Early-Stage Gas Identification Using Convolutional Long Short-Term Neural Network with Sensor Array Time Series Data

Zhou, Kai; Liu, Yixin

doi:10.3390/s21144826

Cited by 15 publications

(13 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Consequently, machine learning has significantly improved biosensors, such as through the analysis of sensing data for anomaly detection, noise reduction, classification, and pattern recognition [ 13 ]. Identifying ultra-low levels of biological species is a critical objective in improving biosensors in medical diagnostics and therapy [ 14 ].…”

Section: Introductionmentioning

confidence: 99%

Predicting Analyte Concentrations from Electrochemical Aptasensor Signals Using LSTM Recurrent Networks

et al. 2022

View full text Add to dashboard Cite

Nanomaterial-based aptasensors are useful devices capable of detecting small biological species. Determining suitable signal processing methods can improve the identification and quantification of target analytes detected by the biosensor and consequently improve the biosensor’s performance. In this work, we propose a data augmentation method to overcome the insufficient amount of available original data and long short-term memory (LSTM) to automatically predict the analyte concentration from part of a signal registered by three electrochemical aptasensors, with differences in bioreceptors, analytes, and the signals’ lengths for specific concentrations. To find the optimal network, we altered the following variables: the LSTM layer structure (unidirectional LSTM (LSTM) and bidirectional LSTM (BLSTM)), optimizers (Adam, RMSPROP, SGDM), number of hidden units, and amount of augmented data. Then, the evaluation of the networks revealed that the highest original data accuracy increased from 50% to 92% by exploiting the data augmentation method. In addition, the SGDM optimizer showed a lower performance prediction than that of the ADAM and RMSPROP algorithms, and the number of hidden units was ineffective in improving the networks’ performances. Moreover, the BLSTM nets showed more accurate predictions than those of the ULSTM nets on lengthier signals. These results demonstrate that this method can automatically detect the analyte concentration from the sensor signals.

show abstract

Section: Introductionmentioning

confidence: 99%

Predicting Analyte Concentrations from Electrochemical Aptasensor Signals Using LSTM Recurrent Networks

et al. 2022

View full text Add to dashboard Cite

show abstract

“…To demonstrate the accuracy to distinguish between the normal tissue and different grades of glioma tissue via the Vis–NIR photodetectors, the machine learning method is adopted to explore the recognition accuracy under single-wavelength detection and multiwavelength detection. The support vector machine (SVM) was chosen because of its reliability and accuracy in small sample data set training. , The data were collected by recording the photocurrents of the photodetectors under 5 V bias as typical wavelengths of light passing vertically through various regions of different tissue sections. Photocurrent data were further normalized by , where is the photocurrent obtained by incident light passing through the tissue area and is the photocurrent obtained by incident light passing through the blank area of the sections (Table S1).…”

Section: Resultsmentioning

confidence: 99%

“…The support vector machine (SVM) was chosen because of its reliability and accuracy in small sample data set training. 52,53 The data were collected by recording the photocurrents of the photodetectors under 5 V bias as typical wavelengths of light passing vertically through various regions of different tissue sections. Photocurrent data were further normalized by I I / effective control , where I effective is the photocurrent obtained by incident light passing through the tissue area and I control is the photocurrent obtained by incident light passing through the blank area of the sections (Table S1).…”

Section: Phmentioning

confidence: 99%

Lateral Heterostructured Vis–NIR Photodetectors with Multimodal Detection for Rapid and Precise Classification of Glioma

Xie

Pan

et al. 2022

ACS Nano

View full text Add to dashboard Cite

Precise diagnosis of the boundary and grade of tumors is especially important for surgical dissection. Recently, visible and near-infrared (Vis–NIR) absorption differences of tumors are demonstrated for a precise tumor diagnosis. Here, a template-assisted sequential printing strategy is investigated to construct lateral heterostructured Vis–NIR photodetectors, relying on the up-conversion nanoparticles (UCNPs)/perovskite arrays. Under the sequential printing process, the synergistic effect and co-confinement are demonstrated to induce the UCNPs to cover both sides of the perovskite microwire. The side-wrapped lateral heterogeneous UCNPs/perovskite structure exhibits more satisfactory responsiveness to Vis–NIR light than the common fully wrapped structure, due to sufficient visible-light-harvesting ability. The Vis–NIR photodetectors with R reaching 150 mA W–1 at 980 nm and 1084 A W–1 at 450 nm are employed for the rapid classification of glioma. The detection accuracy rate of 99.3% is achieved through a multimodal analysis covering the Vis–NIR light, which provides a reliable basis for glioma grade diagnosis. This work provides a concrete example for the application of photodetectors in tumor detection and surgical diagnosis.

show abstract

“…For the classification task, we implemented a softmargin SVM model with a linear kernel. [46] The whole model was trained within 10 epochs. Table S6, Supporting Information, shows the detection accuracy of 10 groups based on the SVM algorithm.…”

Section: D Spatial Explorationmentioning

confidence: 99%

Circular Subwavelength Photodetectors for 3D Space Exploration

Pan

Sun

et al. 2022

Advanced Optical Materials

View full text Add to dashboard Cite

Diverse photodetectors with spatially distributed configuration have been developed for achieving angle detection in 3D space. [7,[10][11][12][13] Currently, the photo sensitive materials are constructed in the conventional linear or planar device structures, which should be folded or embedded into the stereoscopic configuration. [14][15][16][17] Based on the macroscopic spatial distribution of the devices, the resolution and range of angle detection cannot meet the application requirements. [8,16,[18][19][20] Therefore, the functional elements should be constructed in bulky 3D models to receive differentiated light and output corresponding signals. However, it is difficult to integrate the complex and bulky 3D photo detector into the limited chip space. [21][22][23] Miniaturizing the device size and simplifying 3D structure would inevitably lead to reduced detection range and light sensitivity. It is still a challenge to design a 3D photodetector with chip-level architecture and spatial recognition function.Subwavelength-scale semiconductor structures have the characteristics of optical resonance, which can regulate the light-matter interaction. [24][25][26][27][28][29] As an alternative, many studies based on resonant structures have been conducted to improve the performance of scene-reconstructed photodetectors. [30][31][32][33][34] The subwavelength structures have the characteristics of differentiated light-matter resonance effect, which possess high photoresponse sensitivity for the incident angle. [33] The lightmatter coupling regulates the distribution of light field, which determines light absorption intensity for the photoelectric semiconductor. [35] With the characteristics of optical resonance at both low and high modes, the facade structure with subwavelength width possesses large resonance cross-section, which endows strong light scattering and absorption discrimination ability. [36][37][38][39][40] The light angle recognition in the vertical direction can be achieved by a single facade-based photodetector. [36] Therefore, the micro/nano facade structure minimizes the stacking of components. For practical application, photodetectors should be placed in multiple directions to cover the spatial dimension parallel to the ground.The light-matter interaction of subwavelength structures is affected by the orientation in the space, which induces differentiated spatial distribution of light field on the structure. [1,35,40] Circle, as the geometry with the highest symmetry, can be divided into an infinite number of equal parts along symmetry axis. Circularly distributed facades can subdivide the Detectors with the ability of 3D space exploration greatly promote sensing performance for artificial intelligence, automatic drive, etc. Current photodetectors for spatial angle detection are usually composed of planar stacked light-sensing elements on the bulky 3D configuration. To meet the miniaturization and integration of intelligent systems, it is critical to break the obstacle between chip-level architecture ...

show abstract

Early-Stage Gas Identification Using Convolutional Long Short-Term Neural Network with Sensor Array Time Series Data

Cited by 15 publications

References 51 publications

Predicting Analyte Concentrations from Electrochemical Aptasensor Signals Using LSTM Recurrent Networks

Predicting Analyte Concentrations from Electrochemical Aptasensor Signals Using LSTM Recurrent Networks

Lateral Heterostructured Vis–NIR Photodetectors with Multimodal Detection for Rapid and Precise Classification of Glioma

Circular Subwavelength Photodetectors for 3D Space Exploration

Contact Info

Product

Resources

About