning Objectives

After completing this module, you will be able to:

• Describe the major clinical applications of BCIs
• Explain how P300 and SSVEP spellers enable communication
• Understand motor BCIs for cursor control, robotic arms, and exoskeletons
• Discuss BCI-based rehabilitation approaches for stroke
• Describe seizure prediction and responsive neurostimulation
• Evaluate neurofeedback applications and their evidence base
• Consider ethical implications of clinical neurotechnology

Communication BCIs

For individuals who cannot speak or type due to severe motor impairment, BCIs may provide the only means of communication. This section examines the paradigms that enable BCI-based communication.

Who Needs Communication BCIs?

Several conditions can leave individuals with intact cognition but severely impaired motor function:

Amyotrophic Lateral Sclerosis (ALS): Also known as Lou Gehrig's disease, ALS progressively destroys motor neurons. Early stages affect limbs; late stages may affect all voluntary muscles including those controlling speech, eye movement, and breathing. Approximately 10% of ALS patients reach a "completely locked-in" state where even eye movement is lost.

Locked-in Syndrome: Typically caused by brainstem stroke, locked-in syndrome involves complete paralysis of voluntary muscles except (usually) vertical eye movements and blinking. Patients are fully aware but trapped in an unresponsive body.

Severe Cerebral Palsy: Some individuals with cerebral palsy have severe motor impairment affecting speech and limb control while maintaining normal cognition.

Brainstem Stroke: Strokes affecting the brainstem can damage motor pathways while leaving cognition intact.

For these populations, BCIs offer hope for communication that no other technology can provide.

P300 Speller

The P300 Speller, introduced by Farwell and Donchin in 1988, remains one of the most successful and widely used BCI paradigms.

How It Works:

The classic P300 Speller displays a 6×6 matrix of characters. Rows and columns flash in random sequence (typically 5-10 times each). The user focuses attention on the desired character. When the row or column containing that character flashes, it triggers a P300 response—a positive voltage deflection peaking around 300 milliseconds after the stimulus.

The P300 Response: The P300 is part of the brain's "oddball" response—an involuntary reaction to rare, task-relevant stimuli. Its amplitude increases when:

• The stimulus is infrequent (rare among other stimuli)
• The stimulus is relevant to the user's task
• The user is paying attention

In the speller context, the target character's flash is "rare" (it occurs in only 1/6 of row flashes and 1/6 of column flashes) and "relevant" (it's the character the user wants).

Character Detection: After all rows and columns have flashed (one sequence), a classifier determines which row and column produced the strongest P300 responses. Their intersection identifies the selected character. Multiple sequences (5-15) are typically averaged to improve accuracy.

Performance Metrics:

Typical P300 speller performance:

• Accuracy: 80-95% character accuracy
• Speed: 2-8 characters per minute
• Information Transfer Rate: 10-25 bits/minute

Factors affecting performance:

• Number of sequences averaged (more = slower but more accurate)
• Matrix size (larger matrix = more characters but slower selection)
• User attention and fatigue
• Signal quality and electrode placement

Advantages:

• Requires minimal training (P300 is involuntary)
• Works for most users including those with severe disabilities
• Robust, well-characterized signal
• Decades of research and development

Limitations:

• Relatively slow compared to natural speech or typing
• Requires sustained visual attention
• Dependent on intact vision (though auditory variants exist)
• Can be fatiguing over extended use

Variants:

Auditory P300 Speller: For users with visual impairment, spoken letters or environmental sounds can serve as stimuli. The user attends to their target sound, triggering an auditory P300.

Rapid Serial Visual Presentation (RSVP): Characters appear one at a time in the same location. The target character triggers a P300. Can be faster than matrix presentation.

Predictive Text: Combining P300 with word prediction can dramatically increase communication speed by reducing the number of characters needed.

SSVEP Speller

Steady-State Visual Evoked Potential (SSVEP) spellers offer an alternative paradigm with higher speed potential.

How It Works:

Multiple targets are displayed, each flickering at a distinct frequency (e.g., 8, 10, 12, 14 Hz). The user gazes at the desired target. The visual cortex entrains to the attended flicker frequency, producing oscillations at that frequency and its harmonics. Frequency analysis of the EEG reveals which target is being attended.

Detection Methods:

Power Spectral Analysis: Compute the power spectrum of occipital EEG; the target frequency with highest power indicates the user's selection.

Canonical Correlation Analysis (CCA): Compute correlation between EEG and reference signals at each target frequency. Higher correlation indicates stronger SSVEP response to that frequency.

Performance Metrics:

SSVEP spellers can achieve:

• Accuracy: 90-99%
• Speed: Up to 60+ characters per minute in optimized systems
• Information Transfer Rate: 40-60+ bits/minute

This substantially exceeds P300 speller performance.

Advantages:

• Higher speed than P300
• Very high accuracy achievable
• Minimal training required
• Robust signal detection

Limitations:

• Flickering stimuli can be uncomfortable or annoying
• Risk of photosensitive seizures (rare but serious)
• Requires precise frequency control of display
• Gaze-dependent (user must look at target)

Gaze-Independent SSVEP: Research has explored covert attention SSVEP, where users attend to peripheral targets without moving their eyes. This is harder but enables use by those who can't control eye movement.

Hybrid Communication BCIs

Combining multiple paradigms can address limitations of individual approaches:

P300 + SSVEP: Use SSVEP for rapid navigation/selection; P300 for confirmation or error correction. Combines speed of SSVEP with robustness of P300.

BCI + Eye Tracking: Use eye tracking for rapid gaze-based selection; BCI for confirmation or for when eye tracking fails. Particularly useful for ALS patients whose eye control gradually deteriorates.

BCI + Residual Movement: If users have any reliable movement (even small), use that for primary control with BCI as backup.

Clinical Communication BCI: Case Example

BrainGate Communication Trials:

In BrainGate clinical trials, participants with tetraplegia have used intracortical BCIs to control computer cursors for typing:

• Point-and-click typing using decoded motor intentions
• Speeds up to 6-8 words per minute achieved
• Some participants use BCI for daily email and communication
• Combination with predictive text increases speed

One participant, a woman with ALS, has used her BrainGate system for years for daily communication, demonstrating the long-term viability of implanted communication BCIs.

Motor BCIs

Motor BCIs decode intended movements and use them to control external devices—cursors, robotic arms, exoskeletons, or even the user's own paralyzed limbs through electrical stimulation.

Cursor Control

The simplest motor BCI application is controlling a computer cursor:

Non-Invasive (EEG-based):

• Motor imagery (left/right hand, feet) controls cursor direction
• 2D control achievable; 3D more challenging
• Performance: Task completion, but slower than mouse control
• Typical: 1-3 correct targets per minute in calibrated users

Invasive (Intracortical):

• Decode velocity or position directly from motor cortex neurons
• Near-natural cursor control achievable
• BrainGate participants achieve speeds approaching able-bodied mouse users
• Enables typing, email, web browsing

Robotic Arm Control

A landmark demonstration of motor BCIs was the control of robotic arms by paralyzed individuals.

The Hochberg 2012 Study: In this Nature paper, two participants with tetraplegia used BrainGate implants to control a robotic arm:

• 7 degrees of freedom (shoulder, elbow, wrist, hand)
• Continuous, naturalistic movements
• One participant brought a cup of coffee to her lips—first self-directed drink in 14 years
• Demonstrated feasibility of useful motor restoration

How It Works:

• Utah Array implanted in motor cortex
• Neural population activity decoded to movement velocity
• Kalman filter or neural network translates neural signals to arm commands
• Visual feedback allows user to correct movements

Current Capabilities:

• Reach, grasp, transport objects
• Some dexterous manipulation
• Still slower and less precise than natural movement
• Requires significant training

Exoskeletons

BCIs can control powered exoskeletons that assist or restore movement:

Lower Limb Exoskeletons:

• BCI triggers stepping or standing
• Typically simple on/off control rather than continuous
• Enables walking for some paraplegic individuals
• Research demonstrations; not yet widely clinical

Upper Limb Exoskeletons:

• Support arm weight while BCI controls direction
• Useful for rehabilitation and daily activities
• Can enable self-feeding, object manipulation

Walk Again Project: In a highly publicized demonstration at the 2014 World Cup opening ceremony, a paraplegic individual using an EEG-controlled exoskeleton took the ceremonial first kick. This demonstration sparked debate about the readiness of such technology but highlighted public interest in motor BCIs.

Functional Electrical Stimulation (FES)

Rather than controlling external devices, FES systems stimulate paralyzed muscles to produce movement.

How FES BCIs Work:

• BCI decodes intended movement
• Stimulation electrodes activate appropriate muscles
• User moves their own limb, not a robot

Advantages:

• Uses the user's own body—more natural feeling
• Provides proprioceptive feedback
• Potentially more cosmetically acceptable than robotic limbs

Challenges:

• Muscles must be intact (not atrophied)
• Precise stimulation patterns complex
• Fatigue limits duration of use
• Some spinal cord injuries affect muscles themselves

Clinical Examples: Systems have demonstrated:

• Hand grasp for quadriplegic individuals
• Arm reaching movements
• Combination with exoskeleton for support

One study showed a participant with tetraplegia feeding himself using his own arm controlled through a combined BCI-FES system.

Stroke Rehabilitation

While most motor BCIs aim to restore function through assistive devices, a different approach uses BCIs to promote recovery itself—harnessing neuroplasticity to rewire the damaged brain.

The Rehabilitation Hypothesis

After stroke, motor pathways are damaged but rarely completely destroyed. The brain has remarkable capacity for reorganization (neuroplasticity), with surviving neurons potentially taking over functions of lost ones.

Why BCIs Might Help: Traditional physical therapy requires some existing movement to practice. For severely impaired patients, this creates a paradox: they need to move to recover, but they can't move.

BCIs can detect the intention to move even when movement itself fails. This enables:

• Practice of motor intentions (activating motor cortex patterns)
• Contingent feedback (immediate response when correct pattern detected)
• Reinforcement of motor circuits even without actual movement

Hebbian Learning Principle: "Neurons that fire together, wire together." By pairing motor cortex activation (detected by BCI) with sensory feedback (visual, haptic, or from assisted movement), connections between motor intention and motor outcome may be strengthened.

BCI Rehabilitation Paradigms

Motor Imagery Training:

• Patient imagines moving affected limb
• BCI detects motor imagery
• Feedback provided (visual display, moving cursor, game)
• Repeated practice strengthens motor representations

BCI-Triggered Movement Assistance:

• Patient attempts movement
• When BCI detects intent, robotic device completes movement
• Provides sensory feedback of successful movement
• Associates intention with action

BCI-FES Rehabilitation:

• Patient attempts movement
• BCI triggers electrical stimulation of affected muscles
• Muscles contract, producing actual movement
• Proprioceptive feedback reinforces motor pathways

Clinical Evidence

Multiple randomized controlled trials have evaluated BCI rehabilitation:

Positive Findings:

• Improvements in motor function scores (e.g., Fugl-Meyer Assessment)
• Some evidence of cortical reorganization (increased motor cortex activation)
• Effects may persist after training ends
• Patient engagement high (often described as "motivating")

Meta-Analysis Results: Recent meta-analyses suggest moderate effect sizes for BCI rehabilitation, comparable to or slightly better than intensive conventional therapy.

Limitations:

• Optimal protocols not yet established
• High variability in patient response
• Time-intensive (many sessions required)
• Expensive equipment and setup
• Limited access (few centers offer BCI rehabilitation)

Practical Considerations

Patient Selection: BCI rehabilitation may work best for patients with:

• Moderate to severe impairment (too impaired for standard therapy)
• Ability to understand and follow instructions
• Sufficient attention for BCI tasks
• Adequate time since stroke for some spontaneous recovery

Treatment Protocols:

• Typical: 20-40 sessions over weeks to months
• Each session: 30-60 minutes of BCI training
• Often combined with conventional therapy

Seizure Prediction and Treatment

Epilepsy affects approximately 50 million people worldwide. For about 30%, medications fail to control seizures. Brain-computer interfaces offer new approaches to both predicting and treating seizures.

Seizure Detection

Why Detection Matters: Real-time detection of seizure onset can trigger interventions:

• Alert caregivers
• Warn patients to move to safety
• Trigger responsive stimulation

Detection Approaches: EEG (scalp or intracranial) shows characteristic changes during seizures:

• Increased amplitude
• Rhythmic patterns
• Spread across electrodes

Machine learning classifiers can detect these patterns:

• Sensitivity: 90-95%+ achievable
• Low latency: Detection within seconds of onset

Seizure Prediction

More ambitious than detection, prediction aims to identify a "pre-ictal" state minutes to hours before seizure onset.

Why Prediction is Hard:

• Pre-ictal changes are subtle
• Large variability across patients and seizures
• High false positive rates reduce practical utility
• Requires continuous long-term monitoring

Current Status:

• Better than chance prediction demonstrated
• Not reliable enough for clinical deployment as sole warning
• Research ongoing with improved algorithms and chronic monitoring

NeuroPace RNS System

The RNS (Responsive NeuroStimulation) system is an FDA-approved closed-loop neuromodulation device for drug-resistant epilepsy.

How It Works:

• Intracranial EEG electrodes implanted at seizure focus
• Neurostimulator implanted in skull
• Continuous monitoring for abnormal activity
• When seizure-like patterns detected, brief electrical stimulation delivered
• Stimulation disrupts seizure before it fully develops

Clinical Results:

• Median seizure reduction: 53% at 2 years, 75% at 9 years
• Some patients achieve seizure freedom
• Improved quality of life reported
• Very good safety profile

Limitations:

• Requires invasive surgery
• Expensive (~$30,000 for device plus surgical costs)
• Doesn't work for all patients
• Not a cure—ongoing therapy required

The RNS system represents the first FDA-approved closed-loop BCI, demonstrating that chronic neural recording and stimulation is feasible and beneficial for carefully selected patients.

Neurofeedback

Neurofeedback uses real-time displays of brain activity to enable voluntary self-regulation. While technically a BCI application, it differs from communication or motor BCIs in that the goal is not external control but internal brain state modification.

How Neurofeedback Works

Basic Protocol:

• Measure EEG (typically 1-4 channels)
• Extract features of interest (e.g., theta/beta ratio)
• Display feedback (visual, auditory) reflecting that feature
• User attempts to modify the feedback through mental strategies
• Over many sessions, self-regulation improves

Common Protocols:

SMR Training: Increase sensorimotor rhythm (12-15 Hz) over motor cortex. Originally developed for epilepsy; also used for attention enhancement.

Beta Enhancement: Increase beta (15-20 Hz) activity associated with focus and concentration. Used for ADHD.

Alpha Training: Increase alpha (8-12 Hz) for relaxation and stress reduction.

Alpha/Theta Training: Achieve deep relaxation with increased theta and reduced alpha. Used for anxiety, trauma, and peak performance.

Clinical Applications and Evidence

ADHD: Neurofeedback for ADHD has the most research support:

• Multiple randomized controlled trials
• Meta-analyses show moderate effect sizes
• Some effects persist at follow-up
• Probably effective, though debate continues about specificity

Epilepsy: SMR neurofeedback may reduce seizure frequency:

• Limited but promising evidence
• Not a first-line treatment
• May complement medication

Anxiety and Depression: Alpha asymmetry training and other protocols show promise:

• Preliminary positive results
• More rigorous trials needed
• Complementary to established treatments

Peak Performance: Athletes and performers use neurofeedback for:

• Attention and focus
• Anxiety management
• Optimal arousal states

Evidence is limited but popular in sports psychology.

Critical Evaluation

Neurofeedback remains controversial in mainstream medicine:

Supportive Evidence:

• Many studies show benefits
• Some show specificity over sham treatment
• Reasonable theoretical basis

Concerns:

• Many studies have methodological limitations
• Placebo effects may contribute
• Expensive and time-intensive
• Variability in practitioner training and protocols
• Commercial marketing sometimes exceeds evidence

Current Consensus: Probably useful for ADHD as an adjunct to established treatments. Other applications need more research before strong recommendations.

Disorders of Consciousness

BCIs can serve as diagnostic tools, not just therapeutic ones. In disorders of consciousness, BCIs may detect awareness in patients who appear unresponsive.

The Diagnostic Challenge

After severe brain injury, patients may appear unresponsive—not following commands, not speaking, not moving purposefully. But are they conscious?

Conditions:

• Coma: No signs of awareness; eyes closed
• Vegetative State (Unresponsive Wakefulness): Eyes may open; no purposeful behavior
• Minimally Conscious State: Inconsistent but reproducible signs of awareness

Distinguishing these conditions is difficult: some studies suggest 15-40% of patients diagnosed as vegetative actually show signs of awareness when tested more carefully.

BCI-Based Assessment

The Owen 2006 Study: A landmark study asked a patient in vegetative state to imagine playing tennis (motor imagery) or walking through her house (spatial navigation) while in an fMRI scanner. Her brain showed appropriate activation for each task—indistinguishable from healthy volunteers.

This demonstrated that apparently unresponsive patients may have rich inner lives undetectable by bedside examination.

EEG-Based Assessment: Similar paradigms using EEG:

• Patient instructed to perform motor imagery tasks
• EEG analyzed for task-appropriate patterns
• Detects covert awareness in some patients

Implications:

• May change diagnosis (from vegetative to minimally conscious)
• May influence treatment decisions
• Enables simple communication (imagine task A for "yes," task B for "no")
• Profound ethical and emotional implications for families

Communication for DOC Patients

For patients with confirmed covert awareness:

• BCI-based yes/no communication possible
• Simple P300 or motor imagery paradigms
• Enables basic interaction with family, medical team
• Research ongoing to improve reliability and speed

Ethical Considerations

Clinical BCIs raise profound ethical questions that researchers, clinicians, and society must address.

Informed Consent

Challenges:

• Patients may have communication or cognitive impairments
• Understanding complex BCI technology is difficult
• Long-term implications uncertain
• Surrogate consent may be needed

Approaches:

• Simplified explanations and demonstrations
• Staged consent as capacity allows
• Ongoing consent assessment
• Family involvement in decisions

Privacy and Brain Data

Unique Sensitivity: Neural data is uniquely personal. It could potentially reveal:

• Health conditions
• Emotional states
• Cognitive abilities
• Personal preferences

Concerns:

• Data storage and security
• Access by third parties (employers, insurers)
• Potential for inference beyond intended purpose
• "Mental privacy" as a new frontier

Emerging Protections:

• "Neurorights" legislation proposed or enacted in some jurisdictions
• Chile became first country to enshrine neurorights in constitution (2021)
• GDPR and health data regulations apply to neural data

Access and Equity

Current Situation:

• Invasive BCIs available only at specialized centers
• High costs (surgery, devices, ongoing support)
• Most development in wealthy countries

Justice Concerns:

• Who gets access to beneficial technology?
• Will BCIs increase or decrease disability disparities?
• How to ensure global access as technology matures?

Identity and Agency

Philosophical Questions:

• When a BCI controls an action, who is responsible?
• Do BCIs alter personal identity?
• What are the boundaries between person and machine?

Practical Implications:

• Legal responsibility for BCI-mediated actions
• Impact on sense of self and autonomy
• Integration of device into body image

Practical Exercises

Exercise 6.1: Paradigm Selection

For each patient, recommend an appropriate BCI paradigm:

1. A patient with ALS who has lost speech but retains eye movement
2. A patient with locked-in syndrome who cannot move eyes reliably
3. A stroke patient in rehabilitation who has minimal finger movement
4. A patient with epilepsy not controlled by medication

Discussion:

1. For ALS with eye movement: Gaze-based selection (eye tracking) with P300 or SSVEP for confirmation. Eye tracking is fastest; BCI provides backup as disease progresses. Consider hybrid for robustness.

1. For locked-in without eye control: Auditory P300 speller or motor imagery-based yes/no. Covert attention SSVEP possible but harder. May need extensive calibration to find reliable signals. Consider invasive BCI if non-invasive fails.

1. For stroke with minimal movement: BCI-FES rehabilitation combining motor imagery detection with stimulation of finger muscles. Goal is recovery, not permanent assistive device. Complement with conventional therapy.

1. For drug-resistant epilepsy: Evaluation for NeuroPace RNS if candidate for surgery. Responsive neurostimulation can reduce seizures by detecting and disrupting pre-ictal activity.

Exercise 6.2: Ethical Analysis

A company offers direct-to-consumer neurofeedback for "cognitive enhancement" in healthy individuals:

• No prescription required
• Claims of improved focus, memory, performance
• Limited clinical evidence
• Collects brain data for "product improvement"

Analyze the ethical issues involved.

Discussion:

Concerns:

• Overstated claims without strong evidence
• Potential for harm if users neglect proven interventions
• Data privacy issues with brain data collection
• Vulnerable populations (students, workers) may feel pressured
• Regulatory gaps for non-medical claims
• Potential to worsen inequity (enhancement for those who can afford it)

Considerations:

• Consumer autonomy in purchasing
• Innovation and access outside medical system
• Transparency about evidence and data use
• Need for appropriate regulatory framework

Exercise 6.3: Clinical Evaluation

A new study claims 80% success rate for a motor imagery BCI used in stroke rehabilitation. Critically evaluate:

• 20 patients
• No control group
• 6 weeks of training, 3 sessions per week
• Primary outcome: Fugl-Meyer Assessment

Discussion:

Limitations:

• Small sample size—results may not generalize
• No control group—cannot separate BCI effect from natural recovery, placebo, attention
• 6 weeks post-stroke typically shows spontaneous recovery
• Selection bias possible (motivated patients enrolled)
• Single outcome measure

What would strengthen:

• Larger sample
• Randomized control group (sham BCI or conventional therapy)
• Blinded outcome assessment
• Multiple outcome measures (function, neuroimaging)
• Longer follow-up

Summary and Key Takeaways

This module has explored clinical BCI applications and their impact on patients with neurological conditions.

Communication BCIs:

• P300 spellers enable letter-by-letter communication
• SSVEP offers higher speeds for those who can use it
• Hybrid approaches combine paradigm strengths
• Life-changing for locked-in and ALS patients

Motor BCIs:

• Cursor control enables computer access
• Robotic arms restore reach and grasp
• FES systems move the user's own limbs
• Continuous improvement but not yet matching natural function

Rehabilitation:

• BCIs may enhance neuroplasticity after stroke
• Detect intention even when movement fails
• Moderate evidence of benefit; more research needed

Seizure Management:

• Detection is reliable; prediction remains challenging
• NeuroPace RNS: First FDA-approved closed-loop BCI
• Responsive stimulation reduces seizures in refractory epilepsy

Neurofeedback:

• Self-regulation of brain activity through feedback
• Best evidence for ADHD
• Controversy about specificity and mechanisms

Ethics:

• Consent, privacy, equity, and agency require attention
• Neurorights emerging as new framework
• Responsible development essential

Quiz

1. Compare P300 and SSVEP spellers in terms of speed, accuracy, and user requirements.
2. How does BCI rehabilitation for stroke differ from assistive motor BCIs?
3. What is responsive neurostimulation and how does it treat epilepsy?
4. Describe how BCIs can detect covert awareness in disorders of consciousness.
5. List three ethical considerations for clinical BCI deployment.

Further Reading

• Wolpaw, J. R., & Wolpaw, E. W. (2012). Brain-Computer Interfaces: Principles and Practice. Oxford University Press.
• Hochberg, L. R., et al. (2012). "Reach and grasp by people with tetraplegia using a neurally controlled robotic arm." Nature, 485, 372-375.
• Owen, A. M., et al. (2006). "Detecting awareness in the vegetative state." Science, 313, 1402.
• Ienca, M., & Andorno, R. (2017). "Towards new human rights in the age of neuroscience and neurotechnology." Life Sciences, Society and Policy, 13(5).

Next Module

Continue to Module 7: Building BCI Systems →

In the next module, we'll shift from understanding BCIs to building them—covering development platforms, real-time architecture design, and hands-on implementation guidance.

---

This module is part of the Introduction to Brain-Computer Interfaces course, developed by Wael El Ghazzawi as part of the MIT Professional Education CTO Program.