Embedded Systems for Healthcare and Industrial Applications

Chimaera GmbH develops custom AI solutions for embedded systems, targeting healthcare and industrial applications. By integrating edge computing and AI workflows directly into compact, energy-efficient embedded hardware, Chimaera brings computing power to the data source — increasing efficiency, accuracy, and enabling entirely new application possibilities. Their solutions are tailored for rapid deployment and cost-effective outcomes, serving clinical environments, medical device development, and industrial quality inspection workflows.

The company has been developing individualized AI networks since 2012 and brings over 18 years of experience in radiology and 3D image processing. Their core competency lies in designing compact, high-performance AI models optimized for resource-constrained embedded hardware, including the NVIDIA Jetson product family.

Key Application Areas

• Automated image and signal processing to support medical diagnoses
• Patient monitoring with real-time tracking of vital parameters and rapid anomaly detection
• Prediction and optimization of patient needs
• Quality control and defect detection through AI-based inspection, including non-destructive testing (NDT) of workpieces
• Real-time monitoring and control of production processes, such as scanner systems

Core Solution Capabilities

• Real-time data processing and analysis directly at the data source
• Edge computing for local data processing, reducing dependency on cloud connectivity
• Analysis of large datasets for pattern recognition, result prediction, and recommendations
• Provision of on-site computing power and flexibility for clinical and industrial workflows
• Expertise in 3D AI systems for big data, including CBCT, CT, and MR imaging
• Experience with regulatory affairs for AI applications, including MDR-compliant development of AI software as a medical device

Benefits of Chimaera's Embedded AI Approach

• Application-specific AI architecture optimised for runtime and accuracy performance
• True 3D AI models delivering consistent 3D output for CAD and healthcare applications
• Ultra-fast inference times through integrated, optimised AI models
• Flexible integration into existing clinical and industrial environments
• On-site data processing supporting GDPR compliance and simplified risk management
• Easy scalability when adding additional edge devices
• Energy-efficient, cost-effective, and highly reliable hardware solutions

Embedded Systems for Healthcare and IoMT

• Edge computing combined with AI workflows provides an effective platform for the Internet of Medical Things (IoMT)
• Decentralised processing enables immediate analysis and faster response to critical patient care situations
• Facilitates risk management of sensitive patient data through local processing
• Supports clinical decision-making and improves treatment outcomes and access to services
• All AI development services account for regulatory requirements and compliance aspects from the outset

Industrial Quality Inspection and Non-Destructive Testing

• Embedded systems guarantee optimal quality and safety in industrial inspection, enhancing efficiency and minimising downtime
• Supports robot-assisted inspections and IoT-based embedded systems with networked, remotely monitored quality controls
• Real-time data analysis and machine learning enable immediate identification of material defects and structural anomalies
• AI models can be trained on a limited number of available training samples, reducing labelling costs
• Rapid adaptation of AI models to product changes, new defect classes, or new product lines through augmentation and simulation
• Expertise in CT reconstruction algorithms including FDK and iterative methods, adaptable to specific task requirements
• Simulated reconstruction data from CAD models can be combined with real images to cost-effectively train AI models for new components

Developing Scalable Embedded Hardware Solutions

• Target accuracy rate of over 99% for customised embedded platforms
• Training data requirements are minimised from the outset to keep development costs low; more real data improves model quality
• Detection complexity depends on defect appearance, shape deviation, or anomaly type; accuracy scales with data volume
• Image quality of scans directly influences reconstruction accuracy; reduced-quality CBCT scans typically require more training data
• Balanced defect class frequency improves detection accuracy; imbalanced classes may increase development time
• Trained AI models are compressed from 32-bit single precision to integer or 16-bit floating point (float16) for optimal embedded performance, reducing memory consumption while maintaining accuracy
• Post-training quantisation techniques are applied to enable advanced AI models to run on very small embedded systems at attractive price points

Hardware and Platform Expertise

• Primary platform: NVIDIA Jetson product family, each a complete system-on-module (SOM) including GPU, CPU, memory, power management, and high-speed interfaces
• Configurations range from compact, low-power devices such as the Jetson Nano — suitable even for drone deployment — to high-performance modules such as the NVIDIA Jetson Orin for processing large 3D image or video data
• Stand-alone system architecture makes embedded solutions well-suited for medical device risk management

Chimaera operates from Erlangen, Germany, and develops AI solutions in compliance with MDR regulations, offering AI software as a medical device. Their deep specialisation in 3D AI model development — capturing full volumetric context rather than processing data layer by layer — results in higher recognition rates and is a defining differentiator for both healthcare and industrial embedded applications.