AI Breast Cancer Screening Device with Multimodal Imaging
ThermoMind’s Vision One is an AI breast cancer screening device that combines multimodal imaging with advanced AI analysis. The system uses thermal, near-infrared, and 3D sensing to capture physiological and structural information for breast screening research.
To acquire and process multiple high-resolution sensor streams in real time, ThermoMind integrated Gidel’s FantoVision edge computing platform into the Vision One system.
ThermoMind Vision One and Multimodal Breast Imaging
ThermoMind developed Vision One as a supplementary breast imaging technology that combines multiple sensing modalities with AI-based analysis.
The system captures more than 300 data points using infrared and depth-sensing technologies to create a digital model of the breast area and analyze thermal, vascular, and structural patterns.
The imaging process is contact-free and does not use ionizing radiation. Vision One integrates long-wave infrared, near-infrared, and 3D imaging within a single system.
AI-Enhanced Thermal and 3D Imaging
The Vision One imaging structure uses multiple infrared and depth sensors positioned around the patient to capture a wide field of view across the breast and axillary areas.
AI algorithms analyze the acquired thermal and depth data to identify imaging patterns and digital biomarkers for evaluation within ThermoMind’s clinical studies.
The system also includes technician-facing software and supports digital workflows for clinical review and consultation.
FantoVision40-CXP12 for High-Bandwidth Medical Imaging
ThermoMind partnered with Gidel to address the challenge of acquiring multiple high-resolution sensor streams simultaneously while maintaining real-time processing.
Gidel’s FantoVision40-CXP12 combines NVIDIA Jetson computing with an Altera Arria 10 FPGA and integrated CoaXPress-12 acquisition in a compact embedded system.
The platform measures approximately 134 × 90 × 60 mm and weighs about 750 g, allowing it to be integrated into space-constrained medical imaging equipment.
Vision One uses CoaXPress-connected imaging sensors, making FantoVision40-CXP12 well suited to its high-bandwidth acquisition requirements.
15 Sensors with Real-Time FPGA Acquisition
The Vision One architecture uses 15 sensors capturing high-resolution image streams at approximately 30 to 60 frames per second.
FantoVision acquires these streams through its FPGA-based imaging pipeline and processes the data before downstream analysis.
The FPGA architecture provides deterministic acquisition and allows pixel-intensive operations to run close to the sensors rather than relying entirely on downstream CPU or GPU processing.
FantoVision40-CXP12 acquires directly from high-bandwidth CoaXPress cameras, supporting multi-sensor medical imaging systems that require deterministic acquisition and real-time FPGA processing.
Real-Time Compression for Multimodal Imaging Data
High-resolution multimodal imaging generates large amounts of data, creating significant bandwidth, processing, storage, and transfer requirements.
Gidel’s real-time image compression operates directly in the FPGA during acquisition.
This allows Vision One to reduce the amount of image data that must move further through the system while maintaining real-time acquisition and processing.
ThermoMind identified real-time compression as an important part of the Vision One architecture because it allows image data to be processed during acquisition rather than only afterward.
Why FantoVision Fits the Vision One Architecture
Vision One required a compact platform capable of combining high-bandwidth image acquisition, FPGA processing, real-time compression, and embedded computing.
FantoVision provides these capabilities in a single embedded architecture:
- High-bandwidth CoaXPress-12 image acquisition
- Deterministic FPGA processing
- Real-time image compression
- NVIDIA Jetson computing for embedded AI and application processing
- Compact 134 × 90 × 60 mm form factor
This integration reduces the need for separate acquisition, processing, and embedded computing hardware inside the medical imaging system.
Clinical Evaluation of the ThermoMind Vision One Device
ThermoMind is evaluating Vision One through international multicenter clinical studies.
The studies evaluate thermal video streams combined with AI algorithms and compare the approach with established breast diagnostic procedures. ThermoMind has reported collaboration with institutions including University Hospital Heidelberg, MD Anderson Cancer Center, and Assuta Medical Centers.
The goal is to evaluate the diagnostic performance of the Vision One approach and its potential role as a supplementary breast imaging and screening technology.
Gidel FPGA Technology for Medical Imaging Systems
The ThermoMind case study demonstrates how Gidel FPGA-based imaging technology can support medical systems that require simultaneous multi-sensor acquisition, deterministic processing, real-time compression, and embedded AI computing.
Gidel Medical Imaging Solutions support high-bandwidth imaging architectures including compact embedded systems, FPGA frame grabbers, and custom image-processing platforms.
Learn more about FantoVision: FantoVision Edge AI Systems
Watch the ThermoMind Vision One case study video:
