Dyson CameraJet Brings AI-Powered Precision to Oral Care

The AI-powered Dyson CameraJet combines real-time imaging, machine learning, and precision fluid delivery to identify interdental gaps and target them while a user brushes.

The more significant story is the technology architecture, with a consumer device turning a routine health behaviour into a machine-perception problem, with potential applications for personalised care, longitudinal data, and connected health.

Image

How Dyson CameraJet Uses AI to See Where Brushing Misses

The central technology of the CamerJet toothbrush is Gap Optical Targeting, Dyson’s machine-learning system for detecting, tracking, and predicting gaps between teeth.

A 100,000-pixel macro camera captures 28 live images per second, while stroboscopic illumination helps compensate for the movement of the brush head. When the system identifies a gap, the device can trigger a targeted burst of mouthrinse.

Dyson says the response occurs within 100 milliseconds of detecting a gap.

The model was developed using more than 470,000 dental images, while the complete device runs on 16 million lines of code. These numbers show that the product is not simply an electric toothbrush with a camera attached, but, rather, it’s a tightly integrated sensing, inference, and actuation system.

This architecture is important from a HealthTech perspective.

The camera generates a visual signal and machine learning interprets that signal in context. A physical actuator then responds to the model output.

This is a closed loop in which perception leads directly to an intervention.

That pattern is increasingly relevant across healthcare. Connected devices can move from collecting data to responding to it.

The same principle appears in passive health monitoring, where sensors capture routine biological signals and software interprets them over time. Pharmatica has explored this shift through its Insights into Throne Science and other digital health technologies.

Precision Engineering Makes AI Useful

The machine-learning layer is only one part of CameraJet. Dyson has paired it with fluid dynamics, brushing mechanics, imaging, and connectivity.

The precision jet uses a broad conical spray rather than a narrow, needle-like stream.

Dyson says its engineers developed a proxy plaque system with the National University of Singapore’s Faculty of Dentistry to test plaque removal during development. This allowed repeated engineering iterations without requiring a clinical trial for every design change.

This is important because AI does not create the health outcome by itself. Its value depends on the physical system that converts a prediction into an action.

CameraJet also uses an anti-gravity tank and diaphragm pump to maintain fluid delivery when the toothbrush is tilted.

Its variable sonic oscillation is designed to reduce brush-head stalling, while a target lens expands the camera’s field of view, and the brush head contains Radio Frequency Identification (RFID) technology to recognise the head and track its use.

This is a useful model for pharma leaders assessing connected health technologies. The intelligence layer cannot be separated from the sensing and intervention layers.

A highly capable algorithm has limited clinical value if the sensor produces poor data or the intervention cannot be delivered consistently.

View interview How Should Pharma Companies Audit AI Vendors?

How Should Pharma Companies Audit AI Vendors?

“This is one of the riskiest industries there is because you're dealing with human lives, you're dealing with experimental protocols, and you're dealing with regulatory bodies where you might not get another shot at that clinical trial."

Image

A Consumer Device to Connected Health Platform

CameraJet also illustrates how health technology can be built into ordinary every-day routines.

The toothbrush connects through Wi-Fi and Bluetooth to the MyDyson app. Users can access guided cleaning, coverage mapping, brushing and flossing insights, and personalised feedback.

Dyson says the camera is active only during live viewing or automatic jetting, and images are not stored on the device or in the cloud.

This approach is significant because it places a form of computer vision inside a highly repetitive health behaviour. Instead of asking users to remember what they did, the system observes a defined task and provides feedback during or after it.

That resembles the wider evolution of passive and near-passive monitoring. Throne Science, for example, uses optical and acoustic sensing to capture bathroom-related health signals without requiring continuous manual logging.

Pharmatica’s broader Insights into digital health technologies in clinical drug development similarly highlights the growing importance of data collected outside conventional clinical settings.

The implication for pharma is not that a toothbrush provides a clinical endpoint. Instead it’s that consumer devices can create new channels for real-world behavioural and physiological data.

The strategic question is whether those data can be made reliable, interpretable, and relevant to a defined clinical or research purpose.

Image

Oral Health Creates a Large Digital Health Opportunity

The potential scale of the problem gives this technology a broader context.

The World Health Organization estimates that oral diseases affect nearly 3.7 billion people globally. Oral conditions are largely preventable, yet access to prevention and treatment remains uneven, particularly in low- and middle-income countries.

Dyson’s technology addresses a narrower behavioural problem in that people often do not clean effectively between teeth. The company cites research indicating that nine in ten people do not floss as regularly as recommended, while average brushing time is only around 45 seconds.

CameraJet therefore targets an important gap between what people know they should do and what they consistently do. The technology attempts to reduce that gap by making the device responsive to the user’s actual brushing environment.

That is a broader healthtech pattern worth watching. Digital systems can make adherence less dependent on memory, motivation, or manual reporting by embedding guidance into the activity itself.

This could eventually matter in areas such as patient support, adherence monitoring, remote intervention, and decentralised research.

But the evidentiary threshold changes when a consumer wellness technology demonstrates medical use.

Data generated during everyday behaviour must be assessed for accuracy, reproducibility, bias, privacy, and clinical relevance before it can support regulated decisions.

Image

The Real Opportunity: The Closed-Loop Model

CameraJet is particularly interesting because it demonstrates a closed-loop model at consumer scale.

Its sequence is straightforward:

• A camera captures the oral environment.
• Machine learning identifies and predicts an interdental gap.
• The device triggers a targeted fluid jet.
• The app provides feedback on cleaning behaviour and coverage.

Each step creates an opportunity for engineering optimisation. It also creates a potential failure point.

Computer vision must work across different mouths, lighting conditions, brushing angles, and movement patterns.

Therefore, the training data must represent the population in which the technology is deployed. Additionally, the system must distinguish meaningful anatomical features from artefacts created by motion or fluid. And the physical response must remain safe and reliable when the device is used repeatedly.

These issues are familiar to organisations developing AI-enabled products.

Model performance is only one component of system performance. Data provenance, representativeness, human factors, cybersecurity, privacy, and post-market monitoring can become equally important as the technology matures.

Dyson says 661 engineers worked on CameraJet and that the company filed 38 patents during development. It also says 30 dental professionals contributed globally.

Those impressive figures illustrate the multidisciplinary nature of the system, spanning machine learning, optics, mechanical engineering, fluid dynamics, and dental science.

View interview AI in Pharma: Why the Future of Healthcare Starts With Patients, Not Tech

AI in Pharma: Why the Future of Healthcare Starts With Patients, Not Tech

Kate O’Reilly, President & Chair of the Healthcare Businesswomen’s Association (HBA) Dublin-Ireland Chapter & Healthcare Transformation Partner at Roche, discusses AI in pharma, patient engagement, and the future of healthcare innovation.

Image

Building Every-day Sensing and Machine Learning

CameraJet shows that sophisticated sensing and machine learning can be structured well enough to operate inside everyday products.

That creates a new healthtech layer between traditional medical devices and consumer electronics. The device can sense a biological or behavioural context, interpret it through connected software, and provide an immediate response.

This could influence how future patient engagement and evidence strategies are designed.

Connected devices could potentially support treatment adherence, symptom tracking, patient-reported outcomes, or remote monitoring, provided the intended use and evidence requirements are clearly defined.

However, more data does not automatically mean better evidence. The clinical usefulness of a digital signal depends on whether it measures something meaningful, whether it performs consistently across relevant populations, and whether it can be connected to an outcome that matters.

Pharmatica understands that the next phase will be defined by technologies that connect sensing, algorithms, human behaviour, clinical evidence, and responsible data governance into one coherent system.

Pharmatica: Insight. Connection. Impact.