Ultrasound and artificial intelligence, explained
Ultrasound is the most accessible medical imaging modality. It is portable, affordable and free of ionizing radiation. It is also the most dependent on the person holding the probe: the system shows grayscale images, and knowing which anatomy is on the screen has always been the user's job.
Artificial intelligence gives the system its own understanding of that anatomy. Here is what it does, where it helps and how it is regulated.
What it does
SmartAlpha delivers software that understands ultrasound images as they are acquired. Models trained on large sets of expert-annotated scans recognize anatomical structures such as nerves, vessels, muscles, tendons and bones, and label them on the live image.
On top of that, it can help the user position the probe, judge whether the image is good enough, and confirm when the right view has been captured. Later steps such as measurements and reporting can be supported as well.
All of this happens live. The probe moves and the image changes many times per second, so the software has to keep up with the user's hand. That is why it runs in real time, on the ultrasound system itself.
Why it matters
Reading sonoanatomy reliably takes years of training, and there are not enough experienced sonographers and specialists to meet demand. Patients wait for exams, clinicians work through full schedules, and results vary from one user to the next.
Part of the reason is how an exam works today. Every scan is a close collaboration between an expert and a machine that does not understand what it shows. The expert finds the right view, recognizes the anatomy, takes the measurements and documents the result, all by hand. Much of that time goes into routine steps rather than clinical judgment.
A machine with more intelligence changes that balance. When the system itself recognizes the anatomy, guides the probe to the right view and supports routine measurements, it simplifies what the expert must do at every step. Their attention can then go to what only an expert can do: interpreting what they see and deciding what it means for the patient. Each exam takes less effort, more patients can be seen, and waiting times come down. Routine scans can happen earlier, by the clinician already with the patient, and expert time is kept for complex cases.
This is the real answer to the shortage: enable beginners and empower experts. Ultrasound is moving to the bedside, and anesthesiologists, emergency physicians, primary care doctors and nurses scan more often, usually without a radiologist or sonographer in the room. With expert knowledge built into the device, newer users can perform exams with confidence, while experienced sonographers and specialists spend less time on routine steps and can see more patients.
Where it helps
- Regional anesthesia, where anesthesiology AI highlights nerves, arteries and other landmarks to help anesthesiologists perform ultrasound-guided nerve blocks.
- Musculoskeletal ultrasound of the shoulder, elbow, wrist and hand, knee, and ankle and foot, in sports medicine, orthopedics and rehabilitation.
- General imaging, including abdominal ultrasound of the liver, gallbladder, kidneys and other organs, where liver AI can simplify quantification and measurements, such as fatty liver grading, to simplify the workflow.
- Point-of-care settings such as emergency departments, primary care clinics and nursing homes, where clinicians scan occasionally.
- Training, where annotated images help clinicians learn sonoanatomy faster.
- Robotic ultrasound, since a robot also needs to know what anatomy is in view and whether it has the image the exam requires.
How it is regulated
AI that supports clinical work with ultrasound is regulated as a medical device. In the United States, most products reach the market through FDA 510(k) clearance. In the European Union, they need CE marking under the Medical Device Regulation and are registered in the EUDAMED database.
Our work
SmartAlpha builds the intelligence layer for ultrasound imaging. Our aim is to simplify what the expert must do during every exam, with real-time AI that recognizes anatomy and guides each step at the point of care, whether the user is a person or a robot. It is delivered as part of ultrasound systems from imaging vendors.
Our initial offerings are already cleared for clinical use as medical devices in many countries. Our use-case portfolio is continuously expanding, with new developments in point-of-care and general imaging use cases such as AI for regional anesthesia, orthopedics, sports medicine and abdominal ultrasound, including liver exams and measurements.
You can read more about why we do this in Putting Expert Knowledge Inside the Ultrasound, or contact us about availability in your region.