StressNet: a spatial-spectral-temporal deformable attention-based framework for water stress classification in maize.

In recent years, monitoring the health of crops has been greatly aided by deploying highthroughput crop monitoring techniques that integrate remotely captured imagery and deep learning techniques. Most methods rely mainly on the visible spectrum for analyzing the abiotic stress, such as water deficiency in crops. In this study, we carry out experiments on maize crop in a controlled environment of different water treatments. We make use of a multispectral camera mounted on an Unmanned Aerial Vehicle for collecting the data from the tillering stage to the heading stage of the crop. A pre-processing pipeline, followed by the extraction of the Region of Interest from orthomosaic is explained. We propose a model based on a Convolution Neural Network, added with a deformable convolutional layer in order to learn and extract rich spatial and spectral features. These features are further fed to a weighted Attention-based Bi-Directional Long Short-Term Memory network to process the sequential dependency between temporal features. Finally, the water stress category is predicted using the aggregated Spatial-Spectral-Temporal Characteristics. The addition of multispectral, multi-temporal imagery significantly improved accuracy when compared with mono-temporal classification. By incorporating a deformable convolutional layer and Bi-Directional Long Short-Term Memory network with weighted attention, our proposed model achieved best accuracy of 91.30% with a precision of 0.8888 and a recall of 0.8857. The results indicate that multispectral, multi-temporal imagery is a valuable tool for extracting and aggregating discriminative spatial-spectral-temporal characteristics for water stress classification.

Multiview spatial compounding using lens-based photoacoustic imaging system.

Recently, an acoustic lens has been proposed for volumetric focusing as an alternative to conventional reconstruction algorithms in Photoacoustic (PA) Imaging. Acoustic lens can significantly reduce computational complexity and facilitate the implementation of real-time and cost-effective systems. However, due to the fixed focal length of the lens, the Point Spread Function (PSF) of the imaging system varies spatially. Furthermore, the PSF is asymmetric, with the lateral resolution being lower than the axial resolution. For many medical applications, such as in vivo thyroid, breast and small animal imaging, multiple views of the target tissue at varying angles are possible. This can be exploited to reduce the asymmetry and spatial variation of system the PSF with simple spatial compounding. In this article, we present a formulation and experimental evaluation of this technique. PSF improvement in terms of resolution and Signal to Noise Ratio (SNR) with the proposed spatial compounding is evaluated through simulation. Overall image quality improvement is demonstrated with experiments on phantom and ex vivo tissue. When multiple views are not possible, an alternative residual refocusing algorithm is proposed. The performances of these two methods, both separately and in conjunction, are compared and their practical implications are discussed.

Two-sided residual refocusing for an acoustic lens-based photoacoustic imaging system.

In photoacoustic (PA) cameras, an acoustic lens-based system can form a focused image of an object plane. A real-time C-scan PA image can be formed by simply time gating the transducer response. While most of the focusing action is performed by the lens, residual refocusing is needed to image multiple depths with high resolution simultaneously. However, a refocusing algorithm for a PA camera has not been studied so far in the literature. In this work, we reformulate this residual refocusing problem for a PA camera into a two-sided wave propagation from a planar sensor array. One part of the problem deals with forward wave propagation while the other deals with time reversal. We have chosen a fast Fourier transform (FFT) based wave propagation model for the refocusing to maintain the real-time nature of the system. We have conducted point spread function (PSF) measurement experiments at multiple depths and refocused the signal using the proposed method. The full width at half maximum (FWHM), peak value and signal to noise ratio (SNR) of the refocused PSF is analyzed to quantify the effect of refocusing. We believe that using a two-dimensional transducer array combined with the proposed refocusing can lead to real-time volumetric imaging using a PA camera.

Characterization of lens based photoacoustic imaging system.

Some of the challenges in translating photoacoustic (PA) imaging to clinical applications includes limited view of the target tissue, low signal to noise ratio and the high cost of developing real-time systems. Acoustic lens based PA imaging systems, also known as PA cameras are a potential alternative to conventional imaging systems in these scenarios. The 3D focusing action of lens enables real-time C-scan imaging with a 2D transducer array. In this paper, we model the underlying physics in a PA camera in the mathematical framework of an imaging system and derive a closed form expression for the point spread function (PSF). Experimental verification follows including the details on how to design and fabricate the lens inexpensively. The system PSF is evaluated over a 3D volume that can be imaged by this PA camera. Its utility is demonstrated by imaging phantom and an ex vivo human prostate tissue sample.

Affordable low complexity heart/brain monitoring methodology for remote health care.

This paper introduces a dual-mode low complex on-chip methodology for processing of ECG (Electrocardiogram) and EEG (Electroencephalography) signals, wherein based on the input switch the architecture can be dynamically configured to operate either as an ECG bio-marker or EEG signal de-noising system. In both the modes the signal processing technique depends on the output of the DWT (Discrete Wavelet Transform), hence a low complex methodology has been developed in which both ECG and EEG processing blocks sharing the same DWT block resulting in low area and low power consumption. The integrated ECG and EEG methodology has been implemented in Matlab, for verifying the ECG processing block the ECG database is taken from MIT-BIH PTBDB and IITH DB, similarly for EEG processing block the EEG signals are taken from PhysioNet database. The outcome of methodology in Matlab is equal to the results obtained from individual ECG and EEG blocks.