NON-INVASIVE GLUCOSE MEASUREMENT SYSTEM THROUGH SIMULTANEOUS FUSION OF FACIAL BIOMECHANICAL AND OPTICAL SIGNALS

Abstract

The invention encompasses an advanced system that enables non-invasive, continuous, and motion-based blood glucose measurement by simultaneously capturing biomechanical vibrations generated during involuntary micro-expressions and emotional reflexes of facial muscles, and glucose-induced optical changes in the subcutaneous interstitial fluid, through integrated accelerometer, gyroscope, and multispectral optical sensors, and fusing them with a deep learning-based hybrid neural network (CNN-LSTM).

Description

1 TARIFF SIMULTANEOUS FUSION OF FACIAL BIOMECHANICAL AND OPTICAL SIGNALS WITH NON- INVASIVE GLUCOSE MEASUREMENT SYSTEM Technical Area The invention relates to 5 involuntary micro-expressions of the facial muscles and emotional reflexes that occur during these processes. biomechanical vibrations and glucose-induced optical changes in the subcutaneous interstitial fluid, through integrated accelerometer, gyroscope and multispectral optical sensors simultaneously captured and processed by a deep learning-based hybrid neural network (CNN-LSTM) Non-invasive, continuous, and uninterrupted blood glucose measurement even during movement, achieved through fusion. It includes an advanced system that enables this. 10 The invention is particularly relevant in the field of non-invasive glucose monitoring technologies, as it aims to activate existing systems. signal noise, discontinuities in skin contact, and reliance solely on optical data. An innovative system that aims to overcome its fundamental limitations, such as low specificity in measurements. It offers. Previous Technique 15 In the field of non-invasive glucose monitoring technologies, current systems are basically divided into three main categories. The approach is shaped around: optical spectroscopy-based methods (infrared, Raman, photoplethysmography), bioimpedance analysis and subcutaneous fluid (interstitial fluid) electrochemical or microfluidic-based analyses. The common goal of these methods is to analyze invasive blood samples. Without the need for sampling, the glucose concentration in tissue or fluid can be measured at 20 Obtaining clinically meaningful data on blood sugar levels by measuring them indirectly or directly. However, each of these methods has its own limitations, especially in everyday use and mobility. Due to technical limitations encountered during the procedure, the accuracy offered by invasive systems is limited. and has not yet achieved sufficient reliability. Optical spectroscopy-based systems, under the skin... Light of specific wavelengths that penetrates the interstitial fluid is reacted by glucose molecules 25 It is based on the principle of absorption or scattering. The most important technique of these systems The problem is that glucose molecules interact with skin tissue components such as water, melanin, hemoglobin, lipids, and collagen. It has a relatively weak optical signature compared to other molecules found, in particular. In infrared systems operating in the 900-1100 nm wavelength range, glucose absorption... Its spectrum almost completely overlaps with the dominant absorption band of water, which is why 30 It dramatically reduces the signal-to-noise ratio. Additionally, changes in skin temperature, Sweating, differences in blood flow rate, and individual variations in tissue density, optical This makes signal calibration extremely complex, even requiring user-based personalization. 2 Generalizability of the system is impossible unless calibration models are created. This situation makes optical-based non-invasive glucose monitoring devices clinically acceptable. producing consistent results even within the margin of error (e.g., between 15-20%). to prevent, especially, critical threshold values ​​such as hypoglycemia or hyperglycemia This poses serious security risks in its determination. 5 Bioimpedance analysis-based systems, on the other hand, detect low-level alternating currents passing through tissue. glucose concentration can be determined by measuring changes in the resistance and capacitance characteristics of the current. It attempts to predict the presence of glucose molecules in the extracellular fluid. It alters the dielectric properties, and this change can be theoretically detected by impedance spectroscopy. It can be detected. However, the biggest technical disadvantage of this approach is the impedance 10 the signal depends not only on glucose, but also on electrolyte balance (especially sodium, potassium ion concentrations), body water ratio, skin moisture, sweat gland activity, blood also seriously affects volume and even micro-mechanical pressure changes that occur during movement. The extent to which it is affected. Many of these parameters are independent of glucose levels. It is changing, and the separation of the glucose-derived component from the impedance signal is high 15 This presents a highly complex signal processing problem. Furthermore, bioimpedance... Ensuring continuous and stable contact between the electrodes used for measurements and the skin, user It becomes quite difficult during movement, instantaneous changes in contact resistance are measured. This leads to significant deviations in the results. Therefore, bioimpedance-based systems, In practical use, it often exhibits low repeatability, even under static conditions. 20 A reliable alternative for users requiring continuous monitoring (CGM). It is unable to create. Based on the analysis of subcutaneous interstitial fluid using microfluidic or electrochemical methods. These systems are minimally invasive because they require direct access to the fluid sample. They are classified and do not offer a truly non-invasive solution. These systems, 25 Microneedling or iontophoresis methods can penetrate the skin barrier and reach the interstitial fluid. reaching it, then detecting the glucose in this fluid with enzymatic or electrochemical sensors. It measures. Technically, these approaches are the closest to invasive systems in terms of accuracy. Even though they perform well, the skin barrier is breached or subcutaneous each time. The presence of installed sensors limits user comfort, and furthermore, the sensors number 30 biocompatibility and performance loss due to fouling (biological surface contamination) over time This brings with it problems such as those mentioned above. The biggest technical limitation of these systems is interstitial space. It is the physiological time delay between fluid glucose levels and blood glucose levels (typically 5- (minutes). This delay is particularly noticeable in situations where rapid glucose changes occur (for example 35 in instantaneous decision-making processes (after exercise or after rapid-acting insulin administration) This creates a critical security vulnerability. 3 These technical limitations are common to current non-invasive glucose monitoring systems. its denominator is based on only a single physical modality (optical only, electrical only, or only mechanical) durability and especially resistance to the user's daily activities movement artifacts, resulting from voluntary and involuntary contractions of facial muscles. System 5 analyzes biomechanical noise and the complex relationship between tissue-metabolic changes. Their inability to analyze at the micro level, especially in the facial area, of the mimic muscles. movements are related to articulation during speech, chewing, and emotional reflexes. Involuntary muscle contractions affect both the distance between the optical sensors and the skin, and photoplethysmography. It momentarily disrupts the stability of the signals, and in current systems, these disruptions affect the signal. This leads to its complete rejection or misinterpretation. This situation, especially in 10 Individuals who require continuous glucose monitoring should consciously remain immobile in order to take measurements. This forces them to stay there, which hinders the integration of the system into daily life. It severely restricts. The present invention solves all of these technical problems, namely biomechanical movements in the facial region. simultaneous and synchronized optical changes in the subcutaneous interstitial fluid in multiple 15 with sensors (accelerometer, gyroscope, photoplethysmography sensor, multispectral optical reader) capture, timestamp of these two different physical quantities (mechanical and optical) by mapping and combining them in a data fusion layer, and then these fused data... the data, trained to distinguish motion artifacts from glucose-induced signaling. In hybrid deep learning architecture (Convolutional Neural Network – CNN and Long Short-Term Memory – 20 It solves the problem by processing (LSTM). Thus, the invention is not purely optical or purely mechanical. Unlike traditional data-driven systems, the system captures the noise caused by motion. by making it a part of the working principle, the user's speech, facial expressions, head movements uninterrupted, high-accuracy, and clinically accurate results during all daily activities such as movements. It offers a meaningful non-invasive glucose monitoring capability. This approach is the only one currently available in the field of technology. The most fundamental limitations of multimodal systems are motion sensitivity and signal stability. the problem, multimodal sensor fusion and artificial neural network-based dynamic noise By overcoming this through modeling, a qualitative leap in non-invasive glucose monitoring technologies. It provides. Patent document US20240000346A1 states that the osmotic pressure difference of the interstitial fluid (ISF) is 30. and collection using paper-based microfluidic channels and electrochemical methods A patch system for analysis is described. The invention relates to a patch that comes into contact with the skin surface. Microneedles have a hydrogel or backing that provides osmotic pressure. a glycerol-gel layer and a paper-based substrate that absorbs liquid through capillary action between these two layers. It consists of a microfluidic channel. The system delivers subcutaneous fluid non-invasively or minimally invasively. 35 invasively collects glucose and electrochemical sensors integrated into the duct (glucose, 4 It analyzes biomarker concentrations (such as lactate) and transmits the results to a user device. Unlike the invention described in the aforementioned document, this invention is specific to the facial region. The sensor placement anticipates involuntary micro-expressions of facial muscles. It presents a data fusion approach that correlates biomechanical vibrations with glucose signaling and Multispectral imaging using active illumination at different wavelengths to detect interstitial light. It captures the optical signatures of metabolic markers in the fluid. Patent document number US20230293101A1 describes a method for determining a person's blood sugar level. The text describes a system that combines bioimpedance analysis and optical spectroscopy. The invention measures tissue impedance through electrodes that come into contact with the skin surface. An optical 10 that collects optical spectroscopy data from the same region, along with a bioimpedance circuit. By integrating the sensor array, the data from both modalities form a data fusion. By combining layers, an estimate of glucose concentration can be provided. The system is intended for use especially in areas such as the wrist, arm, or fingers. It is designed in the form of a wearable device. It differs from the invention described in the document. Therefore, this invention is not based on the bioimpedance method, but on involuntary 15°C in the facial region. biomechanical vibrations of micromimics as a glucose-related parameter to evaluate and eliminate artifacts related to facial movements using an accelerometer and The gyroscope data was not only used for filtering purposes but was directly input into the glucose prediction model. It uses it in a way that will create it. Patent document number WO2024112342A1 describes how to raise a diabetic person's glucose level to 20 A system and method for managing is described. The invention is a sensor unit (invasive (or non-invasive glucose sensor), consists of a control unit and a user interface. It processes glucose data received from the sensor and provides the user with insulin dose recommendations or warnings. The system also tracks the user's daily activities, dietary habits, and It works integrated with a mobile application that tracks physical activity levels, glucose 25 It aims to provide a comprehensive platform for management. In the document in question... Unlike the invention described above, this invention relates to a new method for collecting glucose data. It offers a sensor architecture that simultaneously detects biomechanical and optical signals in the facial region. It involves a measurement technique based on fusion, and only data processing and decision support It is not a software platform focused on algorithms. 30 Patent document number US20240065588A1 describes a wearable device for non-invasive glucose monitoring. The device and method are described. The invention is an optical sensor that comes into contact with the user's skin. array (photoplethysmography and infrared spectroscopy), a temperature sensor and an accelerometer It includes and the data collected from these sensors are processed in a machine learning model. Glucose levels are estimated. The device is designed to be worn specifically on the wrist or arm. It is designed to detect the user's movement status using accelerometer data and optical signals. It is used for filtering purposes to remove motion artifacts. The aforementioned Unlike the invention described in the document, in this invention, accelerometer data is obtained solely from optical sources. It is not used to filter motion-induced noise in signals, face biomechanical vibrations resulting from involuntary micro-expressions of muscles in the region 5 glucose prediction model in data fusion layer as a parameter related to glucose It provides direct input and multispectral analysis with anatomical segmentation specific to the facial region. It includes the use of lighting. The investigations revealed that existing systems in the field of non-invasive glucose monitoring technologies, basically a single physical modality (optical spectroscopy only, bioimpedance 10 only) (based on analysis or simply photoplethysmography), this situation is particularly relevant to the user's daily resulting from voluntary or involuntary movements of facial muscles during activities motion artifacts make it difficult to separate them from the signal and reduce measurement accuracy. It appears to limit this. In optical spectroscopy-based systems, glucose molecules are observed to interact with water. It is quite weak compared to other molecules found in skin tissue, such as melanin and hemoglobin. Because it has an optical signature, the signal-to-noise ratio is reduced, and also the skin Changes in temperature, sweating, and differences in blood flow rate affect optical signals. This makes its calibration extremely complex. Based on bioimpedance analysis. In these systems, the impedance signal affects glucose as well as electrolyte balance and body water percentage. Skin moisture and micro-mechanical pressure changes during movement also significantly affect skin moisture. 20 The interference makes it difficult to separate the glucose-derived component. In current solutions... Even if mechanical sensors like accelerometers exist, these sensors only detect signals in optical signals. It is used to filter motion-induced noise and biomechanical vibrations. It is not considered a glucose-related parameter. Furthermore, the existing non- Almost all invasive systems are located in parts of the body such as the wrist, arm, fingertip, or earlobe. It involves taking measurements from specific anatomical regions, particularly those of the face. Segmentation and the association of involuntary microexpressions in this region with glucose signaling. It does not offer any technical solution to this problem. This invention addresses these technical shortcomings. all of it, from involuntary micro-expressions of the muscles in the facial region (emotional reflex arc) The triggered micro-contractions of the orbicularis muscle around the eye, micro-contractions in the nasolabial region 30 biomechanical vibrations (caused by vibrations) and glucose in the subcutaneous interstitial fluid Simultaneous and synchronized optical changes caused by multiple sensors (accelerometer, gyroscope, photoplethysmography sensor, active at 530nm, 660nm, 940nm and 1020nm wavelengths. Capturing these two different images with a multispectral optical reader that provides illumination Data fusion by timestamping physical quantities (mechanical and optical) 35 merging them at the layer and then this fused data, motion artifacts, glucose 6 in a hybrid deep learning architecture trained to distinguish it from the source signal (By processing via Convolutional Neural Network – CNN and Long Short-Term Memory – LSTM) It solves the problem. Thus, the invention is based on neither purely optical nor purely mechanical data. Unlike traditional systems, the noise caused by movement is integrated into the operating principle of the system. by making it a part of it, biomechanical vibrations are used as a glucose-related parameter. 5 It provides direct input to the glucose prediction model, facial region-specific anatomical features. using multispectral illumination with segmentation (nasolabial region, periorbital region) capturing the optical signatures of metabolic markers in interstitial fluid and allowing the user uninterrupted throughout all daily activities such as speaking, facial expressions, and head movements, It offers a highly accurate and clinically significant non-invasive glucose monitoring capability. 10 Ultimately, the problems mentioned above, which cannot be solved with the current technology, are related to the technical aspects. This has made it necessary to make an innovation in the field. Purposes and Brief Description of the Invention The main objective of the invention is to overcome the challenges faced by existing systems in the field of non-invasive glucose monitoring. The fundamental technical problem is that it is based on only one physical modality (optical spectroscopy, 15 Low accuracy and motion in measurements based on bioimpedance analysis or photoplethysmography. The aim is to offer a fundamental solution to the problem of sensitivity to artifacts. In this way, the face subcutaneous tissue through biomechanical vibrations resulting from involuntary micro-expressions of the muscles synchronous and instantaneous optical changes caused by glucose in the interstitial fluid By capturing and fusing it, all 20 features of the user's speech, facial expressions, head movements, etc. uninterrupted, motion artifact-free and high-quality even during daily activities. By providing an accurate glucose measurement, the most critical limitation of the current technique is overcome. Another aim of the invention is to replace traditional non-invasive glucose tolerance testing, which relies solely on optical modality. measurement systems measure glucose molecules in skin proteins such as water, melanin, hemoglobin, and collagen. It has an extremely weak optical signature compared to other molecules found in its tissue. 25 The aim is to overcome the low signal-to-noise ratio problem caused by multispectral technology. lighting unit (active lighting at 530nm, 660nm, 940nm and 1020nm wavelengths) characteristic optical signatures of metabolic markers such as glucose and lactate in interstitial fluid capture, then fusion of this optical data with biomechanical vibration data to create a hybrid. Thanks to processing in a deep learning architecture (CNN-LSTM), the other 30 within the texture Noise originating from molecules is effectively separated, thereby improving the specificity of the glucose signal. and its sensitivity is significantly increased. 7 Another aim of the invention is to replace accelerometers, which are similar to those in existing non-invasive glucose monitoring systems. mechanical sensors only filter motion-induced noise in optical signals Despite being used for this purpose, biomechanical vibrations themselves are associated with glucose. The aim is to present an innovative sensor fusion approach that considers the face as a parameter. from involuntary micro-expressions of the muscles in the region (triggered by the emotional reflex arc 5 (Microcontractions of the orbicularis oculi muscle, microvibrations in the nasolabial region) biomechanical vibrations resulting from high-precision MEMS-based accelerometers and Simultaneous capture with the gyroscope and data fusion of this data with optical data. By processing it in such a way as to directly provide input to the glucose prediction model at the layer, By integrating motion artifacts into the system's operating principle, measurement stability is increased to 10. and repeatability is maximized. Another aim of the invention is to improve almost all existing non-invasive glucose monitoring systems. taking measurements from body parts such as the wrist, arm, fingertip, or earlobe In contrast to this, anatomical segmentation specific to the facial region and involuntary movements in this region... A new technical approach that enables the correlation of facial expressions with glucose signaling. 15 The goal is to develop the nasolabial region (between the side of the nose and the cheek) and the periorbital region (below the eye). (like a lid) it is densely vascularized and the interstitial fluid is closest to the epidermal layer. anatomical regions are automatically segmented using AI-powered facial segmentation algorithms. identification and simultaneous biomechanical and optical measurements in these regions Thanks to this, the anatomical 20 that responds most quickly to glucose changes By enabling measurements from various locations, the system's sensitivity and response time are optimized. is being done. Another aim of the invention is to improve the use of skin temperature in existing non-invasive glucose measurement systems. changes, sweating, differences in blood flow rate, and individual tissue density calibration complexity caused by external factors such as variations and individual 25 The goal is to eliminate the necessary dependence on calibration models. From multiple sensors (photoplethysmography, multispectral optical sensor, accelerometer, gyroscope, temperature sensor, humidity) the data obtained from the sensor is processed in a hybrid deep learning architecture (CNN-LSTM) and By training this architecture on a large population, individual physiological generalizable regardless of differences, requiring no user-specific calibration, and 30 It offers a clinically significant and accurate glucose monitoring option. 8 Another aim of the invention is to provide applications in invasive or minimally invasive glucose monitoring systems. (microneedles, subcutaneous sensors, iontophoresis-based systems) encountered skin barrier loss of user comfort due to overcoming obstacles, biocompatibility issues, sensor wear over time resulting biological surface fouling and interstitial fluid glucose and blood glucose 5 Our goal is to offer a completely non-invasive and real-time glucose monitoring solution for the face area. directly via directed multispectral optical sensors and biomechanical sensors By enabling the real-time detection of metabolic changes in the subcutaneous interstitial fluid, the user does not experience any physical discomfort, skin integrity is preserved and blood is drawn Data fusion and forward signaling of the time delay between glucose and interstitial fluid glucose 10 A viewing experience is provided where processing techniques are minimized. Another aim of the invention is to enable glucose measurement from facial movements using the current technique. Given the lack of any existing system for this, this is an entirely new technical field. The goal is to introduce a groundbreaking innovation in the wearable health technology sector by opening smart glasses. 15 in an augmented reality (AR) glasses or near-face wearable sensor form factor Designed to be applicable, this system has a high degree of integration into daily life. by providing continuous and uninterrupted glucose monitoring that is imperceptible to the user It aims to bring about a revolutionary transformation in diabetes management, and at the same time... significant contribution to the health technology ecosystem through the licensing and commercialization of technology It creates added value. 20 Another aim of the invention is to investigate the activation of the sympathetic nervous system in cases of hypoglycemia. micro-vibrations occurring in the facial muscles (involuntary vibrations in the orbicularis oculi muscle) micro-expressions (micro-contractions in the nasolabial area) and their relationship with glucose levels, artificial intelligence The aim is to use it through a supported model. This relationship is described in the literature during a hypoglycemic attack. Due to increased sympathetic nervous system activation, skeletal muscles (especially the face 25 measurable changes in micro-vibration frequency and amplitude (in the muscles) It is based on the findings. The hybrid deep learning model (CNN-) developed within the scope of the invention. LSTM uses multispectral technology to analyze these biomechanical vibration patterns captured by accelerometers and gyroscopes. It fuses simultaneously with glucose-derived optical signals obtained by an optical reader, Thus, biomechanical changes caused by mood, fatigue, or other external factors 30 It separates the noise from the glucose-induced signal. All the purposes mentioned above and those that will emerge from the detailed explanation below. The present invention aims to achieve this by using biomechanical vibrations in the facial region and subcutaneous tissue. Non-invasive glucose measurement based on the simultaneous fusion of optical changes in interstitial fluid. 9 It relates to the system and method. The system in question will either make contact with the user's face or... Configured to be positioned at close range, containing sensors a portable imaging and sensing unit, the aforementioned portable imaging and located within the sensing unit, from involuntary micro-expressions of facial muscles En 5 detects biomechanical vibrations and triaxial linear acceleration changes caused by vibrations. a small accelerometer, mentioned within the portable imaging and sensing unit. from involuntary micro-expressions of facial muscles, positioned simultaneously with the accelerometer At least one gyroscope that detects the resulting angular velocity changes and orientation information, The aforementioned portable imaging and sensing unit includes the mentioned accelerometer and Simultaneously positioned with the aforementioned gyroscope, video 10 of the user's face At least one high-resolution camera captured the image of the aforementioned portable device. The accelerometer, gyroscope, and other sensors mentioned within the imaging and sensing unit the skin positioned simultaneously with the aforementioned high-resolution camera the change over time in the intensity of light reflected from the surface or scattered back from the subcutaneous tissue at least 15 that detect microcirculation changes in blood volume by sensing a photoplethysmography sensor, the aforementioned portable imaging and detection unit the accelerometer mentioned, the gyroscope mentioned, the high-resolution mentioned 530 positioned simultaneously with the camera and the aforementioned photoplethysmography sensor active at wavelengths of 1020 nanometers, 660 nanometers, 940 nanometers, and 1020 nanometers. By providing illumination, the 20 caused by glucose molecules in the subcutaneous interstitial fluid At least one multispectral optical reader capable of detecting light scattering and absorption variations, the aforementioned accelerometer, the aforementioned gyroscope, the aforementioned high-resolution camera, obtained from the aforementioned photoplethysmography sensor and the aforementioned multispectral optical reader A data collection and pre-processing system that synchronizes the collected raw data by performing timestamp mapping. The processing unit receives synchronized 25 data from the aforementioned data collection and preprocessing unit. A data fusion layer that combines mechanical and optical data to create a common data structure. It takes fused data from the aforementioned data fusion layer as input, Convolutional neural network layers extract spatial features and represent long-term and short-term memory. Predicting glucose levels by processing dynamic changes in time series using layers. a hybrid deep learning model and 30 by the aforementioned hybrid deep learning model The predicted glucose output value is provided to the user as visual, auditory, or tactile feedback. It includes a user interface unit that provides a presentation. The best way to utilize the advantages of the existing invention, together with its structure and additional elements. For it to be understood, it must be considered together with the figures explained below. Brief Description of Figures 35 Figure 1: The subject of the invention is the simultaneous fusion of facial biomechanical and optical signals with non- This is a schematic view of an invasive glucose monitoring system. Reference Numbers 1 - Portable imaging and sensing unit 11 - Accelerometer 5 12 - Gyroscope 13 - Camera 14 - Photoplethysmography sensor - Multispectral optical reader 2 - Data collection and preprocessing unit 10 3 - Data fusion layer 4 - Hybrid deep learning model - User interface unit Detailed description of the invention 15 The invention enables non-invasive glucose measurement through the simultaneous fusion of facial biomechanical and optical signals. related to the system,  In a way that will touch the user's face or be positioned close to the face. a portable imaging and sensing system configured with sensors unit (1), 20  Located within the aforementioned portable imaging and sensing unit (1), biomechanical vibrations resulting from involuntary micro-expressions of facial muscles and at least one accelerometer that detects triaxial linear acceleration changes (11),  The portable imaging and sensing unit (1) mentioned within it Involuntary 25 of the facial muscles positioned simultaneously with the accelerometer (11) angular rotation velocity changes and orientation information resulting from micro-mimics at least one gyroscope detecting (12),  The portable imaging and sensing unit (1) mentioned within it positioned simultaneously with the accelerometer (11) and the aforementioned gyroscope (12), at least one camera (13) that captures a video image of the user's face, 30  The portable imaging and sensing unit (1) mentioned within it accelerometer (11), the aforementioned gyroscope (12) and the aforementioned camera (13) simultaneously positioned, reflected from the skin surface or scattered back from the subcutaneous tissue 11 microcirculation in blood volume by detecting changes in light intensity over time at least one photoplethysmography sensor that detects changes (14),  The portable imaging and sensing unit (1) mentioned within it accelerometer (11), the aforementioned gyroscope (12), the aforementioned camera (13) and the aforementioned 530 nanometer, 5 positioned simultaneously with the photoplethysmography sensor (14) Active illumination at wavelengths of 660 nanometers, 940 nanometers, and 1020 nanometers. by providing light caused by glucose molecules in the subcutaneous interstitial fluid at least one multispectral optical reader capable of detecting scattering and absorption variations (15),  The aforementioned accelerometer (11), the aforementioned gyroscope (12), the aforementioned camera (13), 10 the aforementioned photoplethysmography sensor (14) and the aforementioned multispectral optic by timestamping the raw data obtained from the reader (15) a data collection and preprocessing unit that synchronizes (2),  Synchronized data obtained from the mentioned data collection and preprocessing unit (2) Data fusion is a process that combines mechanical and optical data to create a common data structure. 15 layer (3),  The fused data taken from the mentioned data fusion layer (3) as input The field extracts spatial features using convolutional neural network layers and displays long and short features. by processing dynamic changes on the time series with temporary memory layers a hybrid deep learning model that predicts glucose levels (4) and 20  The glucose prediction value produced by the aforementioned hybrid deep learning model (4) a user interface that provides the user with visual, auditory, or tactile feedback unit (5) It includes. In traditional non-invasive glucose measurement approaches, usually only a single modality is used. (for example, only optical spectroscopy or only biomechanical sensors) are being used, This situation is particularly noticeable during the user's daily activities when facial muscles are used voluntarily or involuntarily. This makes it difficult to separate artifacts resulting from involuntary movements from the signal. and limits the accuracy of measurement. This technical problem is addressed in the development of the invention. Involuntary micro-expressions triggered by the emotional reflex arc (e.g., orbicularis 30 around the eyes) biomechanics created by micro-contractions of the muscle (micro-vibrations in the nasolabial region) Vibration patterns are detected by an integrated, high-precision MEMS-based accelerometer and gyroscope. portable device (smart glasses, near-face wearable sensor or smartphone camera) Simultaneous capture by integrated mechanical sensor array, the same anatomical features at the same time. Multispectral optical readers from regions (530nm, 660nm, 940nm and 1020nm wavelengths) 35 12 (photoplethysmography and tissue reflection spectroscopy with active illumination in their lengths) light scattering caused by glucose molecules in the interstitial fluid and This is overcome by recording the changes in absorption. These two different physical quantities The raw data stream undergoes timestamp mapping in a preprocessor unit located within the device. This is done by synchronizing them and then using Convolutional 5, a hybrid deep learning architecture. A combination of Neural Network (CNN) and Long Short-Term Memory (LSTM) networks. The CNN layer is fused in the model. It transmits optical signals to tissue metabolism. When extracting spatial features, the LSTM layer analyzes the time series of biomechanical vibrations. By processing the dynamic changes in muscle activity, the noise associated with muscle activity is determined by the system. It enables "learning" and differentiation from the actual glucose signal. This synergistic 10 Thanks to this approach, the invention is based on neither purely optical nor purely mechanical sensors. Unlike systems that allow the user to speak, make facial expressions, or move their head in everyday situations... uninterrupted, free from motion artifacts, and high-quality even during activities. It offers an accurate glucose measurement. Thus, the invention provides an alternative to invasive methods. For individuals requiring continuous glucose monitoring (CGM), 15 years of age and free from restricted movement. A critical advancement in current technology by offering a wearable and clinically accurate monitoring solution. It records. The invention relates to biomechanical vibrations in the facial region and optical vibrations in the subcutaneous interstitial fluid. Non-invasive glucose monitoring system based on the simultaneous fusion of changes, working principle. in terms of simultaneous data obtained from multiple modalities (mechanical and optical) 20 collection, synchronization, fusion, and in a hybrid deep learning model It is based on the principle of estimating glucose levels by processing data. The system allows the user to... portable imaging devices that will touch the face or are positioned in close proximity to the face, and by the simultaneous activation of the sensors located in the sensing unit (1) is started. At this stage, the accelerometer (11) detects involuntary micro-expressions of the facial muscles (orbicularis 25 micro-contractions of the oculi muscle (micro-vibrations in the nasolabial region) gyroscope detects biomechanical vibrations and triaxial linear acceleration changes. (12) angular rotation velocity changes and orientation information related to the same muscle activities They detect simultaneously. Data from these two mechanical sensors analyzes the user's face. It creates a comprehensive biomechanical profile of their movements. At the same time, preferably a high 30 high-resolution camera (13) captures video image of the user's face When capturing, the photoplethysmography sensor (14) captures the light reflected from the skin surface or from the subcutaneous tissue. microcirculation in blood volume by detecting changes in the intensity of backscattered light over time It detects changes. However, the multispectral optical reader (15) detects changes at 530 nanometers. Active illumination at wavelengths of 660 nanometers, 940 nanometers, and 1020 nanometers 35 by providing light caused by glucose molecules in the subcutaneous interstitial fluid 13 It detects scattering and absorption changes. The portable imaging and sensing mentioned... all these sensors (accelerometer (11), gyroscope (12), camera (13), inside unit (1), raw material obtained from photoplethysmography sensor (14) and multispectral optical reader (15)) data is timestamped by the data collection and preprocessing unit (2) This synchronization process ensures that mechanical and optical data are synchronized within the same time frame. by ensuring that it is aligned in such a way that it will belong to the same place, the accuracy of the subsequent fusion process. This ensures that the synchronized data is received by the data fusion layer (3) and mechanical (accelerometer (11) and gyroscope (12) data) and optical (high-resolution camera (13), two photoplethysmography sensors (14) and multispectral optical readers (15) data) Data from different physical quantities are combined under a common data structure. This fusion process is 10 This is critical for differentiating motion artifacts from glucose-induced signaling. The fused data generated by the aforementioned data fusion layer (3) is hybrid deep The learning model (4) is taken as input by the learning model. This model is with convolutional neural network layers. spatial features within fused data (e.g., specific features in the face region) (Optical signatures of anatomical areas) while extracting layers of long and short-term memory and time 15 dynamic changes on the series (e.g., the time variation of biomechanical vibrations and It predicts glucose levels by processing changes in the glucose signaling pathway. This prediction During the process, the hybrid deep learning model (4) uses only accelerometer (11) and gyroscope (12) data. Not for noise filtering purposes, but directly estimated as a parameter related to glucose. It evaluates the model in a way that will provide input. Finally, the aforementioned hybrid deep learning 20 The glucose prediction value produced by the model (4) is displayed by the user interface unit (5). This information is collected and presented to the user as visual, auditory, or tactile feedback. In this way, during all of the user's daily activities such as speaking, facial expressions, and head movements A continuous, motion artifact-free, and highly accurate non-invasive glucose tolerance test. A viewing experience is provided. 25

Claims

14 REQUESTS 1. Micro-expressions of facial muscles and emotional reflexes that occur during these processes. biomechanical vibrations and glucose-induced optical effects in the subcutaneous interstitial fluid Simultaneous capture of changes, a deep learning-based hybrid neural network By fusing it with, non-invasive, continuous and uninterrupted blood sugar monitoring even during movement 5 It is a glucose measurement system that enables glucose measurement; its feature is:  It will make contact with the user's face or be positioned close to the face. a portable device configured in this way and containing sensors imaging and sensing unit (1),  Contained within the aforementioned portable imaging and sensing unit (1), 10 biomechanics resulting from involuntary micro-expressions of facial muscles at least capable of detecting vibrations and triaxial linear acceleration changes. an accelerometer (11),  Contained within the aforementioned portable imaging and sensing unit (1) and working simultaneously with the aforementioned accelerometer (11), the facial muscles 15 angular rotational velocity changes resulting from involuntary micro-expressions and At least one gyroscope that detects orientation information (12),  Contained within the aforementioned portable imaging and sensing unit (1) and simultaneously with the aforementioned accelerometer (11) and the aforementioned gyroscope (12) the employee has at least one camera that captures a video image of the user's face 20 (13),  Contained within the aforementioned portable imaging and sensing unit (1) and the aforementioned accelerometer (11), the aforementioned gyroscope (12) and the aforementioned camera (13) working simultaneously with, reflected from the skin surface or subcutaneous By detecting the time-dependent change in the intensity of light scattered back from the tissue, the blood... 25 detecting at least changes in microcirculation in volume a photoplethysmography sensor (14),  Contained within the aforementioned portable imaging and sensing unit (1) and the aforementioned accelerometer (11), the aforementioned gyroscope (12), the aforementioned camera (13) and working simultaneously with the aforementioned photoplethysmography sensor (14), 30 by providing active illumination at various nanometer wavelengths under the skin Light scattering caused by glucose molecules in the interstitial fluid and At least one multispectral optical reader that detects absorption changes (15),  The mentioned accelerometer (11), the mentioned gyroscope (12), the mentioned camera (13), the mentioned photoplethysmography sensor (14) and the mentioned multispectral optic 35 by timestamping the raw data obtained from the reader (15) a data collection and preprocessing unit that synchronizes (2),  Synchronized data obtained from the mentioned data collection and preprocessing unit (2) a data that combines mechanical and optical data to create a common data structure fusion layer (3), 5  Input the fused data taken from the mentioned data fusion layer (3). It takes as its field, extracts spatial features using convolutional neural network layers, and Dynamics on time series with long and short-term memory layers a hybrid deep learning that predicts glucose levels by processing changes model (4) 10 It is characterized by its inclusion.

2. The system is compliant with claim 1 and its feature is the aforementioned hybrid deep learning model (4) The glucose forecast produced is returned to the user visually, audibly, or tactilely. It includes a user interface unit (5) that presents the notification.

3. The system is compliant with Claim 1 and its feature is that the mentioned camera (13) has a high resolution 15 It is the fact that.

4. The system is compliant with Claim 1 and its feature is the aforementioned multispectral optical reader (15) active illumination is provided at wavelengths of 530 nanometers and / or 660 nanometers. and / or 940 nanometers and / or 1020 nanometers.