This AI company wants to replace MRIs with a 60-second dip in the spa. Can that really work?
The pitch is irresistible: instead of a loud, claustrophobic, 30–60 minute MRI inside a multi-million-dollar magnet, you take a one-minute soak in a warm, spa-like bath while an AI system maps your insides. Lower cost, faster throughput, no radiation, happier patients. It sounds like the kind of disruption healthcare could use. It also raises a hard question: can any water-based, AI-accelerated imaging system truly replace MRI across the breadth of clinical use?
Short answer: not broadly, not soon. Longer answer: in specific body regions and indications, a water-coupled imaging system—likely based on ultrasound, photoacoustics, or electrical/magnetic impedance—could plausibly complement or even replace MRI for targeted tasks, if it clears significant technical, clinical, and regulatory hurdles.
What MRI does that’s hard to beat
– Soft-tissue contrast: MRI excels at differentiating water, fat, edema, fibrosis, blood products, and more through dozens of pulse sequences and parameters (T1, T2, diffusion, perfusion, spectroscopy).
– Depth and versatility: From brain and spinal cord to joints, liver, prostate, and heart, MRI images deep anatomy through bone and air-filled structures where ultrasound struggles.
– Resolution and quantitation: Isotropic sub-millimeter resolution is routine; quantitative biomarkers (e.g., T1/T2 mapping, diffusion metrics, extracellular volume) are increasingly standardized.
– Safety profile: No ionizing radiation, and most scans require no contrast agents.
Any challenger has to match or beat enough of these attributes for specific use cases, or offer a dramatically better cost, speed, or access profile that justifies trade-offs.
What a “60-second spa dip” might actually be
Water immersion isn’t magic, but it hints at certain physics:
– Ultrasound tomography: Arrays of transducers transmit and receive sound around a body part immersed in water (excellent acoustic coupling). By modeling how sound speeds and attenuates through tissue, a tomographic image can be reconstructed. Strengths: speed, safety, cost. Limitations: bone and air block/reflect ultrasound; image quality drops with depth; full torso imaging is hard.
– Photoacoustic (optoacoustic) tomography: Rapid laser pulses heat absorbers like hemoglobin by tiny amounts, launching ultrasound waves that are detected by transducers in water. This can map microvasculature and oxygenation several centimeters deep, with far richer contrast than ultrasound alone. Strengths: vascular and functional contrast; fast frame rates. Limitations: light penetration in tissue restricts depth; safety/eye protection needed; torso-scale imaging is challenging.
– Electrical or magnetic impedance tomography: Patterns of currents or magnetic fields are applied externally and tissue conductivities are inferred. Water can simplify uniform contact. Strengths: very fast, safe, cheap. Limitations: low spatial resolution (centimeters), prone to artifacts; best for monitoring (e.g., lung ventilation), not diagnostics that need fine detail.
All three become more capable with AI that fills in missing data, denoises signals, and reconstructs images from sparse measurements. But the underlying physics still limits what can be seen.
Could this replace MRI? Break it down by body region and clinical job
– Brain and spine: MRI dominates because it sees through skull and vertebrae with exquisite soft-tissue contrast and quantitative diffusion. Ultrasound-based methods can’t traverse bone effectively; photoacoustics and impedance methods can’t provide diagnostic-grade brain/spine detail. Replacement: no.
– Abdomen and pelvis: MRI is preferred for liver lesions, bile ducts, pancreas, pelvis, bowel fistulas, and more. Ultrasound struggles with bowel gas; photoacoustics is depth-limited. Some narrow wins are possible (e.g., liver fat/fibrosis staging with ultrasound elastography already replaces many MRIs), but wholesale replacement is unlikely. Replacement: selective.
– Breast: A genuine battleground. Ultrasound tomography and photoacoustic imaging have shown promise for lesion characterization and vascular biomarkers without compression or radiation. Several systems aim to complement or reduce MRI use in certain patients. Replacement: plausible in defined subgroups if large trials show non-inferiority.
– Musculoskeletal (joints, soft tissue): High-resolution ultrasound already diagnoses many tendon and muscle injuries and guides procedures. Deep cartilage, internal derangements, marrow edema, and occult fractures still favor MRI. Replacement: partial, indication-specific.
– Cardiovascular: Echocardiography is first-line; cardiac MRI remains gold-standard for tissue characterization (scar, myocarditis, infiltration). Water-bath systems wouldn’t change ultrasound’s established role, and replacing cardiac MRI’s quantitative tissue mapping is unlikely. Replacement: no, but ultrasound remains strong.
– Vascular and perfusion: Photoacoustics can map microvasculature and oxygenation for superficial tissues (e.g., skin, breast, peripheral vasculature). Potential to replace niche MR perfusion studies in superficial targets. Replacement: niche.
– Whole-body screening: MRI-based whole-body scans detect marrow lesions, sarcomas, and metastases. No water-based modality can match that depth and contrast today. Replacement: no.
Can AI close the physics gap?
AI already accelerates MRI and CT, improves ultrasound image quality, and reconstructs images from fewer measurements. But replacing information that doesn’t exist—like high-fidelity brain anatomy through bone—with inference is risky. Regulators increasingly require that AI reconstructions be faithful, generalize across populations and devices, and avoid “hallucinated” structures. Expect AI to:
– Make ultrasound/photoacoustic images cleaner, faster, and more quantitative.
– Improve tomographic reconstructions from sparse measurements.
– Assist triage and decision support (e.g., flagging concerning patterns).
But AI won’t negate the fundamental limits of light and sound penetration or the need for ground-truth validation against MRI, histopathology, or outcomes.
The 60-second claim: what is actually feasible?
– Acquisition time: With dense transducer arrays and parallel capture, a cross-sectional scan can be milliseconds; covering an organ can be seconds. A whole limb or breast in under a minute is plausible. A full torso at diagnostic resolution in 60 seconds is far harder due to geometry, attenuation, motion, and data volume.
– Motion management: Breathing and cardiac motion degrade images. MRI solves this with gating; water-based systems would need similar strategies or accept lower resolution.
– Throughput and setup: “One minute” often excludes preparation—undressing, positioning, sanitizing the bath, and drying—easily adding 10–15 minutes. Clinical throughput matters more than raw acquisition time.
Safety, infection control, and workflow
– Water hygiene: Warm water is a microbial playground. Medical-grade circulation, filtration, and disinfection protocols are essential. This adds cost and complexity.
– Electrical and laser safety: Photoacoustics entails high-energy light; shielding, eyewear, and interlocks are required. Electrical safety in wet environments is non-negotiable.
– Accessibility: Not all patients can enter and exit a bath (mobility limits, wounds, catheters). A “spa” may be less inclusive than a table-based scanner.
Regulatory and evidence bar
Replacing MRI isn’t a marketing word; it’s a regulatory claim. To displace MRI for a given indication, a system must demonstrate:
– Clinical performance: Non-inferiority or superiority to MRI (or standard of care) on diagnostic accuracy, sensitivity/specificity, or patient outcomes.
– Generalizability: Multicenter studies across patient sizes, skin tones, comorbidities, and devices.
– Reliability: Robust performance across operators and environments, with quality control.
– Safety: Proven exposure limits, biocompatible materials, and infection control.
Most likely pathway: claim “adjunctive” or “screening/triage” use, then expand indications as evidence accumulates.
Economics: where a “spa dip” could win
– Cost per scan: Ultrasound/photoacoustic hardware can be far cheaper than MRI, with lower room build-out requirements and power needs.
– Access: Faster, quieter, less intimidating, potentially deployable in ambulatory settings or rural clinics.
– Throughput: If setup/hygiene workflow is streamlined, one clinician could run many scans per day.
Even if it never replaces MRI, a water-coupled modality that reliably answers 60–70% of common soft-tissue questions at a fraction of the cost could meaningfully relieve MRI backlogs.
A realistic roadmap
– Start with focused indications where physics aligns:
– Breast lesion characterization and therapy monitoring.
– Peripheral vascular assessment, diabetic foot, wound perfusion.
– Musculoskeletal soft tissue and superficial oncology staging.
– Prove value as an adjunct:
– Show that negative scans safely avoid MRI in low-risk pathways.
– Use AI to flag “MRI-needed” cases with high sensitivity, keeping false negatives low.
– Build quantitative biomarkers:
– Standardize measurements (e.g., speed-of-sound maps, oxygen saturation) that correlate with histology or MRI markers.
– Iterate hardware:
– Dense, conformal transducer arrays; better light delivery; motion compensation.
– Scale evidence and reimbursement:
– Multicenter trials, cost-effectiveness analyses, payer coverage decisions.
The bottom line
– Replace MRI outright? No—MRI’s deep, versatile soft-tissue imaging isn’t broadly replicable with water-coupled modalities.
– Replace some MRIs? Yes—targeted, well-chosen indications where ultrasound or photoacoustic physics shine could shift substantial volume if high-quality evidence supports it.
– A 60-second spa dip as the new default? Unlikely. But as a fast, comfortable, lower-cost first-line exam that rules in or out common problems and reserves MRI for the harder questions, it’s a credible and potentially important addition to the imaging toolbox.
Questions to ask any startup making this claim
– Which specific indications are you targeting first, and what’s your reference standard?
– What is your depth and resolution at those targets, across BMI ranges?
– How do you handle motion, bone, and air interfaces?
– What are your multicenter clinical results versus MRI (sensitivity, specificity, NPV)?
– What’s the full room-time per patient, including hygiene and setup?
– How is the AI trained, validated, and monitored for drift and bias?
– What regulatory pathway and reimbursement codes are you pursuing?
If the answers are strong, a one-minute dip won’t replace MRI—but it could change when and why we need one.
