Pub Date : 2025-11-07DOI: 10.1109/LSENS.2025.3630485
Mohannad K Sabir;Bashar S. Falih;Łukasz Gierz;Aymen Saad;Mohammed Ahmed Subhi;Montadar Abas Taher
European bee-eaters (Merops genus) pose significant challenges to beekeepers by preying on worker bees, reducing hive productivity. In this letter, a new approach for European bee-eater sound recognition employing convolutional neural networks (CNNs) based on classically trained classification models is presented. The short-time Fourier transform computes the time–frequency representation of the bird sounds, which acts as input to CNNs. The precision of the classifier was confirmed over 1000 spectrogram images per bird species and done on 11 families. The proposed method obtained 98.45% accuracy for the 11 bird species and 100% for identifying bee-eater sounds. The resultant algorithm could be applied on a small, minicomputer type of device such as Raspberry Pi, with an incorporated frightening function for beekeepers, which helps in preserving their hives and harvesting more honey.
{"title":"Deep and Machine Learning-Based Detection of European Bee-Eaters Using Bird Sounds","authors":"Mohannad K Sabir;Bashar S. Falih;Łukasz Gierz;Aymen Saad;Mohammed Ahmed Subhi;Montadar Abas Taher","doi":"10.1109/LSENS.2025.3630485","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3630485","url":null,"abstract":"European bee-eaters (<italic>Merops</i> genus) pose significant challenges to beekeepers by preying on worker bees, reducing hive productivity. In this letter, a new approach for European bee-eater sound recognition employing convolutional neural networks (CNNs) based on classically trained classification models is presented. The short-time Fourier transform computes the time–frequency representation of the bird sounds, which acts as input to CNNs. The precision of the classifier was confirmed over 1000 spectrogram images per bird species and done on 11 families. The proposed method obtained 98.45% accuracy for the 11 bird species and 100% for identifying bee-eater sounds. The resultant algorithm could be applied on a small, minicomputer type of device such as Raspberry Pi, with an incorporated frightening function for beekeepers, which helps in preserving their hives and harvesting more honey.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"10 2","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11234890","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146082226","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-06DOI: 10.1109/LSENS.2025.3630120
Pi-Yun Chen;Chun-Yu Lin;Ping-Tzan Huang;Neng-Sheng Pai;Chao-Lin Kuo;Chien-Ming Li;Chia-Hung Lin
Clinical assessment methods for Parkinson's disease (PD) commonly rely on the Movement Disorder Society-Unified Parkinson's Disease Rating Scale and the Health-Related Quality of Life questionnaire. Both methods employ structured question-and-answer assessments to evaluate the severity and progression of patients with related PD by assessing the nonmotor and motor experiences, movement disorders, and motor complications, along with complications of therapy. However, these methods need face-to-face interaction and are time-consuming (typically taking >20 min). Moreover, the assessment outcomes are often influenced by the clinician's expertise and subjective judgments. In addition, these methods also lack the capability to objectively and automatically quantify both tremor severity level and tremor classification in PD patients. To overcome the aforementioned limitations, this letter intends to implement a W-band (76–81 GHz) millimeter-wave-based noncontact biosensor that extracts the echo features for upper limb tremor classification. A deep learning method, cascade convolutional neural network-based classifier with combined feature extraction and pattern recognition tasks, is employed to identify tremor feature patterns for distinguishing typical tremor frequencies among low-frequency (<4.0 Hz), medium-frequency (4.0–7.0 Hz), and high-frequency (>7.0 Hz) tremors through short-range (<1.0 m) and noncontact measurements.
{"title":"W-Band Millimeter-Wave Echo Features Detection With Cascade CNN-Based Classifier for Parkinson's Disease Tremors Classification","authors":"Pi-Yun Chen;Chun-Yu Lin;Ping-Tzan Huang;Neng-Sheng Pai;Chao-Lin Kuo;Chien-Ming Li;Chia-Hung Lin","doi":"10.1109/LSENS.2025.3630120","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3630120","url":null,"abstract":"Clinical assessment methods for Parkinson's disease (PD) commonly rely on the Movement Disorder Society<bold>-</b>Unified Parkinson's Disease Rating Scale and the Health-Related Quality of Life questionnaire. Both methods employ structured question-and-answer assessments to evaluate the severity and progression of patients with related PD by assessing the nonmotor and motor experiences, movement disorders, and motor complications, along with complications of therapy. However, these methods need face-to-face interaction and are time-consuming (typically taking >20 min). Moreover, the assessment outcomes are often influenced by the clinician's expertise and subjective judgments. In addition, these methods also lack the capability to objectively and automatically quantify both tremor severity level and tremor classification in PD patients. To overcome the aforementioned limitations, this letter intends to implement a W-band (76–81 GHz) millimeter-wave-based noncontact biosensor that extracts the echo features for upper limb tremor classification. A deep learning method, cascade convolutional neural network-based classifier with combined feature extraction and pattern recognition tasks, is employed to identify tremor feature patterns for distinguishing typical tremor frequencies among low-frequency (<italic><</i>4.0 Hz), medium-frequency (4.0–7.0 Hz), and high-frequency (>7.0 Hz) tremors through short-range (<1.0 m) and noncontact measurements.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145560707","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Most commercial optocouplers integrate at least two optical elements, such as a light-emitting diode (LED) and a silicon photodiode (PD), which function as a simple signal switch. However, this hybrid approach not only makes high-level device integration difficult but also increases fabrication complexity and decreases reliability. To achieve a compact, high-performance optocoupler, this study integrates an LED and a PD on a sapphire-based gallium nitride (GaN) epi-wafer into a single chip using a monolithic microfabrication process. The design involves patterning an annular interdigitated microstructure in which the LED is surrounded by the PD. This method is suitable for batch fabrication and enhances coupling efficiency by enlarging the active area via the annular and interdigitated structures. Measurement results revealed that the proposed chip with annular interdigitated structures generated a photocurrent of 0.176 mA when an 80 mA current was applied to the emitting element. A high current transfer ratio of 0.23% was achieved, indicating excellent performance. In addition, the proposed optocoupler requires fewer PDs, thereby reducing chip size and simplifying packaging.
{"title":"An Improved Monolithic GaN-Based Optocoupler With Annular Interdigitated Microstructures","authors":"Jhihfong Liou;Yuwei Chen;Huiqi Xie;Shengyung Wang;Chengshiun Liou;Chingfu Tsou","doi":"10.1109/LSENS.2025.3629077","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3629077","url":null,"abstract":"Most commercial optocouplers integrate at least two optical elements, such as a light-emitting diode (LED) and a silicon photodiode (PD), which function as a simple signal switch. However, this hybrid approach not only makes high-level device integration difficult but also increases fabrication complexity and decreases reliability. To achieve a compact, high-performance optocoupler, this study integrates an LED and a PD on a sapphire-based gallium nitride (GaN) epi-wafer into a single chip using a monolithic microfabrication process. The design involves patterning an annular interdigitated microstructure in which the LED is surrounded by the PD. This method is suitable for batch fabrication and enhances coupling efficiency by enlarging the active area via the annular and interdigitated structures. Measurement results revealed that the proposed chip with annular interdigitated structures generated a photocurrent of 0.176 mA when an 80 mA current was applied to the emitting element. A high current transfer ratio of 0.23% was achieved, indicating excellent performance. In addition, the proposed optocoupler requires fewer PDs, thereby reducing chip size and simplifying packaging.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145560701","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-03DOI: 10.1109/LSENS.2025.3627849
Felipe Hornung;Walter O. C. Flores;Katia Christina Zuffellato-Ribas;André Eugenio Lazzaretti;Marcia Muller;José Luís Fabris
Nanotechnology has been increasingly applied in agriculture to optimize crop performance. By combining irrigation with nanoparticles and appropriate lighting, plant development can be improved. This work shows that when the lighting overlaps the plasmonic resonances of silver and gold nanoparticles, tomato leaves exhibit higher chlorophyll content than under nonselective broadband lighting. Whereas chlorophyll can be quantified via destructive assays, the nondestructive method proposed in this work uses deep learning regression to estimate chlorophyll directly from reflectance spectroscopy of tomato leaves. This methodology avoids pigment extraction and tissue damage, being a more suitable tool for field applications. The deep neural network trained with leaf reflectance spectra from 400 to 800 nm achieved R$^{2}$ = 0.8925 for chlorophyll estimation. These findings can pave the way to increase crop yield, with optimized conditions through precision agriculture.
{"title":"Optical Sensing of Chlorophyll Content in Tomato Plants Exposed to Metal Nanoparticles Under Selective Lighting","authors":"Felipe Hornung;Walter O. C. Flores;Katia Christina Zuffellato-Ribas;André Eugenio Lazzaretti;Marcia Muller;José Luís Fabris","doi":"10.1109/LSENS.2025.3627849","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3627849","url":null,"abstract":"Nanotechnology has been increasingly applied in agriculture to optimize crop performance. By combining irrigation with nanoparticles and appropriate lighting, plant development can be improved. This work shows that when the lighting overlaps the plasmonic resonances of silver and gold nanoparticles, tomato leaves exhibit higher chlorophyll content than under nonselective broadband lighting. Whereas chlorophyll can be quantified via destructive assays, the nondestructive method proposed in this work uses deep learning regression to estimate chlorophyll directly from reflectance spectroscopy of tomato leaves. This methodology avoids pigment extraction and tissue damage, being a more suitable tool for field applications. The deep neural network trained with leaf reflectance spectra from 400 to 800 nm achieved R<inline-formula><tex-math>$^{2}$</tex-math></inline-formula> = 0.8925 for chlorophyll estimation. These findings can pave the way to increase crop yield, with optimized conditions through precision agriculture.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145510187","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-31DOI: 10.1109/LSENS.2025.3627636
David V. Thiel;Anish Kumar;Krishnasamy T. Selvan;Hugo G. Espinosa
There has been significant global interest in the use of nonlethal methods to repel sharks during ocean-based activities. Given that sharks possess an electrosensory system for detecting prey, quasi-static electric fields were investigated as a potential wearable deterrent. A series of controlled experiments were conducted in a water tank (900 × 400 × 400 mm3) using a pulsed electric field (PEF) generator (5000 V at 8.5 kHz, 10 μs pulsewidth), with three conductivity values based on different salinity concentrations: 0.059, 0.149, and 1.042 S/m. The arc distance was approximately 1 mm, and the detector consisted of a germanium diode in parallel with a 130 kΩ resistor feeding a digital voltmeter. All equipment was battery-powered to minimize cable induction effects. The transmitter and receiver were enclosed in waterproof plastic bags under 40 mm of water. These data were fitted to a log–log power law (slope = −1.85, r2 = 0.96). The received voltage power law was less than the theoretical prediction from the geophysical resistivity method (slope = −3.0), likely due to side reflections in the water tank. Water conductivity had a minimal effect on the results, suggesting the findings are representative of saline water conditions. Given the small, portable, and insulated nature of the equipment, it is feasible to extrapolate the electric field strength at a distance in open water for potential shark-deterrent applications. Unlike permanent magnets, electric signals can be easily manipulated to minimize shark habituation.
{"title":"Toward an Electric Shark Deterrent: Electric Field Attenuation in Saline Water","authors":"David V. Thiel;Anish Kumar;Krishnasamy T. Selvan;Hugo G. Espinosa","doi":"10.1109/LSENS.2025.3627636","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3627636","url":null,"abstract":"There has been significant global interest in the use of nonlethal methods to repel sharks during ocean-based activities. Given that sharks possess an electrosensory system for detecting prey, quasi-static electric fields were investigated as a potential wearable deterrent. A series of controlled experiments were conducted in a water tank (900 × 400 × 400 mm<sup>3</sup>) using a pulsed electric field (PEF) generator (5000 V at 8.5 kHz, 10 μs pulsewidth), with three conductivity values based on different salinity concentrations: 0.059, 0.149, and 1.042 S/m. The arc distance was approximately 1 mm, and the detector consisted of a germanium diode in parallel with a 130 kΩ resistor feeding a digital voltmeter. All equipment was battery-powered to minimize cable induction effects. The transmitter and receiver were enclosed in waterproof plastic bags under 40 mm of water. These data were fitted to a log–log power law (slope = −1.85, <italic>r</i><sup>2</sup> = 0.96). The received voltage power law was less than the theoretical prediction from the geophysical resistivity method (slope = −3.0), likely due to side reflections in the water tank. Water conductivity had a minimal effect on the results, suggesting the findings are representative of saline water conditions. Given the small, portable, and insulated nature of the equipment, it is feasible to extrapolate the electric field strength at a distance in open water for potential shark-deterrent applications. Unlike permanent magnets, electric signals can be easily manipulated to minimize shark habituation.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145510186","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-31DOI: 10.1109/LSENS.2025.3626751
Fangfang Zhang;Hao Sun;Jinzhu Peng;Haijing Wang
Millimeter-wave radar is widely used for indoor human activity recognition due to its privacy-preserving nature, with point cloud data effectively capturing target geometry. However, the sparsity and dynamic nature of these point clouds leads to unstable feature extraction, and resource constraints challenge large-scale neural network deployment. To address this, this letter proposes the Lightweight PointNet-BiLSTM with SE-Net (LPBS-Net), a lightweight network integrating a squeeze-and-excitation (SE) attention mechanism and bidirectional long short-term memory (BiLSTM) into a streamlined PointNet backbone, enhancing spatiotemporal feature modeling for dynamic point clouds. To overcome PointNet's need for fixed input point counts and its sensitivity to sparse distributions, we introduce Gaussian-based intensity and repeat padding, which selects base points by reflection intensity and uses Gaussian perturbation and repeated sampling to mitigate sparsity-induced feature degradation. Experiments on two public datasets show that LPBS-Net achieves 97.11% accuracy on the MMActivity dataset with only 0.176 M parameters, reducing model size by 84% compared to PointNet-BiLSTM, and outperforming existing methods, with maximum accuracy improvements exceeding 30%. The proposed lightweight network offers high accuracy and computational efficiency, evidenced by its low parameter count and floating point operations (FLOPs), making it suitable for deployment on resource-constrained edge devices.
{"title":"LPBS-Net: A Lightweight Network for Human Activity Recognition From Sparse Millimeter-Wave Radar Point Clouds","authors":"Fangfang Zhang;Hao Sun;Jinzhu Peng;Haijing Wang","doi":"10.1109/LSENS.2025.3626751","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3626751","url":null,"abstract":"Millimeter-wave radar is widely used for indoor human activity recognition due to its privacy-preserving nature, with point cloud data effectively capturing target geometry. However, the sparsity and dynamic nature of these point clouds leads to unstable feature extraction, and resource constraints challenge large-scale neural network deployment. To address this, this letter proposes the Lightweight PointNet-BiLSTM with SE-Net (LPBS-Net), a lightweight network integrating a squeeze-and-excitation (SE) attention mechanism and bidirectional long short-term memory (BiLSTM) into a streamlined PointNet backbone, enhancing spatiotemporal feature modeling for dynamic point clouds. To overcome PointNet's need for fixed input point counts and its sensitivity to sparse distributions, we introduce Gaussian-based intensity and repeat padding, which selects base points by reflection intensity and uses Gaussian perturbation and repeated sampling to mitigate sparsity-induced feature degradation. Experiments on two public datasets show that LPBS-Net achieves 97.11% accuracy on the MMActivity dataset with only 0.176 M parameters, reducing model size by 84% compared to PointNet-BiLSTM, and outperforming existing methods, with maximum accuracy improvements exceeding 30%. The proposed lightweight network offers high accuracy and computational efficiency, evidenced by its low parameter count and floating point operations (FLOPs), making it suitable for deployment on resource-constrained edge devices.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145560705","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-30DOI: 10.1109/LSENS.2025.3627309
Wenbo Pan;Zhiwei Chen;Xianghua Fan
Accurate 3-D reconstruction and localization are essential for enhancing the sensing capabilities of vehicle-mounted monocular camera systems in intelligent transportation and smart city sensor networks. Existing 3-D Gaussian splatting (GS)-based methods, designed for small-scale indoor sensing, often fail in large outdoor roadside environments due to limited depth cues and motion variability. This letter presents an adaptive GS simultaneous localization and mapping (GS-SLAM) framework that directly improves monocular sensor-based localization and perception in complex outdoor scenarios. A temporally consistent structure predictor enables robust pose estimation without additional depth sensors, while a differentiable joint optimization integrates 3-D Gaussian rendering with feature-guided pose refinement to enhance geometric consistency. A viewpoint-aware learning rate scheduler further stabilizes tracking under varying vehicle motions. Experimental results on the Waymo dataset and vehicle-mounted tests demonstrate significant improvements in sensor-based localization and 3-D environment reconstruction accuracy over existing monocular SLAM systems, offering a scalable and efficient solution for real-time roadside mapping.
{"title":"Gaussian Splatting SLAM for Enhanced Monocular Vehicle Sensor Localization and Roadside Scene Reconstruction","authors":"Wenbo Pan;Zhiwei Chen;Xianghua Fan","doi":"10.1109/LSENS.2025.3627309","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3627309","url":null,"abstract":"Accurate 3-D reconstruction and localization are essential for enhancing the sensing capabilities of vehicle-mounted monocular camera systems in intelligent transportation and smart city sensor networks. Existing 3-D Gaussian splatting (GS)-based methods, designed for small-scale indoor sensing, often fail in large outdoor roadside environments due to limited depth cues and motion variability. This letter presents an adaptive GS simultaneous localization and mapping (GS-SLAM) framework that directly improves monocular sensor-based localization and perception in complex outdoor scenarios. A temporally consistent structure predictor enables robust pose estimation without additional depth sensors, while a differentiable joint optimization integrates 3-D Gaussian rendering with feature-guided pose refinement to enhance geometric consistency. A viewpoint-aware learning rate scheduler further stabilizes tracking under varying vehicle motions. Experimental results on the Waymo dataset and vehicle-mounted tests demonstrate significant improvements in sensor-based localization and 3-D environment reconstruction accuracy over existing monocular SLAM systems, offering a scalable and efficient solution for real-time roadside mapping.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778181","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-30DOI: 10.1109/LSENS.2025.3626822
Parthesh Patil;Sangeeta Palekar;Jayu Kalambe
Monitoring serum calcium levels is essential, as deviation from normal levels can disrupt body functions and can cause severe health issues. Due to the high prevalence of calcium-related disorders in India, there is an urgent need for the development of low-cost, portable, easily accessible calcium testing devices that enable early detection and timely treatment. To address the issue, this study presents a portable calcium-sensing device that combines principles of colorimetric chemistry, image processing, and machine learning to measure calcium concentration. The proposed device uses the Arsenazo III reagent, which changes color from blue to purple on an increase in calcium concentration. The sensing platform is designed as a custom 3-D-printed controlled light enclosure for capturing reproducible and consistent images of Arsenazo III-calcium solutions. For calcium measurement, color features were extracted from multiple color space models (red, green, and blue; hue, saturation, value; and Lab), and then these features were mapped to calcium concentrations using supervised regression algorithms. Feature selection was applied to identify the three most effective predictors, which simplified the model without affecting accuracy. By evaluating various models with standard metrics, it was found that ensemble-based models, particularly random forest (R2 = 0.9979) and gradient boosting (R2 = 0.9962), performed better than other models. The developed calcium-sensing platform is highly sensitive, capable of detecting calcium levels as low as 1 mg/dL (limit of detection), and works effectively across a clinically relevant range of 1–20 mg/dL. Validation tests were also performed comparing the proposed device with a commercial biochemistry analyzer. These tests showed strong agreement, confirming the potential of the proposed system for accurate calcium prediction.
{"title":"Smartphone-Based Portable Calcium-Sensing Platform Using Image Processing and Machine Learning","authors":"Parthesh Patil;Sangeeta Palekar;Jayu Kalambe","doi":"10.1109/LSENS.2025.3626822","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3626822","url":null,"abstract":"Monitoring serum calcium levels is essential, as deviation from normal levels can disrupt body functions and can cause severe health issues. Due to the high prevalence of calcium-related disorders in India, there is an urgent need for the development of low-cost, portable, easily accessible calcium testing devices that enable early detection and timely treatment. To address the issue, this study presents a portable calcium-sensing device that combines principles of colorimetric chemistry, image processing, and machine learning to measure calcium concentration. The proposed device uses the Arsenazo III reagent, which changes color from blue to purple on an increase in calcium concentration. The sensing platform is designed as a custom 3-D-printed controlled light enclosure for capturing reproducible and consistent images of Arsenazo III-calcium solutions. For calcium measurement, color features were extracted from multiple color space models (red, green, and blue; hue, saturation, value; and Lab), and then these features were mapped to calcium concentrations using supervised regression algorithms. Feature selection was applied to identify the three most effective predictors, which simplified the model without affecting accuracy. By evaluating various models with standard metrics, it was found that ensemble-based models, particularly random forest (<italic>R</i><sup>2</sup> = 0.9979) and gradient boosting (<italic>R</i><sup>2</sup> = 0.9962), performed better than other models. The developed calcium-sensing platform is highly sensitive, capable of detecting calcium levels as low as 1 mg/dL (limit of detection), and works effectively across a clinically relevant range of 1–20 mg/dL. Validation tests were also performed comparing the proposed device with a commercial biochemistry analyzer. These tests showed strong agreement, confirming the potential of the proposed system for accurate calcium prediction.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145510184","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In a recent study, we introduced a fully bioresorbable Mo + MoOx thin-film electrode for tissue health monitoring, microfabricated using a two-step sputtering process. We demonstrated its enhanced short-term electrochemical stability (<24>x electrode annealed at 450 °C, under physiologically relevant conditions: in simulated biofluids, under 6% O2 and at body temperature. For comparison, control experiments were conducted in phosphate buffer solution at both 6% and 21% O2. The electrodes tested in physiologically mimicking conditions exhibited superior longevity, likely due to first, the lower solubility of Mo compounds in O2-deficient environments, and second, differences in composition between buffer solution and simulated biofluids, playing an important role in corrosion mitigation. These results underscore the dual importance of postdeposition annealing as a strategy to fine-tune the electrochemical stability, in a solution, of thin-film electrodes and of evaluating the said stability under conditions that better mimic the intended physiological environment.
{"title":"Thermal Annealing as a Key Strategy for Enhancing the Electrochemical Stability of Fully Bioresorbable Mo and MoOx Electrodes in Physiologically Mimicking Conditions","authors":"Catarina Fernandes;Anna Altafin;Filippo Franceschini;Irene Taurino","doi":"10.1109/LSENS.2025.3626966","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3626966","url":null,"abstract":"In a recent study, we introduced a fully bioresorbable Mo + MoO<sub>x</sub> thin-film electrode for tissue health monitoring, microfabricated using a two-step sputtering process. We demonstrated its enhanced short-term electrochemical stability (<24>x</sub> electrode annealed at 450 °C, under physiologically relevant conditions: in simulated biofluids, under 6% O<sub>2</sub> and at body temperature. For comparison, control experiments were conducted in phosphate buffer solution at both 6% and 21% O<sub>2</sub>. The electrodes tested in physiologically mimicking conditions exhibited superior longevity, likely due to first, the lower solubility of Mo compounds in O<sub>2</sub>-deficient environments, and second, differences in composition between buffer solution and simulated biofluids, playing an important role in corrosion mitigation. These results underscore the dual importance of postdeposition annealing as a strategy to fine-tune the electrochemical stability, in a solution, of thin-film electrodes and of evaluating the said stability under conditions that better mimic the intended physiological environment.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145510219","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Image contrast is a critical factor for machine vision tasks. A promising approach for enhancing contrast involves the use of algorithmically optimized, spectrally tunable illumination. However, the very definition of “contrast” is often rooted in principles of human perception, which may not be optimal for a machine observer. For an algorithm, contrast is an objective, task-driven metric that can be mathematically defined. To investigate the impact of this definition, we first use eigenvalue-based optimization algorithms to compute optimal illumination spectra. We then systematically evaluate these spectra using four distinct, physically realizable contrast formulations. Our analysis reveals that the performance of a given optimization algorithm is entirely dependent on the subsequent choice of evaluation metric. An illumination spectrum considered optimal under one metric can be significantly suboptimal when measured by another. This demonstrates that the choice of contrast metric is not a passive measurement, but an active design parameter with tangible physical consequences. From a machine learning perspective, the choice of this “loss function” should be codesigned with the physical hardware and the ultimate downstream task to achieve true system-level optimization.
{"title":"Rethinking the Concept of Pixel Intensity Contrast From a Machine Learning Perspective","authors":"Sanush Abeysekera;Melanie Po-Leen Ooi;Ye-Chow Kuang;Shah Faisal;Yaminn Thawdar;Geoffrey Holmes;Dale Fletcher;Peter Reutemann","doi":"10.1109/LSENS.2025.3627240","DOIUrl":"https://doi.org/10.1109/LSENS.2025.3627240","url":null,"abstract":"Image contrast is a critical factor for machine vision tasks. A promising approach for enhancing contrast involves the use of algorithmically optimized, spectrally tunable illumination. However, the very definition of “contrast” is often rooted in principles of human perception, which may not be optimal for a machine observer. For an algorithm, contrast is an objective, task-driven metric that can be mathematically defined. To investigate the impact of this definition, we first use eigenvalue-based optimization algorithms to compute optimal illumination spectra. We then systematically evaluate these spectra using four distinct, physically realizable contrast formulations. Our analysis reveals that the performance of a given optimization algorithm is entirely dependent on the subsequent choice of evaluation metric. An illumination spectrum considered optimal under one metric can be significantly suboptimal when measured by another. This demonstrates that the choice of contrast metric is not a passive measurement, but an active design parameter with tangible physical consequences. From a machine learning perspective, the choice of this “loss function” should be codesigned with the physical hardware and the ultimate downstream task to achieve true system-level optimization.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":"9 12","pages":"1-4"},"PeriodicalIF":2.2,"publicationDate":"2025-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11222939","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145560703","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：