Method and system for in-bed contact pressure estimation via contactless imaging
Inventors
Ostadabbas, Sarah • LIU, Shuangjun
Assignees
Northeastern University Boston
Publication Number
US-12226203-B2
Publication Date
2025-02-18
Expiration Date
2042-06-03
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
Provided herein are systems and methods for estimating contact pressure of a human lying on a surface including one or more imaging devices having imaging sensors oriented toward the surface, a processor and memory, including a trained model for estimating human contact pressure trained with a dataset including a plurality of human lying poses including images generated from at least one of a plurality of imaging modalities including at least one of a red-green-blue modality, a long wavelength infrared modality, a depth modality, or a pressure map modality, wherein the processor can receive one or more images from the imaging devices of the human lying on the surface and a source of one or more physical parameters of the human to determine a pressure map of the human based on the one or more images and the one or more physical parameters.
Core Innovation
The technology provides a contact-less approach, termed pressure eye (PEye), for estimating the contact pressure between a human body and the surface they are lying on, utilizing vision signals such as RGB and long wavelength infrared (LWIR) images. Instead of using expensive and maintenance-intensive pressure mats, this system leverages imaging devices to capture images and incorporates physical parameters such as weight, body measurements, and gender, applying a trained model to reconstruct a dense, high-resolution pressure map of the subject lying on the surface.
The key problem addressed is the costly and difficult nature of current in-bed contact pressure monitoring systems, which rely on direct pressure mats. Such systems are not scalable for routine use in environments where monitoring is crucial, like preventing pressure ulcers in bed-bound patients. Automatic and accurate contact pressure estimation through imaging seeks to provide a cheaper, easier-to-maintain, and non-invasive alternative, enabling more effective and individualized patient care.
The PEye network employs a multi-stage dual encoding, shared decoding architecture, allowing simultaneous integration of visual cues and non-visual physical parameters to generate the contact pressure map. Enhancements such as a pixel-wise resampling (PWRS) strategy, based on a Naive Bayes assumption, and evaluation metrics like percentage of correct sensing (PCS) are introduced to improve estimation performance, particularly for sparsely distributed high-pressure regions. This end-to-end system avoids complex mechanical modeling by relying on deep learning regression from multimodal signals.
Claims Coverage
The patent claims cover several inventive features related to systems and methods for generating contact pressure maps of a human subject lying on a surface using contactless imaging and the integration of physical parameters.
System for generating a contact pressure map using imaging devices and physical parameter data
A system comprising one or more imaging devices oriented toward a surface, which are capable of generating RGB, LWIR, or depth images, and a processor with memory holding a trained model. The model is trained on datasets including human lying poses, their physical parameters, pressure maps, and corresponding images from at least one imaging modality. The processor receives images and physical parameter data of a human lying on the surface and generates a pressure map based on this combined input.
Dual encoding and shared decoding of image and physical parameter signals
The processor encodes signals representing images and physical data (physical parameters) of the human separately, concatenates the encoded signals, and decodes them jointly to produce the contact pressure map, allowing integration of multimodal data for improved estimation.
Method for generating a contact pressure map based on trained models and combined input
A method that provides a processor and memory with a trained model, receives one or more images of a subject lying on a surface from imaging devices, receives physical data representing one or more physical parameters, and then generates the contact pressure map using the trained model and the combined input.
Model training with datasets comprising physical parameters, pressure maps, and multimodal images
A method of generating and using a dataset that includes physical parameters, pressure maps, and corresponding images from one or more imaging modalities for training a model to estimate or generate contact pressure maps.
Model loss function incorporating pixel-wise resampling and physical constraints
A loss function for the trained model expressed as Ltotal=λpwrsL2-1pwrs+λphyL2phy, where L2-1pwrs is a pixel-wise resampling loss weighted inversely by pixel value density and smoothed, and L2phy is a physical loss constraining the integration of pressure over the contact area to the subject's total weight, promoting physical plausibility in predictions.
System and method operability with various physical parameters and modalities
Physical parameters can include weight, height, gender, bust, waist, hip, upper arm circumference, lower arm circumference, thigh circumference, and shank circumference. The imaging devices and surface types are versatile, supporting integration of data from multiple sensor modalities and surfaces such as beds, surgical tables, gurneys, kennels, cribs, or bassinets.
System coupling with patient posture adjustment and medical professional instructions
The system is operable to transmit instructions to medical professionals or devices for patient repositioning based on estimated pressure maps, including the use of posture scheduling algorithms and integration with repositionable beds.
In summary, the claims protect a suite of technologies for contact pressure estimation based on contactless imaging and physical parameters, with innovations in joint data encoding, model optimization strategies, dataset creation, and applications in patient monitoring and care.
Stated Advantages
Enables non-contact estimation of in-bed contact pressure, eliminating the need for expensive pressure mats.
Easy to maintain due to the absence of contact-based sensors.
Low cost compared to traditional pressure map-based approaches.
Functional with different imaging modalities, including RGB and LWIR, allowing operation even in complete darkness.
Sufficiently accurate to localize high pressure concentration areas for targeted intervention.
Provides high-resolution pressure mapping capable of supporting long-term in-bed monitoring.
Reduces the burden on nursing staff by enabling efficient workload management and individualized patient care.
System is robust to occlusions, such as blankets or sheets, maintaining estimation performance under covered conditions.
Allows for integration with posture scheduling and patient monitoring systems for prevention of pressure ulcers and related syndromes.
Documented Applications
Patient monitoring and prevention of pressure related syndromes such as pressure ulcers.
Pressure causal studies using the inferred pressure distribution obtained from vision signals.
Integration with medical professional devices or repositionable beds to automate or assist in patient repositioning based on estimated pressure maps.
Monitoring and analysis in surfaces such as hospital beds, residential beds, surgical tables, cots, gurneys, floors of kennels or crates for animals, and cribs or bassinets.
Interested in licensing this patent?