A portable brain scanner detected THC impairment more accurately than field sobriety tests | Research

The takeaway

A machine learning model using portable brain imaging (fNIRS) identified THC-impaired individuals with 76.4% accuracy and only a 10% false-positive rate, outperforming standard field sobriety evaluations.

Law enforcement, driving safety researchers, and policymakers seeking cannabis impairment detection methods.

76.4% accuracy with 10% false positive rate vs 35.4% for field sobriety

What the researchers found

In 169 cannabis users given oral THC or placebo in a crossover design, prefrontal cortex oxygenated hemoglobin increased after THC only in participants classified as impaired. ML models using fNIRS data achieved 76.4% accuracy and 69.8% positive predictive value with a 10% false-positive rate, compared to field sobriety exams at 67.8% accuracy, 35.4% PPV, and 35.4% false-positive rate.

Why it matters

There is currently no evidence-based method to detect cannabis-impaired driving. Blood THC levels do not reliably predict impairment. A portable brain-based measure could fill this critical gap.

The numbers in context

169 participants. fNIRS ML model: 76.4% accuracy, 69.8% PPV, 10% false positive. Field sobriety: 67.8% accuracy, 35.4% PPV, 35.4% false positive.

How the study worked

Double-blind, randomized, crossover study with 169 cannabis users aged 18-55. fNIRS measured prefrontal cortex activation before and after oral THC and placebo. Impairment defined by convergent clinical ratings and an algorithm based on heart rate and self-rated "high." Machine learning models compared to drug recognition evaluator field sobriety exams.

What this study cannot tell us

Impairment was operationalized using clinical ratings and physiological markers, not actual driving performance. Oral THC has different pharmacokinetics than inhaled. Specificity to THC versus other impairment sources not yet determined.

How to read the evidence

Well-designed double-blind crossover trial with large sample, though impairment definition was proxy-based, not driving-performance based.

When this study was published

Published in 2022.

The bigger picture

Unlike blood or urine tests that only detect THC presence, brain imaging captures actual functional impairment, making it more relevant for safety-critical decisions like driving.

Questions still open

Would this approach work roadside in real-world conditions? Is the neural signature specific to THC or shared with other forms of impairment?

Common questions

How does fNIRS detect impairment?

fNIRS measures blood oxygenation changes in the prefrontal cortex using light. After THC, impaired individuals showed distinct activation patterns and connectivity changes that machine learning models could identify.

Is this ready for roadside use?

Not yet. The technology is portable and the results are promising, but further work is needed to confirm specificity to THC impairment and validate the approach in real-world roadside conditions.

Read the original research

Identification of ∆9-tetrahydrocannabinol (THC) impairment using functional brain imaging.

Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology, 47(4), 944-952

View on PubMed ↗Publisher’s paper ↗

Citation

Gilman, Jodi M; Schmitt, William A; Potter, Kevin; Kendzior, Brian; Pachas, Gladys N; Hickey, Sarah; Makary, Meena; Huestis, Marilyn A; Evins, A Eden. (2022). Identification of ∆9-tetrahydrocannabinol (THC) impairment using functional brain imaging.. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology, 47(4), 944-952. https://doi.org/10.1038/s41386-021-01259-0