Detection, Inc. — Autonomous Medical Robotics

Physical medicine, self-directed.
No operator in the loop.

MedsApp.ai trains robotic systems on archived procedure data, then executes real invasive and non-invasive medical, dental, cosmetic, dermatology, ophthalmology, and veterinary procedures — comparing every step against that training in real time, without a clinician guiding the instrument.

MARKED-POSITION TRIANGULATION — CARTESIAN REFERENCE FRAME TRACKING
Z X M1 x14.221 y08.903 z02.114 M2 x41.008 y19.442 z02.098 INSTRUMENT POSE validated · Δ0.004mm
Method
Marked Positions
Reference
Photodetector + laser spot, triangulated
Function
Validates instrument pose before each step
The problem

Doctors are trained the same way they've always been — by watching, repeating, and sometimes getting it wrong. MedsApp is trained the same way, minus the third option.

01
Human-guided robotics still means human errorEven when a robot's mechanical arms do the work, a person is still steering, monitoring, or correcting — and mistakes in notes, charts, and prep still make it into procedures.
02
Training doesn't compoundA doctor's skill lives in one person. A trained model's skill is archived, reused, and refined after every procedure it runs.
03
Staffing caps repetitive careNon-invasive, repetitive procedures are exactly where clinical staff time is most constrained — and most automatable.
System architecture

Two systems, one procedure.

MedsApp.ai is built from two embodiments that work together: a system that learns, and a system that acts. Everything below is drawn directly from the filed claims.

Embodiment 1

Training System

Builds and refines the archived data pool a procedure will be measured against — from human-performed demonstrations, self-training, or both.

  • Trains and selects one or more procedure models
  • Stores images, video, depth data, point clouds, ultrasound, CBCT, MRI, force, torque, and tactile signals
  • Refines models after every completed procedure
Embodiment 2

Autonomous Physical Procedures

Executes the procedure through a robotic apparatus, comparing real-time sensor data to the archived pool at every checkpoint — with no real-time human guidance.

  • Actuator-driven instrument control with live sensor feedback
  • Controller determines and updates the procedure plan step by step
  • Autonomously issues control commands to the robotic apparatus
Proprietary methods

Three ways MedsApp decides what happens next.

Named and defined in the filing — the mechanisms the controller uses to plan, validate, and correct a procedure in progress.

METHOD 01

Instructional Twin Procedures

Step-by-step instructional sequences established from training data before a procedure ever starts — a rehearsed twin of the procedure to come.

STATUS → pre-procedure planning
METHOD 02

Bouncing

Real-time patient data is continuously compared — "bounced" — against archived data from many prior patients, thousands of times per second, to select the next action.

STATUS → in-procedure comparison
METHOD 03

Marked Positions

Measured distances from reference markers on the patient, the operating surface, and the instrument — triangulated in a Cartesian frame to validate pose before every step.

STATUS → spatial validation
Engineered for containment

Autonomy with a hard stop.

The controller doesn't just execute — it watches its own confidence, and escalates the moment something doesn't check out.

400,000
evaluation loops before forced escalation to a human operator
7 min
maximum time on an unresolved challenge condition — whichever limit hits first
Δ<0.01mm
target pose tolerance from triangulated marker validation
Qubits
quantum-assisted computation layered with classical algorithms for procedure planning
Images & video Depth data Point clouds Ultrasound CBCT MRI Force & torque Tactile signals Physiological signals Instrument pose & trajectory UMMDA matter detection
Where it applies

Filed across six clinical domains.

Invasive and non-invasive procedures, plus the automated intake and vitals work that surrounds them.

Medical
INVASIVE + NON-INVASIVE
Dental
PROCEDURAL
Cosmetic
NON-INVASIVE
Dermatology
NON-INVASIVE
Ophthalmology
PRECISION
Veterinary
CROSS-SPECIES
Intellectual property

The filings behind the system.

Detection, Inc. has consolidated its autonomous-medicine and matter-detection research into two active patent families.

USPTO 19/532,226
Robots Self-Directed by Quantum Computing and Classical Algorithms for Autonomous Physical Procedures — Medical, Dental, Cosmetic, Dermatology, Ophthalmology & Veterinary Applications
FILED FEB 2026
PCT/US26/14456
International companion filing to 19/532,226, extending priority across the same claim set
FILED FEB 2026
US2023035583A1 (18/141,374)
Universal Multipurpose Matter Detection AI ("UMMDA") — blood, urine, sputum and biological specimen analysis
NON-PROVISIONAL
WO/2023/224791
International companion filing for UMMDA (PCT/US23/202501)
WIPO

Built for the innovators and funders ready to move past electronic charts.

MedsApp autonomous procedures become available for select non-invasive applications as funding closes.