Intelligent automated imaging system
Inventors
BOKADIA, Pratik • CHAWARE, Amey • Horstmeyer, Roarke • Kim, Kanghyun • KONDA, Pavan
Assignees
Publication Number
US-12288158-B2
Publication Date
2025-04-29
Expiration Date
2041-01-22
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
The present disclosure describes a computational imaging system that uses a supervised learning algorithm to jointly process the captured image data to identify task-optimal hardware settings and then uses the task-optimal hardware settings to dynamically adjust its hardware to improve specific performance. The primary application of this device is for rapid and accurate automatic analysis of images of biological specimens.
Core Innovation
The invention is a computational imaging system that utilizes a supervised learning algorithm to process captured image data and dynamically identify task-optimal hardware settings. These settings are then used to automatically adjust hardware components, such as a programmable LED array and an electro-tunable lens, to enhance imaging performance for specific analytical tasks. The system includes an optical hardware system with tunable elements, a visual detector system, and a computing system operating an artificial intelligence algorithm that analyzes image data to inform hardware adjustments.
The primary problem addressed is the limitation of traditional optical microscopes, which are predominantly optimized for human observation and require laborious and error-prone mechanical scanning to inspect large areas at high resolution. Additionally, conventional microscopy techniques have difficulties resolving important features in transparent biological samples and typically involve increased cost and size for enhanced capabilities. There is a need for portable, low-cost, and high-quality imaging solutions for biological samples, which can overcome these limitations and perform rapid, accurate, and automated analysis.
By integrating dynamically tunable hardware and an artificial intelligence algorithm configured to both process digital images and optimize task-specific hardware settings, the system enables iterative optimization during image acquisition. The AI-driven approach reduces the need for manual intervention, accelerates analysis, and can be tailored for various tasks such as detecting viruses, segmenting nuclei, or counting white blood cells. The platform also supports methods that first scan at low resolution to detect regions of interest, followed by targeted high-resolution imaging, further improving efficiency and accuracy.
Claims Coverage
The patent claims broadly cover three main inventive features related to automated imaging systems, methods, and computer-readable media for AI-driven optimization of tunable hardware elements to enhance detection and analysis of biological samples.
Automated imaging system with AI-optimized programmable LED array and dynamically addressable lens element
The system comprises: - An optical hardware system featuring a patterned illumination unit that includes a programmable LED array and a dynamically addressable lens element. - A visual detector system. - A computing system that analyzes images using an artificial intelligence algorithm and changes parameters of the programmable LED array and the dynamically addressable lens element in response to automated decisions. - The artificial intelligence algorithm optimizes hardware settings for specific tasks such as detecting viruses, segmenting nuclei, detecting and counting white blood cells, and detecting tumors from tissue by reducing a task-specific loss function.
Method for detecting objects or features of interest using AI to optimize hardware settings
The method includes: 1. Using an optical hardware system with a programmable LED array and an electro-tunable lens, together with a visual detector system, to create an image of a biological sample. 2. Analyzing the image with an artificial intelligence algorithm to generate an automated decision. 3. Adjusting parameters of both the programmable LED array and the electro-tunable lens based on this decision to maximize detection of the feature of interest. 4. Optimization is aimed at specific tasks selected from detecting viruses, segmenting nuclei, detecting and counting white blood cells, and detecting tumors from tissue, where the algorithm optimizes to reduce a task-specific loss function.
Non-transitory computer readable medium for AI-driven optimization of imaging hardware
A computer-readable medium stores program code that, when executed by a processor, performs a method including: - Using an optical hardware system with a programmable LED array and an electro-tunable lens, and a visual detector system, to acquire an image of a sample. - Analyzing images with an artificial intelligence algorithm to produce automated decisions. - Adjusting at least one parameter of the programmable LED array and at least one parameter of the electro-tunable lens to maximize detection of features of interest for tasks such as detecting viruses, segmenting nuclei, detecting and counting white blood cells, and detecting tumors from tissue, with task-specific settings optimized to reduce a relevant loss function.
In summary, the claims cover systems, methods, and software implementing artificial intelligence algorithms to analyze images and dynamically control programmable imaging hardware—specifically an LED array and an electro-tunable lens—to optimize detection and analysis of biological sample features for a defined set of analytical tasks.
Stated Advantages
Reduces dependence on trained human intervention for analysis and interpretation of imaging data.
Improves speed and accuracy of diagnosis by automating image analysis and hardware optimization.
Provides portable, low-cost imaging capabilities, allowing production of compact and affordable devices.
Eliminates the need for chemical reagents or additional staining agents in sample analysis.
Enhances modularity and simplifies maintenance due to dynamically controllable hardware components.
Enables real-time and task-specific optimization of imaging conditions using artificial intelligence.
Facilitates connectivity between pathologists and remote laboratories for improved telemedicine and remote diagnostics.
Allows generalization to other pathology domains with minor modifications.
Documented Applications
Rapid and accurate automatic analysis of images of biological specimens, including blood smears.
Detection of viruses such as malaria and tuberculosis within biological samples.
Segmentation of nuclei and detection and counting of white blood cells.
Detection of tumors from tissue samples and analysis for cytology and histopathology.
Determination of blood parameters including mean corpuscular Hb concentration (MCHC), mean corpuscular Hb content (MCH), mean corpuscular volume (MCV), and RBC distribution width (RDW) using Fourier ptychographic tomography.
In vitro imaging applications for biologists and biomedical engineers, such as analysis of healing microbeads placed inside tissue.
Interested in licensing this patent?