前沿专利观察 / Patent Frontier Watch
第006期：AI前沿：视觉、算力、医疗与场景化智能
本期重写为逐件研读版，重点不是罗列AI热词，而是看每件专利如何把模型、数据、验证、安全、算力或场景任务写成可执行的技术链条。
为什么选这些专利
本期选择标准不是随机编号，也不是标题看起来热门，而是每件专利都对应一个可被拆解的工程问题：它必须能说明一个具体痛点，并且在公开文本中给出结构、流程、材料组合、数据处理或控制策略。下面按逐件研读方式展开。
1. Artificial Intelligence Apparatus and Method for monitoring prescribed Movements
代表专利：US-2025256158-A1。申请人/权利人：SALSABILI HODA (IN) HOSSEINI MOHAMMAD (US)。优先权日：2024/02/12；公开日：2025/08/14。
它真正发现的问题
这件专利的问题落在“Artificial Intelligence Apparatus and Method for monitoring prescribed Movements”这个具体AI场景：This invention provides an artificial intelligence (AI) apparatus and method for prescription, real-time monitoring and tracking individualized physical therapy exercise plan and classes to remote users. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：This invention provides an artificial intelligence (AI) apparatus and method for prescription, real-time monitoring and tracking individualized physical therapy exercise plan and classes to remote users. In one aspect, the present disclosure comprising a method utilizing the AI apparatus for prescribing body movements to individuals and/or groups that can be accessed via a digital communication network by a user at a remote location, sending digital video and audio content comprising AI-prescribed individualized e...
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
2. Scalable neural network processing engine
代表专利：US-2025165747-A1。申请人/权利人：APPLE INC (US)。优先权日：2018/05/04；公开日：2025/05/22。
它真正发现的问题
这件专利的问题落在“Scalable neural network processing engine”这个具体AI场景：Embodiments relate to a neural processor circuit with scalable architecture for instantiating one or more neural networks. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：Embodiments relate to a neural processor circuit with scalable architecture for instantiating one or more neural networks. The neural processor circuit includes a data buffer coupled to a memory external to the neural processor circuit, and a plurality of neural engine circuits. To execute tasks that instantiate the neural networks, each neural engine circuit generates output data using input data and kernel coefficients. A neural processor circuit may include multiple neural engine circuits that are selectively a...
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
3. Pixel-based deformation of fashion items
代表专利：US-12536751-B2。申请人/权利人：SNAP INC (US)。优先权日：2023/08/16；公开日：2026/01/27；授权日：2026/01/27。
它真正发现的问题
这件专利的问题落在“Pixel-based deformation of fashion items”这个具体AI场景：Methods and systems are disclosed for using machine learning models to perform pixel-based deformation of fashion items. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：Methods and systems are disclosed for using machine learning models to perform pixel-based deformation of fashion items. The methods and systems receive one or more images depicting a first person in a first pose and receive a source image depicting a target fashion item worn on a portion of a body of a second person in a second pose. The methods and systems process, using one or more machine learning models, the one or more images together with the source image to generate a flow field indicating existence and lo...
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
4. Data retrieval and validation for asset onboarding and deriving asset characteristics
代表专利：US-12380503-B1。申请人/权利人：TRETE INC (US)。优先权日：2023/03/24；公开日：2025/08/05；授权日：2025/08/05。
它真正发现的问题
这件专利的问题落在“Data retrieval and validation for asset onboarding and deriving asset characteristics”这个具体AI场景：The present invention extends to methods, systems, and computer program products for data retrieval and validation for asset onboarding and deriving asset characteristics. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：The present invention extends to methods, systems, and computer program products for data retrieval and validation for asset onboarding and deriving asset characteristics. A first set of data associated with an asset is collected. Identifiers associated with the first set of data are created. A second set of data associated with the asset is collected based on the identifiers. The first set of data set and the second set of data are compared based on the identifiers. The first set of data is validated based on the...
给我们的启示
启示是，RAG和企业AI的壁垒常在数据链路：检索什么、如何验证、怎样回填业务系统、错误如何处理，都是可保护点。
5. Memory device including 2-transistor memory cell structure for neural network
代表专利：US-2025029638-A1。申请人/权利人：MICRON TECHNOLOGY INC (US)。优先权日：2023/07/20；公开日：2025/01/23。
它真正发现的问题
这件专利的问题落在“Memory device including 2-transistor memory cell structure for neural network”这个具体AI场景：Some embodiments include apparatuses and methods of operating the apparatuses. One of the apparatuses includes a first memory cell and a second memory cell, each of the first and second memory cells including a first transistor including a first region and a first charge storage structure separated from the first region; a second transistor including a seco... 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：Some embodiments include apparatuses and methods of operating the apparatuses. One of the apparatuses includes a first memory cell and a second memory cell, each of the first and second memory cells including a first transistor including a first region and a first charge storage structure separated from the first region; a second transistor including a second region formed over the first charge storage structure; a first data line coupled to the first memory cell configured to provide a first sum based on current ...
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
6. Systems and methods for weakly supervised training of a model for monocular depth estimation
代表专利：US-11386567-B2。申请人/权利人：TOYOTA RES INST INC (US)。优先权日：2019/07/06；公开日：2022/07/12；授权日：2022/07/12。
它真正发现的问题
这件专利的问题落在“Systems and methods for weakly supervised training of a model for monocular depth estimation”这个具体AI场景：System, methods, and other embodiments described herein relate to semi-supervised training of a depth model for monocular depth estimation. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：System, methods, and other embodiments described herein relate to semi-supervised training of a depth model for monocular depth estimation. In one embodiment, a method includes training the depth model according to a first stage that is self-supervised and that includes using first training data that comprises pairs of training images. Respective ones of the pairs including separate frames depicting a scene of a monocular video. The method includes training the depth model according to a second stage that is weakl...
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
7. AI intelligent sound wave pulse resonance mouse repeller and mouse repellent method
代表专利：US-12382947-B2。申请人/权利人：XIE LIHAI (TW)。优先权日：2023/03/13；公开日：2025/08/12；授权日：2025/08/12。
它真正发现的问题
这件专利的问题落在“AI intelligent sound wave pulse resonance mouse repeller and mouse repellent method”这个具体AI场景：An AI intelligent sound wave pulse resonance mouse repeller and a mouse repellent method are provided, the mouse repeller includes an AI intelligent chip, a power converter, an indicator light, and an ultrasonic horn. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：An AI intelligent sound wave pulse resonance mouse repeller and a mouse repellent method are provided, the mouse repeller includes an AI intelligent chip, a power converter, an indicator light, and an ultrasonic horn. The AI chip can freely generate a high-speed and ultra-high frequency pulse wave with random changes, can perform AI calculation with environmental sound, and then perform random scanning. A generated pulse signal resonates with a sound frequency of animals such as mice and bats, is superimposed with...
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
8. System and method for cybersecurity threat detection utilizing static and runtime data
代表专利：US-12278825-B2。申请人/权利人：WIZ INC (US)。优先权日：2022/01/31；公开日：2025/04/15；授权日：2025/04/15。
它真正发现的问题
这件专利的问题落在“System and method for cybersecurity threat detection utilizing static and runtime data”这个具体AI场景：A system and method for improved endpoint detection and response (EDR) in a cloud computing environment initiates inspection based on data received from a sensor deployed on a workload. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：A system and method for improved endpoint detection and response (EDR) in a cloud computing environment initiates inspection based on data received from a sensor deployed on a workload. The method includes: configuring a resource, deployed in a cloud computing environment, to deploy thereon a sensor, the sensor configured to detect runtime data; detecting a potential cybersecurity threat on the resource based on detected runtime data received from the sensor; and initiating inspection of the resource for the poten...
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
9. Machine-learning-based meal detection and size estimation using continuous glucose monitoring (cgm) and insulin data
代表专利：US-2025259727-A1。申请人/权利人：UNIV OREGON HEALTH & SCIENCE (US)。优先权日：2022/04/29；公开日：2025/08/14。
它真正发现的问题
这件专利的问题落在“Machine-learning-based meal detection and size estimation using continuous glucose monitoring (cgm) and insulin data”这个具体AI场景：Disclosed is a meal detection and meal size estimation machine learning technology. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：Disclosed is a meal detection and meal size estimation machine learning technology. In some embodiments, the techniques entail applying to a trained multioutput neural network model a set of input features, the set of input features representing glucoregulatory management data, insulin on board, and time of day, the trained multioutput neural network model representing multiple fully connected layers and an output layer formed from first and second branches, the first branch providing a meal detection output and t...
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
10. System to detect foot abnormalities
代表专利：US-12419521-B2。申请人/权利人：EMPO HEALTH INC (US)。优先权日：2020/08/21；公开日：2025/09/23；授权日：2025/09/23。
它真正发现的问题
这件专利的问题落在“System to detect foot abnormalities”这个具体AI场景：Systems, devices, and methods for detecting a foot abnormality includes a platform configured to be stood upon by a user, an imaging device within the platform, and a processor connected to the imaging device. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：Systems, devices, and methods for detecting a foot abnormality includes a platform configured to be stood upon by a user, an imaging device within the platform, and a processor connected to the imaging device. The processor is configured to detect a foot abnormality from images gathered by the imaging device.
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
11. Utility usage prediction and optimization
代表专利：US-2025085673-A1。申请人/权利人：IBM (US)。优先权日：2023/09/12；公开日：2025/03/13。
它真正发现的问题
这件专利的问题落在“Utility usage prediction and optimization”这个具体AI场景：Techniques are provided for utility usage prediction and optimization. In one embodiment, the techniques involve receiving user data, receiving utility data, generating, via a trained physics-informed neural network (PINN), a utility usage prediction based on the utility data, generating a utility plan based on the user data and the utility usage prediction... 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：Techniques are provided for utility usage prediction and optimization. In one embodiment, the techniques involve receiving user data, receiving utility data, generating, via a trained physics-informed neural network (PINN), a utility usage prediction based on the utility data, generating a utility plan based on the user data and the utility usage prediction, wherein the utility plan includes limits or restrictions of a utility usage, and controlling the utility usage based on the utility plan.
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
12. Assay reading method
代表专利：US-12373946-B2。申请人/权利人：FORSITE DIAGNOSTICS LTD (GB)。优先权日：2020/05/11；公开日：2025/07/29；授权日：2025/07/29。
它真正发现的问题
这件专利的问题落在“Assay reading method”这个具体AI场景：A method for reading a test region of an assay includes: capturing a plurality of images of an assay with an imaging device; from each image of the plurality of images, extracting a region of interest comprising pixels of the image associated with a test region of the assay; from each extracted region of interest, estimating respective intensity values of a. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：A method for reading a test region of an assay includes: capturing a plurality of images of an assay with an imaging device; from each image of the plurality of images, extracting a region of interest comprising pixels of the image associated with a test region of the assay; from each extracted region of interest, estimating respective intensity values of at least a portion of the pixels; grouping the estimated intensity values into one or more clusters, said grouping comprising determining a total number of inten...
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
13. Wearable sensor and healthcare management system using a wearable sensor
代表专利：US-12354747-B2。申请人/权利人：EMFIT OY (FI)。优先权日：2019/03/31；公开日：2025/07/08；授权日：2025/07/08。
它真正发现的问题
这件专利的问题落在“Wearable sensor and healthcare management system using a wearable sensor”这个具体AI场景：A modular artificial intelligence (AI) engine for a healthcare system includes clinical analysis modules, each clinical analysis module is directed to a respective predetermined medical area. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：A modular artificial intelligence (AI) engine for a healthcare system includes clinical analysis modules, each clinical analysis module is directed to a respective predetermined medical area. The respective clinical analysis module analyzes healthcare data of a patient with a respective learned algorithm directed to the respective predetermined medical area. Each clinical analysis module includes software interfaces for incoming healthcare data and outgoing processed data to integrate the respective clinical analy...
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
14. Optimizing a prognostic-surveillance system to achieve a user-selectable functional objective
代表专利：US-12488265-B2。申请人/权利人：ORACLE INT CORP (US)。优先权日：2021/07/28；公开日：2025/12/02；授权日：2025/12/02。
它真正发现的问题
这件专利的问题落在“Optimizing a prognostic-surveillance system to achieve a user-selectable functional objective”这个具体AI场景：The disclosed embodiments relate to a system that optimizes a prognostic-surveillance system to achieve a user-selectable functional objective. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：The disclosed embodiments relate to a system that optimizes a prognostic-surveillance system to achieve a user-selectable functional objective. During operation, the system allows a user to select a functional objective to be optimized from a set of functional objectives for the prognostic-surveillance system. Next, the system optimizes the selected functional objective by performing Monte Carlo simulations, which vary operational parameters for the prognostic-surveillance system while the prognostic-surveillance ...
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
15. Method for reconceptualization of deep learning-based translation of t1-weighted image to magnetic resonance angiography (mra) and a system for deep learning-based translation of t1-weighted image to vasculature image of a brain
代表专利：US-2025245956-A1。申请人/权利人：HONG KONG CENTRE FOR CEREBRO CARDIOVASCULAR HEALTH ENGINEERING LTD (HK)。优先权日：2024/01/30；公开日：2025/07/31。
它真正发现的问题
这件专利的问题落在“Method for reconceptualization of deep learning-based translation of t1-weighted image to magnetic resonance angiography (mra) and a system for deep learning-based translation of t1-weighted image to vasculature image of a brain”这个具体AI场景：The present invention provides an artificial intelligence engine configured to a UNet model including an encoder and a decoder that can synthesize 3D MRA and MIP of MRA using acquired single contrast MR image (T1-w MR image) for the same subject while preserving the continuity of vascular anatomy and important vascular morphological features. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：The present invention provides an artificial intelligence engine configured to a UNet model including an encoder and a decoder that can synthesize 3D MRA and MIP of MRA using acquired single contrast MR image (T1-w MR image) for the same subject while preserving the continuity of vascular anatomy and important vascular morphological features.
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
16. Systems, methods, and devices for medical image analysis, diagnosis, risk stratification, decision making and/or disease tracking
代表专利：US-12324695-B2。申请人/权利人：CLEERLY INC (US)。优先权日：2020/01/07；公开日：2025/06/10；授权日：2025/06/10。
它真正发现的问题
这件专利的问题落在“Systems, methods, and devices for medical image analysis, diagnosis, risk stratification, decision making and/or disease tracking”这个具体AI场景：The disclosure herein relates to systems, methods, and devices for medical image analysis, diagnosis, risk stratification, decision making and/or disease tracking. 研读重点是它如何把模型能力约束到可验证的数据处理、判断或部署流程中。
它的技术方案
公开摘要显示，方案的具体构成是：The disclosure herein relates to systems, methods, and devices for medical image analysis, diagnosis, risk stratification, decision making and/or disease tracking. In some embodiments, the systems, devices, and methods described herein are configured to analyze non-invasive medical images of a subject to automatically and/or dynamically identify one or more features, such as plaque and vessels, and/or derive one or more quantified plaque parameters, such as radiodensity, radiodensity composition, volume, radiodens...
给我们的启示
启示是，AI专利应把模型放回具体任务闭环，写清数据、模型、验证、反馈和部署载体之间的关系。
本期结论
这些AI专利共同说明，真正可布局的不是“用了大模型/机器学习”，而是模型外层的验证、安全、数据入口、硬件执行和具体业务闭环。
资料来源 / Sources
本期依据 PubChem patent JSON、Google Patents 公开页面和原公开链接研读整理。正式FTO或授权稳定性判断仍需进一步核对官方登记簿、审查历史、同族和权利要求全文。
- US-2025256158-A1 - Artificial Intelligence Apparatus and Method for monitoring prescribed Movements
- US-2025165747-A1 - Scalable neural network processing engine
- US-12536751-B2 - Pixel-based deformation of fashion items
- US-12380503-B1 - Data retrieval and validation for asset onboarding and deriving asset characteristics
- US-2025029638-A1 - Memory device including 2-transistor memory cell structure for neural network
- US-11386567-B2 - Systems and methods for weakly supervised training of a model for monocular depth estimation
- US-12382947-B2 - AI intelligent sound wave pulse resonance mouse repeller and mouse repellent method
- US-12278825-B2 - System and method for cybersecurity threat detection utilizing static and runtime data
- US-2025259727-A1 - Machine-learning-based meal detection and size estimation using continuous glucose monitoring (cgm) and insulin data
- US-12419521-B2 - System to detect foot abnormalities
- US-2025085673-A1 - Utility usage prediction and optimization
- US-12373946-B2 - Assay reading method
- US-12354747-B2 - Wearable sensor and healthcare management system using a wearable sensor
- US-12488265-B2 - Optimizing a prognostic-surveillance system to achieve a user-selectable functional objective
- US-2025245956-A1 - Method for reconceptualization of deep learning-based translation of t1-weighted image to magnetic resonance angiography (mra) and a system for deep learning-based translation of t1-weighted image to vasculature image of a brain
- US-12324695-B2 - Systems, methods, and devices for medical image analysis, diagnosis, risk stratification, decision making and/or disease tracking