CN-122025068-A - AI-assisted urinary testing device

CN122025068ACN 122025068 ACN122025068 ACN 122025068ACN-122025068-A

Abstract

The technical scheme of the disclosure carries out preliminary diagnosis of urinary tract by utilizing historical case information and sweat metabolite signals, and in the case of abnormality, comprehensive processing and judgment are carried out on multi-mode information such as a first time sequence electrochemical signal, a second time sequence electrochemical signal, current information, metabolic waste marker information, inflammation and infection indexes by utilizing a deep detection branch and a dynamic weight distribution branch; further, in the technical scheme, when the depth detection is performed, not only is the basic feature extraction performed by using the time sequence electrochemical signals, but also data such as current generated in the detection process of the target object are utilized, and meanwhile, the dynamic weight configuration is performed based on the preliminary detection result, so that the accuracy of early tumor screening is effectively improved, and the probability of false positive/false negative is reduced.

Inventors

XU BIN
Lei Hanqi
LI JUN
HUANG JINSHENG

Assignees

中山大学附属第七医院(深圳)

Dates

Publication Date: 20260512
Application Date: 20251229

Claims (10)

1. The AI-assisted urinary detection equipment is characterized by comprising a body, a flexible electronic skin patch and a cloud platform, wherein the body comprises a detection component and an edge calculation module in communication connection with the detection component, the flexible electronic skin patch is arranged on the skin of a patient and in communication connection with the edge calculation module, the edge calculation module is in communication connection with the cloud platform, the flexible electronic skin patch is used for detecting and obtaining sweat metabolite signals, and the detection component is used for detecting urine of a target object and obtaining urine detection signals; the cloud platform responds to a detection starting request sent by the edge computing module, analyzes the detection starting request to obtain the identity information of the patient, and invokes the historical case information of the patient based on the identity information to send the historical case information to the edge computing module; The edge calculation module is used for receiving the historical case information, extracting key information in the historical case information by using a trained text processing model, and forming a historical record vector based on the key information, wherein the key information comprises operation history and patient data indexes; The edge calculation module is used for receiving the sweat metabolite signals, denoising and feature extraction are carried out on the received sweat metabolite signals by utilizing a sweat processing branch in a trained detection model to obtain sweat characterization vectors, and feature fusion and classification processing are carried out on the sweat characterization vectors and the historical record vectors by utilizing a primary detection branch in the detection model to obtain primary detection indexes and primary diagnosis data; And under the condition that at least one of the preliminary detection indexes is higher than a corresponding preset threshold value or the preliminary diagnosis data shows abnormality, the edge calculation module is further used for starting a depth detection branch in the detection model to extract feature vectors and dynamically configure weights of the urine detection signals, and determining target inflammation index data and target disease index data based on the extracted feature vectors and the dynamically configured weights.
2. The AI-assisted urinary detection apparatus of claim 1, wherein the detection component comprises a carbon-based nanosensor, a gold nanorod-quantum dot complex, a nanoporous gold electrode, a micro spectrometer, wherein the carbon-based nanosensor comprises a graphene electrode, wherein the urine detection signal comprises a first time-ordered electrochemical signal, a second time-ordered electrochemical signal, current information, metabolic waste marker information, inflammation and infection indicators, tumor-related metabolite information; The carbon-based nano sensor is used for detecting a first time sequence electrochemical signal comprising uric acid and glucose content in a target object, the gold nanorod-quantum dot complex is used for detecting a second time sequence electrochemical signal comprising urine protein and creatine content in the target object, the nano-porous gold electrode is used for measuring current information in the target object, and the micro spectrometer is used for carrying out near infrared spectrum identification treatment on the target object to obtain metabolic waste marker information, inflammation and infection indexes and tumor related metabolite information; The edge calculation module is further used for starting a depth detection branch in the detection model to denoise, normalize, extract and splice vector operations on the first time sequence electrochemical signal, the second time sequence electrochemical signal, the current information, the metabolic waste marker information, the inflammation and infection index and the tumor related metabolite information to obtain component feature vectors, inflammation feature vectors and abnormal feature vectors under the condition that at least one of the preliminary detection indexes is higher than a corresponding preset threshold value or the preliminary diagnosis data shows abnormality, and is further used for starting a dynamic weight distribution branch in the detection model to determine a first weight of the component feature vectors, a second weight of the inflammation feature vectors, a third weight of the abnormal feature vectors and a fourth weight of the historical record vectors based on the preliminary detection indexes and the preliminary diagnosis data, and performing fusion processing and classification processing on the component feature vectors, the inflammation feature vectors, the abnormal feature vectors and the historical record vectors based on the first weight, the second weight, the third weight and the fourth weight to obtain target inflammation index data and target disease index data.
3. The AI-assisted urinary detection apparatus of claim 2, wherein the edge calculation module is further configured to initiate a higher-level task-specific head corresponding to bladder cancer in the detection model to perform feature extraction and classification processing on the second time-series electrochemical signal, the current information, and the tumor-related metabolite information to obtain bladder cancer detection index data, if the target disease index data indicates that the bladder cancer probability is higher than a preset probability, wherein the higher-level task-specific head performs the following operations before performing feature extraction classification processing: denoising the second time-sequence electrochemical signal, the current information and the tumor-related metabolite information respectively, sequencing the second time-sequence electrochemical signal, the current information and the tumor-related metabolite information based on sequencing positions of the duty ratio of the effective signals in the second time-sequence electrochemical signal, the current information and the tumor-related metabolite information, and matching corresponding preset weights according to the sequencing to the second time-sequence electrochemical signal, the current information and the tumor-related metabolite information; and the special head of the high-level task is performed based on the extracted characteristics and the matched preset weight in the classifying process.
4. The AI-assisted urinary detection apparatus of claim 2, wherein the gold nanorod-quantum dot complex comprises a gold nanorod, a core layer and a shell layer, wherein the core layer and the shell layer are wrapped outside the gold nanorod, a transition layer is formed at an interface of the shell layer and the core layer, the transition layer is provided with a doped thiol compound to form a bifunctional connecting arm, one end of the bifunctional connecting arm is anchored with a quantum dot lattice, the other end of the bifunctional connecting arm is exposed with a sulfhydryl group for directional coupling of a nucleic acid probe, and the surface of the shell layer is modified with carboxyl and a hydrophobic group.
5. The AI-assisted urinary testing device of claim 1, further comprising a urethral pressure profile sensing array disposed within the patient's urethra for detecting patient's urinary pressure information and transmitting the detected urinary pressure information to the margin calculation module; the edge calculation module is also used for carrying out feature extraction on the urination pressure information and generating a pressure curve based on the extracted features.
6. The AI assisted urinary testing apparatus of claim 5 further comprising a microfluidic sensor disposed about the periphery of the urethra for detecting urine flow data and communicating to the margin calculation module; The edge calculation module performs feature extraction on the flow data and generates a urine flow rate based on the extracted features.
7. The AI-assisted urinary detection device of claim 6, further comprising a bladder ultrasound imaging module; The bladder ultrasonic imaging module is used for collecting ultrasonic signals of the bladder of the patient and transmitting the ultrasonic signals to the edge computing module, wherein the bladder ultrasonic imaging module comprises a bladder ultrasonic imaging body and a wireless communication module arranged on the body; the edge calculation module is also used for extracting multi-scale features of the ultrasonic signals by using a lightweight CNN, and generating a space attention map and bladder morphology result information based on the multi-scale features by using an encoder; The edge calculation module is also used for processing the pressure curve, the urine flow rate, the space attention map and the bladder morphology result information by using a high-layer task exclusive head corresponding to the prostate in the detection model to generate prostate result index data.
8. The AI-assisted urinary detection device of claim 7, wherein the edge calculation module is further configured to generate augmented reality display data based on bladder morphology result information and to send the augmented reality display data to a display terminal for augmented reality display.
9. The AI-assisted urinary detection apparatus of claim 2, wherein the metabolic waste marker information includes urea/creatinine ratio information, uric acid crystallization information, the inflammation and infection indicators include nitrite/leukocyte esterase information, beta 2-microglobulin information, the tumor-related metabolite information includes tryptophan metabolites, and the sweat metabolite signals include electrolyte information, creatinine information, urea information.
10. The AI-assisted urinary detection apparatus of claim 1, wherein the detection model includes a transducer+cnn hybrid architecture.

Description

AI-assisted urinary testing device Technical Field The present disclosure relates to the field of urinary testing, and in particular to an AI-assisted urinary testing device. Background Urinary detection is derived from multiple dimensions such as clinical diagnosis and treatment, health management and disease prevention, and relates to multiple scenes such as disease diagnosis, treatment monitoring, health screening and personalized medicine. The existing urinary detection technology covers various aspects of imaging examination, laboratory detection, endoscopy and the like, but the defects of low detection sensitivity, high probability of early tumor omission, low probability of false positive/false negative, insufficient standardization, larger influence of experience on detection results and poor detection timeliness still exist. Disclosure of Invention The present disclosure provides at least one AI-assisted urinary testing device to address at least one of the above-described drawbacks. According to one aspect of the disclosure, an AI-assisted urinary testing device is provided, comprising a body, a flexible electronic skin patch and a cloud platform, wherein the body comprises a detection component and an edge computing module in communication connection with the detection component, the flexible electronic skin patch is arranged on the skin of a patient and in communication connection with the edge computing module, the edge computing module is in communication connection with the cloud platform, the flexible electronic skin patch is used for detecting and obtaining sweat metabolite signals, and the detection component is used for detecting urine of a target object and obtaining urine detection signals; the cloud platform responds to a detection starting request sent by the edge computing module, analyzes the detection starting request to obtain the identity information of the patient, and invokes the historical case information of the patient based on the identity information to send the historical case information to the edge computing module; The edge calculation module is used for receiving the historical case information, extracting key information in the historical case information by using a trained text processing model, and forming a historical record vector based on the key information, wherein the key information comprises operation history and patient data indexes; The edge calculation module is used for receiving the sweat metabolite signals, denoising and feature extraction are carried out on the received sweat metabolite signals by utilizing a sweat processing branch in a trained detection model to obtain sweat characterization vectors, and feature fusion and classification processing are carried out on the sweat characterization vectors and the historical record vectors by utilizing a primary detection branch in the detection model to obtain primary detection indexes and primary diagnosis data; And under the condition that at least one of the preliminary detection indexes is higher than a corresponding preset threshold value or the preliminary diagnosis data shows abnormality, the edge calculation module is further used for starting a depth detection branch in the detection model to extract feature vectors and dynamically configure weights of the urine detection signals, and determining target inflammation index data and target disease index data based on the extracted feature vectors and the dynamically configured weights. In one possible implementation, the detection component comprises a carbon-based nano sensor, a gold nanorod-quantum dot complex, a nano porous gold electrode and a micro spectrometer, wherein the carbon-based nano sensor comprises a graphene electrode, and the urine detection signal comprises a first time-sequence electrochemical signal, a second time-sequence electrochemical signal, current information, metabolic waste marker information, inflammation and infection indexes and tumor-related metabolite information; The carbon-based nano sensor is used for detecting a first time sequence electrochemical signal comprising uric acid and glucose content in a target object, the gold nanorod-quantum dot complex is used for detecting a second time sequence electrochemical signal comprising urine protein and creatine content in the target object, the nano-porous gold electrode is used for measuring current information in the target object, and the micro spectrometer is used for carrying out near infrared spectrum identification treatment on the target object to obtain metabolic waste marker information, inflammation and infection indexes and tumor related metabolite information; The edge calculation module is further used for starting a depth detection branch in the detection model to denoise, normalize, extract and splice vector operations on the first time sequence electrochemical signal, the second time sequence electrochemical signal, the current information, the met