1. Clinical Need
Diagnostic medical ultrasound is the most frequently performed imaging modality in the United States, with exam volumes growing from 38.6 million to 59.8 million between 2011 and 2021, a 55.1% increase (AIUM Workforce Study, 2023). This growth has far outpaced the supply of qualified sonographers: during the same period, accredited programs produced only 23% more graduates (4,386 to 5,393 annually), creating a structural workforce deficit that continues to widen.
The consequences of this deficit are measurable. Sonographer vacancy rates reached 16.7% in 2023 before improving to 12.4% in 2025, still well above historical norms (ASRT Staffing Survey, 2025). The Bureau of Labor Statistics projects 13% employment growth for diagnostic medical sonographers from 2024 to 2034, with approximately 5,800 openings per year. Workforce burnout compounds the supply problem: a survey of US and Canadian sonographers found that over 50% reported moderate to severe work-related burnout, driving retention challenges and early career exits.
Ultrasound imaging is uniquely operator dependent. Unlike CT or MRI, where the machine acquires a complete dataset regardless of operator skill, ultrasound image quality depends on the sonographer’s ability to maintain consistent probe pressure, orientation, and anatomical coverage. Inter-operator variability in ultrasound measurements is a well-documented problem: different sonographers scanning the same patient can produce measurements that differ by 20% or more (Zielke et al., 2022). This variability limits diagnostic reproducibility and complicates longitudinal monitoring.
The workforce constraint is most acute in rural and underserved settings. Approximately 62 million Americans live in Health Professional Shortage Areas where access to diagnostic imaging is limited (HRSA, 2024). A fully autonomous scanning system that does not require a trained sonographer would extend diagnostic ultrasound access to primary care clinics, urgent care centers, and rural hospitals that currently lack the personnel to offer the service.
The economic scale of the problem is substantial. The US diagnostic imaging center market alone was valued at $26 billion in 2024 (IBISWorld). Medicare reimburses diagnostic ultrasound examinations through established CPT codes: 76536 for thyroid ultrasound (approximately $120), 76700 for complete abdominal ultrasound (approximately $215), and 93306 for echocardiography with Doppler (approximately $300). A system that maintains or improves diagnostic quality while removing the operator dependency bottleneck addresses a clinical need that grows every year the workforce gap persists.
2. State of the Art
The field of robotic ultrasound has progressed through three generations. First-generation systems (2015 onward) used teleoperation, where a remote physician controls a robotic arm holding an ultrasound probe via joystick or haptic interface. Second-generation systems (2020 onward) added semi-autonomous capabilities, where the robot handles probe pressure and basic positioning while a human supervises and intervenes for complex anatomy.
Third-generation systems, which represent the current research frontier, aim for full autonomy: the robot plans the scan path, controls probe contact force, optimizes image quality, and interprets results without human intervention. This is where reinforcement learning enters.
Autonomous scan path planning
Su et al. (2024, Nature Communications) demonstrated a fully autonomous robotic ultrasound system (FARUS) for thyroid scanning that combines skeleton point recognition, reinforcement learning for target localization, and Bayesian optimization for dynamic probe orientation adjustment. The system was tested on 19 adult patients (mean age 53.05 ± 5.90 years) and produced scans comparable to those of experienced clinicians, with the ability to detect thyroid nodules and compute ACR TI-RADS classifications.
Reproducible force control and measurement accuracy
Lin et al. (2025, Frontiers in Robotics and AI) demonstrated an autonomous scanning system achieving thyroid volume measurements of 34.3 ± 0.3 mL against a CT gold standard of 34.48 mL (p = 0.285, no statistically significant difference). Expert physicians, by contrast, measured 31.0 to 33.0 mL (all p < 0.05, significantly different from ground truth). The autonomous system maintained force control with a coefficient of variation (COV) of 0.09 to 0.11, compared to expert physician COV of 0.21 to 0.50. AI classification consistency for a benign lesion showed autonomous COV of 0.29 versus expert COV of 0.97 to 2.01.
Intercostal and deep organ navigation
Bi et al. (2026, Scientific Reports) addressed one of the hardest problems in autonomous ultrasound: imaging organs that require scanning between ribs (intercostal windows), where acoustic shadows from bone create diagnostic blind spots. The RL framework uses CT template data to create 3D state representations and trains an agent to plan scanning trajectories that avoid rib shadows while maintaining target coverage. This extends RL-guided ultrasound from superficial structures (thyroid) to deep organs (liver, heart, kidneys) that are clinically more complex.
The gap between what exists in research laboratories and what is commercially available defines the opportunity. The research has demonstrated that autonomous RL-guided ultrasound can match or exceed human performance on specific organs. What does not exist is a commercially available system that integrates these capabilities into a clinical-grade device cleared for diagnostic use.
3. Foundational Research
Su K, Liu J, Ren X, Huo Y, Du G, Zhao W, Wang X, Liang B, Li D, Liu PX. (2024). “A fully autonomous robotic ultrasound system for thyroid scanning.” Nature Communications, 15, 4004. DOI: 10.1038/s41467-024-48421-y. PMID: 38734697.
System: UR3 6-DOF manipulator, linear ultrasound probe, Kinect depth camera for patient localization, 6-axis force/torque sensor for contact control. The RL component handles localization and approach planning: given depth camera input, the agent determines the optimal probe trajectory to the thyroid region. Bayesian optimization dynamically adjusts probe tilt and rotation during scanning to maximize image quality. Clinical testing on 19 patients (age 53.05 ± 5.90 years) demonstrated autonomous location of the thyroid, maintenance of appropriate contact pressure, and acquisition of diagnostic-quality images comparable to manual scans by experienced sonographers. The system also computed ACR TI-RADS scores for detected nodules, demonstrating end-to-end diagnostic capability from approach planning through clinical classification.
Lin X-X, Li M-D, Ruan S-M, Ke W-P, Chen L-D, Huang Q-H, Wang W et al. (2025). “Autonomous robotic ultrasound scanning system: a key to enhancing image analysis reproducibility and observer consistency in ultrasound imaging.” Frontiers in Robotics and AI, 12, 1527686. PMCID: PMC11835693.
System: Franka Emika Panda 7-DOF manipulator, 1 kHz control loop, Robotiq FT 300-S 6-axis force sensor, SonoHealth D5CL 7.5 MHz wireless probe. Compared against 4 expert physicians (>5 years experience) and 4 non-expert physicians (<3 years). Force control: mean contact force of 1.9 ± 0.2 N (phantom) and 2.1 ± 0.5 N (human thyroid) with COV of 0.09 to 0.11, compared to expert physician COV of 0.21 to 0.50 and non-expert COV of 0.20 to 0.48. Consistent force across all tested scanning speeds (1.0 to 11.0 mm/s). Thyroid volume accuracy: 34.3 ± 0.3 mL against 34.48 mL CT gold standard (0.5% error, p = 0.285), while expert measurements ranged from 31.0 to 33.0 mL (4.3% to 10.1% error, all p < 0.05). Reproducible radiomics features: 75.73% versus expert 73.43% versus non-expert 70.70%. Classification COV: 0.29 versus expert 0.97 to 2.01.
Bi Y, Qian C, Zhang Z, Navab N, Jiang Z. (2026). “Autonomous path planning for intercostal robotic ultrasound imaging using reinforcement learning.” Scientific Reports (Nature). DOI: 10.1038/s41598-026-37702-9.
RL framework for intercostal path planning using CT template data and 3D state representations. The agent plans scanning trajectories that avoid rib acoustic shadows while maintaining target organ coverage. Validated on previously unseen patient models with randomly placed scanning targets, demonstrating generalization to anatomically constrained regions where manual scanning consistency is lowest. This work directly validates the feasibility of RL-guided scanning for abdominal and cardiac targets that require intercostal acoustic windows.
Ning G et al. (2024). “Inverse-reinforcement-learning-based robotic ultrasound active compliance control in uncertain environments.” IEEE Transactions on Industrial Electronics, 71, 1686–1696. DOI: 10.1109/tie.2023.3250767.
Inverse RL applied to learn compliance control policies from expert demonstrations, enabling real-time adaptation of force and posture when scanning curved or irregular body surfaces. Validated for vascular ultrasound, where probe contact must be maintained along non-planar vessel paths without occluding the target vessel. This methodology addresses the requirement for safe, consistent contact on anatomically variable tissue surfaces.
Li et al. (2023). “RL-TEE: Autonomous probe guidance for transesophageal echocardiography based on attention-augmented deep reinforcement learning.” IEEE Transactions on Automation Science and Engineering, 21(2), 1526–1538.
Attention-augmented deep RL agent for transesophageal echocardiography (TEE) probe positioning. TEE is one of the most operator-dependent ultrasound modalities, requiring navigation of a probe through the esophagus to image the heart from posterior windows. This demonstrates that RL-based autonomous guidance is feasible even for the most technically demanding ultrasound applications, where TEE-qualified sonographers are scarce in intensive care and cardiac surgery settings.
4. Competitive Landscape
Cobionix (Waterloo, Canada). Raised $3 million in July 2025. CODI platform is designed for teleoperation: a physician in a centralized location controls the robotic arm at a remote clinic. The product is not autonomous; it still requires a skilled operator for every scan.
iSono Health (San Francisco, CA). FDA-cleared ATUSA is a wearable automated breast ultrasound system. It is a pad-style wearable, not a robotic arm, and performs only breast imaging using a predetermined scanning pattern. Single anatomical target, fixed protocol.
Life Science Robotics (Denmark). ARUS is a robotic arm with joystick control, haptic feedback, and adaptive pressure management. Explicitly designed to be human-operated, reducing musculoskeletal strain on sonographers rather than replacing the need for a sonographer.
No company currently sells or has in clinical trials a fully autonomous RL-guided diagnostic ultrasound system. The competitive landscape consists of teleoperated systems (Cobionix, Life Science Robotics) and a single-organ automated wearable (iSono Health). The space where RL algorithms autonomously plan scan paths, control probe contact, and acquire diagnostic-quality images across multiple anatomical targets is entirely unoccupied.
5. Addressable Scope
Bottom-up calculation (US diagnostic ultrasound)
- US exam volume reached 59.8 million annually as of 2021 (AIUM Workforce Study, 2023); at 3% annual growth, approximately 69.3 million exams per year in 2026.
- Autonomous systems would enter through standardized screening and surveillance: thyroid surveillance (estimated 4.2 million exams/year), abdominal screening (estimated 12.1 million exams/year), obstetric biometry (estimated 8.3 million exams/year). Total: approximately 24.6 million standardized exams per year.
- 15% capture within 5 years of FDA clearance: 3.69 million exams × $180 average reimbursement = $664 million in annual procedural revenue.
- Device pricing at $150,000 per system (comparable to high-end ultrasound carts plus robotic arm), installed base of 5,000 to 10,000 US systems: US device TAM of $750 million to $1.5 billion.
Top-down cross-check
The global robotic ultrasound systems market was valued at $1.52 billion in 2024 and is projected to reach $3.85 billion by 2032 at 16.6% CAGR (Persistence Market Research, 2025). An autonomous RL-guided subsegment capturing 20% to 30% of the broader robotic ultrasound market by 2032 yields $770 million to $1.16 billion, consistent with the bottom-up estimate.
Public benefit framing
The initial serviceable population is US healthcare facilities with existing ultrasound infrastructure but sonographer shortages: over 2,100 critical access hospitals, 11,000+ urgent care centers, and primary care practices piloting point-of-care ultrasound. These settings represent approximately 15,000 to 20,000 potential installation sites in the first 5 years, concentrated in the Health Professional Shortage Areas where 62 million Americans currently lack access to diagnostic imaging.
6. Research Gaps and Opportunity
The published research demonstrates three validated capabilities that have not been integrated into a single system:
Gap 1: Multi-organ generalization
Autonomous scan path planning has been demonstrated for thyroid (Su et al., 2024), vascular structures (Ning et al., 2024), intercostal organs (Bi et al., 2026), and transesophageal cardiac imaging (Li et al., 2023). Each study solves path planning for one anatomical region. No system generalizes across multiple organs using a unified RL framework. The research question is whether a single RL architecture, trained on multi-organ anatomical models derived from CT data, can adapt its scanning protocol based on clinical indication, analogous to how a sonographer switches between thyroid, abdominal, and cardiac protocols.
Gap 2: Integrated force control and adaptive planning
Force-compliant contact control has been validated with sub-Newton consistency (Lin et al., 2025: COV 0.09 to 0.11 versus physician COV 0.20 to 0.50). This achievement has not been combined with real-time RL scan planning in a system that adapts force control to patient-specific body habitus, tissue compliance, and positional changes during scanning. The research question is how to co-optimize force compliance and scan coverage within a single policy network.
Gap 3: Lab-to-production hardware translation
Academic prototypes use general-purpose research manipulators (UR3, Franka Emika Panda) costing $30,000 to $80,000, weighing 10 to 18 kg, and occupying the footprint of a desktop workstation. A clinical system requires a purpose-built robotic arm optimized for weight, reach envelope, cable routing, and sterilization, designed for manufacturing at volume, and integrated into a system architecture that meets IEC 62304 software lifecycle and IEC 60601 electrical safety standards. The originating research laboratories lack the manufacturing engineering expertise to close this gap. This is the translation barrier where most funded research stalls.
Research thesis: The group that integrates multi-organ RL-guided scanning, compliant force control, and production-ready hardware into a single FDA-cleared platform establishes the foundation for autonomous diagnostic ultrasound. The window is defined by the regulatory timeline: whoever receives De Novo authorization first defines the predicate device for all subsequent entrants.
7. Comparable Funded Projects
| Source | PI / Entity | Amount | Focus |
|---|---|---|---|
| NSF SBIR Phase I | Award #2212911 | ~$275K | AI-enabled ultrasound for musculoskeletal imaging and diagnosis |
| NIBIB Trailblazer R21 | Ongoing program | $400K / 3yr | Early-stage investigators in biomedical imaging; active portfolios in “Robotics” and “Ultrasound: Diagnostic and Interventional” |
| NSF NRI-3.0 | Active solicitation | $500K–$1.5M / 3–4yr | Collaborative and autonomous medical robotics, human-robot interaction |
| ARPA-H Open BAA | $2.5B initial appropriation | Variable | Breakthrough health technologies for underserved populations; Health Science Futures focus area |
Government agencies have invested substantially in adjacent domains (AI-assisted imaging, teleoperated ultrasound, medical robotics), establishing both technical validation and funder familiarity with the component technologies. The convergence of these components into fully autonomous diagnostic scanning represents the next step in a funded research trajectory.
8. Opportunity Assessment
TRL evidence chain
TRL 3 (analytical and experimental proof of concept): RL-based probe guidance demonstrated in simulation and phantom models across multiple anatomical targets (Jiang et al., 2022; Li et al., 2023; Ning et al., 2024).
TRL 4 (validation in relevant environment): FARUS tested on 19 human patients with diagnostic-quality thyroid scans (Su et al., 2024). auto-RUSS validated on human volunteers with quantitative comparison to expert physicians, achieving 0.5% measurement error versus 4.3% to 10.1% error from experts (Lin et al., 2025).
Top 3 research questions
Generalization across body habitus and anatomical variation
Patients vary in body mass index, tissue composition, and anatomical variants (e.g., ectopic thyroid, horseshoe kidney). Current RL agents are trained on limited patient populations.
Approach: Domain randomization during simulation training (varying body parameters, tissue properties, anatomical positions) combined with sim-to-real transfer, validated by Jiang et al. (2022) for vascular navigation.
ModerateSafety during autonomous patient contact
The probe must maintain safe contact force (typically 2 to 5 N) on sensitive areas (carotid, neonatal fontanelle) without excessive pressure.
Approach: Hardware force limits (mechanical compliance plus software force ceiling), validated by Lin et al. (2025), where the autonomous system maintained more consistent and lower force than human operators (COV 0.09 to 0.11 versus physician COV 0.21 to 0.50).
ModerateSim-to-real transfer fidelity for multi-organ scanning
RL agents trained in simulation must perform equivalently on live patients across diverse anatomical targets. Transfer degradation is the primary technical uncertainty.
Approach: CT-derived anatomical phantoms for progressive validation (simulation, physical phantom, healthy volunteer). Bi et al. (2026) demonstrated successful transfer to unseen patient models for intercostal scanning.
HighRegulatory pathway
Classification: FDA De Novo (Class II, no directly predicate autonomous diagnostic ultrasound system). The IDx-DR autonomous retinopathy system (De Novo DEN180001, April 2018) serves as regulatory precedent for an autonomous diagnostic imaging device that provides clinical decisions without physician oversight.
Algorithm strategy: Locked algorithm approach for initial regulatory submission. The scanning algorithm is fixed after training (not adaptive on-device), qualifying as a Software as a Medical Device (SaMD) with a predetermined algorithm under the FDA’s AI/ML framework. Post-market adaptive updates governed by a Predetermined Change Control Plan (PCCP) per the FDA’s 2023 guidance on AI/ML-based SaMD.
Regulatory moat: De Novo classification creates a 2 to 3 year barrier to entry. The first group to receive De Novo authorization establishes the predicate device for the product category, requiring all subsequent entrants to demonstrate substantial equivalence.
Estimated timeline: 18 to 24 months for De Novo preparation and submission, 12 to 18 months for FDA review. Total: 30 to 42 months to market authorization.
9. Team Capabilities
Successful pursuit of this research direction requires three intersecting capabilities: biomedical domain expertise for clinical problem framing and experimental design, machine learning engineering for the autonomous scanning agent, and manufacturing engineering for the lab-to-production hardware bridge. HHA’s team provides coverage across all three.
Hass Dhia
MS Biomedical Sciences (Wayne State University School of Medicine), with anatomy teaching assistant background. Provides: clinical problem identification (which anatomical targets the RL agent must locate), anatomical and physiological domain knowledge (ultrasound physics, acoustic windows, tissue characteristics), experimental design (clinical validation protocols, IRB-ready study design), and AI system architecture for autonomous diagnostics. Translates clinical requirements into engineering specifications: what constitutes a diagnostically adequate image, how the system’s output integrates with electronic health records, and how imaging protocols map to standard-of-care examination requirements.
Haedar Hadi
MS Computer Science (Boston University, Information Systems focus), with cloud and database architecture expertise. Provides: RL algorithm design (TD3, SAC, PPO for continuous control probe navigation), evaluation methodology (benchmark tasks for regulatory submissions), sim-to-real transfer frameworks, and scalable training infrastructure. The evaluation framework is as critical as the algorithm: defining benchmark tasks, constructing reproducible test scenarios, and establishing quantitative pass/fail criteria for regulatory submissions. Leads technical infrastructure, navigation controller development, and the software engineering required for IEC 62304 compliance.
Ahmed
Director of Manufacturing with deep expertise in design for manufacturability (DFM), production scaling, and quality systems. Provides: purpose-built robotic arm DFM (replacing $30,000 to $80,000 research manipulators with $5,000 to $15,000 clinical-grade arms), probe fixture and force sensor integration, cable management for ultrasound and sensor connections, medical-grade sterilization compatibility, and assembly process design that scales to hundreds and thousands of units per year.
Most research proposals end at “it works in the lab.” This proposal includes explicit DFM milestones at every phase, ensuring that prototype decisions consider production scaling, tolerance analysis, and quality systems from day one. This addresses the valley of death between TRL 4–5 prototypes and TRL 7+ deployable systems, the gap where most funded research stalls. Ahmed’s manufacturing engineering capability directly addresses Gap 3 (lab-to-production hardware translation), the single most common failure mode in translational medical robotics research.
10. Recommended Next Steps
Target funding programs
| Program | Mechanism | Range | Fit |
|---|---|---|---|
| NIH NIBIB Trailblazer R21 | R21 | $400K / 3yr | New/early investigators in biomedical imaging; autonomous robotic ultrasound falls within active “Robotics” and “Ultrasound: Diagnostic and Interventional” portfolios |
| NSF NRI-3.0 | Standard grant | $500K–$1.5M / 3–4yr | Collaborative and autonomous medical robotics; human-robot interaction in clinical settings |
| NSF SBIR Phase I | SBIR | $275K | AI-enabled medical devices; established precedent via Award #2212911 for AI-augmented ultrasound |
| ARPA-H Open BAA | Performer agreement | Variable | Breakthrough health technology for underserved populations; autonomous diagnostics for Health Professional Shortage Areas |
| NIH NIBIB R01 | R01 | $500K/yr / 5yr | Established programs in biomedical imaging and bioengineering; full multi-organ validation and clinical trials |
Estimated total funding range: $275,000 (SBIR Phase I) to $2.5 million (NRI-3.0 combined with R01) over 24 to 36 months for Phase 1 (multi-organ RL validation, custom arm prototype, phantom and healthy volunteer studies).
24-month milestone timeline
- M1–3 Simulation environment construction: CT-derived anatomical models for thyroid, abdominal, and carotid targets. RL training infrastructure setup. Multi-organ reward function design.
- M1–6 Ahmed (DFM): DFM analysis of research-grade components (UR3, Franka Emika Panda). Custom arm concept design targeting $5,000 to $15,000 unit cost. BOM optimization and supplier identification.
- M4–9 RL agent training for multi-organ scanning (thyroid, abdominal, carotid). Domain randomization across body habitus parameters. Sim-to-real transfer methodology development.
- M6–12 Ahmed (Prototype): First prototype arm build. Force sensor integration. Sterilization-compatible housing design. Cable management for ultrasound and sensor connections.
- M10–15 Phantom validation across 3 anatomical targets. Quantitative comparison to manual scanning (force consistency, measurement accuracy, coverage completeness). Publication of multi-organ RL results.
- M12–18 IRB submission for healthy volunteer study. FDA Pre-Sub meeting to confirm De Novo classification strategy and clinical evidence requirements.
- M15–20 Healthy volunteer clinical validation (n = 30 to 50). Quantitative comparison of autonomous scanning to expert sonographers across thyroid, abdominal, and carotid protocols.
- M18–24 Ahmed (Production intent): Second-generation arm with production-intent design. Cost target validation ($5,000 to $15,000). Tolerance analysis and assembly process documentation.
- M20–24 FDA De Novo preparation. Clinical data package compilation. IEC 62304 software lifecycle documentation. Phase 2 funding application for expanded clinical trials and manufacturing qualification.