Breast cancer is a highly lethal form of cancer that primarily affects women. Its early detection has been proven to significantly improve the likelihood of survival. Mammography is the primary diagnostic tool used for breast cancer, but in the early stages, it can be challenging for physicians to accurately identify tumors on mammogram images. To tackle this issue, image enhancement techniques are necessary. In this study, we introduce a novel image enhancement method that combines homomorphic filtering and contrast-limited adaptive histogram equalization (CLAHE). The outcomes obtained were compared to those of conventional filters and deep learning methods. The performance of the different methods was assessed using image dissimilarity parameters such as entropy(E), Michelson contrast (MC), mean square error (MSE), peak signal-to-noise ratio (PSNR), and structural similarity index (SSIM). The proposed method demonstrates superior image quality when compared to the other methods.

Cholangiocarcinoma, a form of liver bile duct cancer, is challenging to detect due to its critically low 5-year survival rate. Conventional imaging modalities, such as Computed Tomography (CT) and Magnetic Resonance Imaging (MRI), are widely used, but recent advancements in Hyperspectral Imaging (HSI) offer a promising, non-invasive alternative for cancer diagnosis. However, supervised learning methods often require large annotated datasets that can be difficult to obtain. To alleviate this limitation, we propose an unsupervised learning strategy using Generative Adversarial Networks (GANs) for cholangiocarcinoma detection. This approach, named Unsupervised Spectral and Spatial Attention-based GAN (USSGAN), employs an unsupervised Spectral-Spatial attention-based GAN to classify and segment cancerous regions without relying on labeled training data. The integration of an adaptive step size into Tasmanian Devil Optimization (TDO) enhances the convergence speed and effectively captures diverse cancerous features. Enhanced Tasmanian Devil Optimization (ETDO) further improves segmentation performance, making the framework robust and computationally efficient. The proposed method was tested on a publicly available multidimensional choledochal cholangiocarcinoma dataset, achieving superior performance compared with existing techniques in the literature. USSGAN demonstrated high accuracy across key metrics such as overall accuracy (OA), average accuracy (AA), and Cohen’s Kappa. Ablation studies confirmed the critical contributions of the proposed enhancements to the overall performance. With an overall accuracy of 98.03%, the USSGAN closely aligns with the assessments of experienced pathologists while maintaining minimal computational requirements. Its lightweight nature ensures real-time deployment, providing results within a minute, making it a practical and effective solution for clinical applications.

Hyperspectral imaging (HSI) has evolved as an important tool for many applications, including remote sensing, crime investigation, target detection, disease diagnosis, and anomaly detection. Among these, hyperspectral underwater target detection (HUTD) presents unique challenges due to spectral distortions caused by water absorption and scattering. Conventional methods often struggle with spectral variability, complicating detection accuracy. This paper introduces spectral variability-aware hybrid autoencoder (SVHAE) for HUTD, a novel autoencoder-based unmixing network incorporating parallel linear and nonlinear decoders to improve underwater target detection. Our method effectively reduces the effect of spectral distortions and addresses variability using a combined loss function integrating Kullback-Leibler divergence, mean squared error, and spectral angle distance. Experimental validation on real-world and simulated datasets demonstrates that our proposed SVHAE outperformed state-of-the-art methods by achieving superior AUC values. These advancements contribute to the progressing field of HUTD, making the way for robust solutions in marine exploration and detecting targets under the water.

Hyperspectral unmixing (HU) is crucial for extracting material information from hyperspectral images (HSI) obtained through remote sensing. Although linear unmixing methods are widely used due to their simplicity, they only address linear mixing effects. Nonlinear mixing models, while more complex, often focus solely on the nonlinear aspects affecting individual pixels. However, in practice, light reflected from materials within a pixel experiences linear and nonlinear interactions, necessitating a hybrid mixing model (HMM) that leverages spatial and spectral information. This work proposes a novel deep learning-based autoencoder (AE) with dual-stream decoders to enhance spectral unmixing. Our approach employs multitask learning (MTL) to process spatial and spectral information concurrently. Specifically, one decoder stream performs linear unmixing from HSI patches, while the other stream utilizes fully connected layers to capture and model the nonlinear interactions within the data. By integrating linear and nonlinear information, our method improves the accuracy of unmixing the mixed spectrum. We validate the effectiveness of our architecture on three real-world HSI datasets and compare its performance against various baseline methods. Experimental results consistently demonstrate that our approach outperforms existing methods, as evidenced by superior spectral angle distance (SAD) and mean squared error (MSE) metrics.

Microalgae-based wastewater treatment has resulted in a paradigm shift toward nutrient removal and simultaneous resource recovery. However, traditionally used microalgal biomass quantification methods are time-consuming and costly, limiting their large-scale use. The aim of this study is to develop a simple and cost-effective image-based method for microalgae quantification, replacing cumbersome traditional techniques. In this study, preprocessed microalgae images and associated optical density data were utilized as inputs. Three feature extraction methods were compared alongside eight machine learning (ML) models, including linear regression (LR), random forest (RF), AdaBoost, gradient boosting (GB), and various neural networks. Among these algorithms, LR with principal component analysis achieved an R2 value of 0.97 with the lowest error of 0.039. Combining image analysis and ML removes the need for expensive equipment in microalgae quantification. Sensitivity analysis was performed by varying the train-test splitting ratio. Training time was included in the evaluation, and accounting for energy consumption in the study leads to the achievement of high model performance and energy-efficient ML model utilization.

This article introduces a novel semi-supervised learning method for Cholangiocarcinoma detection using inherent statistical parameters of the image on the multidimensional Choledochal dataset. Results closely match the pathologist’s annotations, validated by image similarity indices.

The rapid spread of the COVID-19 pandemic necessitated unprecedented collective action against coronavirus disease. In this light,we are proposing a novel online platform for the visualization of epidemiological data incorporating social determinants for understanding the patterns associated with the spread of COVID-19. The current AI computational platform combines modeling methodologies along with temporal geospatial visualization of COVID-19 data, providing real-time sharing of graphic analytical simulation of vulnerable hotspots of recurrent (nowcasting) and emergent (forecasting) infections visualized on a spatiotemporal scale on geoportals. The proposed study will be a secondary data analysis of primary data accessed from the national portal (Indian Council of Medical Research (ICMR)) incorporating 766 districts in India. Epidemiological data related to spatiotemporal visualization of the demographic spread of COVID-19 will be displayed using a compartmental socio-epidemiological model, reproduction number R, epi-curve diagrams as well as choropleth maps for different levels of administrative and development units at the district levels.

In crime investigations, rapid bloodstain identification is crucial. Hyperspectral imaging (HSI) offers a non-destructive solution. Our investigation into preprocessing techniques to improve classification accuracy and reduce computation time reveals that the best options are max normalization and mean filter.

Cholangiocarcinoma is one of the rarest yet most aggressive cancers that has a low 5-year survival rate (2%-24%) and thus often requires an accurate and timely diagnosis. Hyperspectral Imaging (HSI) is a recently developed, promising spectroscopic-based non-invasive bioimaging technique that records a spatial image (x, y) together with wide spectral (λ) information. In this work, for the first time we propose to use a three-dimensional (3D)U-Net architecture for Hyperspectral microscopic imaging-based cholangiocarcinoma detection and classification. In addition to this architecture, we opted for a few preprocessing steps to achieve higher classification accuracy (CA) with minimal computational cost. Our results are compared with several standard unsupervised and supervised learning approaches to prove the efficacy of the proposed network and the preprocessing steps. For instance, we compared our results with state-of-the-art architectures, such as the Important-Aware Network (IANet), the Context Pyramid Fusion Network (CPFNet), and the semantic pixel-wise segmentation network (SegNet). We showed that our proposed architecture achieves an increased CA of 1.29% with the standard preprocessing step i.e., flat-field correction, and of 4.29% with our opted preprocessing steps.

A hyperspectral image (HSI) is a 3D hypercube that incorporates spatial and spectral data. It's an emerging optical imaging method especially in the field of medical imaging, allowing detailed introspection of the spectral data from biological tissues. However, this imaging technique is susceptible to various types of noise, such as thermal noise, speckle noise, Poisson noise, and Gaussian noise. This research article investigates the denoising of hyperspectral images in a multidimensional choledochal liver bile duct cancer dataset that is publicly available in kaggle. Initially, we synthetically add different types of noises to the original data. The noise is then removed using various filters, and the denoising performance is evaluated based on several metrics. The results of this study provide information regarding the efficiency of multiple filters in reducing noise in HSI.

Fundus imaging is a great tool for the detection of diabetic retinopathy; however, it often suffers from poor image quality and fails to show the vascular information which is crucial for precise diagnosis. Photoacoustic (PA) imaging is a recently developed non-invasive bioimaging technique that illuminates tissues using nanosecond laser pulses to generate acoustic waves to obtain deep tissue images with optical imaging resolution. In this study, we synthesize PA images from normal and abnormal (glaucoma-affected) retinal fundus images. One of the major limitations of synthetic vascular PA images is noise. To alleviate this problem, we propose to use a dictionary learning-based denoising technique i.e., the K-Singular Value Decomposition (K-SVD). Results are compared with several standard denoising approaches such as the Median filter, Jerman filter, and Frangi filter together with the other learning-based approaches, e.g., orthogonal matching pursuit (OMP), and sequential generalized K-means algorithms (SGK). Our results demonstrate that the K-SVD denoising method exhibits superior performance in denoising glaucoma-affected abnormal retina PA images and normal retina PA images, offering better reconstruction image quality and noise removal.

The incorporation of fundamental ratios for predicting the stock price, such as price to book (PB), price to sales (PS) and Price to earnings(PE), alongside historical price data has gained considerable attention in stock price prediction. This study aims to investigate the effectiveness of utilizing these fundamental ratios in conjunction with Long-Short Term Memory (LSTM) model for predicting stock prices. In particular, we chose three large-cap companies that are listed in the National Stock Exchange (NSE), in India. scripts such as Reliance, Tata Consultancy Services (TCS), and Imperial Tobacco Company (ITC) are selected as case studies. Historical price data and fundamental ratios are collected over a specific period of the past year i.e. march 2022 to march 2023. To ensure accurate predictions, the collected dataset undergoes preprocessing techniques. By incorporating both fundamental ratios and historical price data into LSTM model, this study aims to explore the potential benefits of combining these factors for improved stock price prediction in terms of the chosen metrics like Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE).

Mammography is an inexpensive and noninvasive imaging tool that is commonly used in detection of breast lesions. However, manual analysis of a mammogramic image can be both time intensive and prone to unwanted error. In recent times, there has been a lot of interest in using computer-aided techniques to classify medical images. The current study explores the efficacy of an Earth Mover’s Distance (EMD)-based mammographic image classification technique to identify the benign and the malignant lumps in the images. We further present a novel leader recognition (LR) technique which aids in the classification process to identify the most benign and malignant images from their respective cohort in the training set. The effect of image diversity in training sets on classification efficacy is also studied by considering training sets of different sizes. The proposed classification technique is found to identify malignant images with up to 80 % sensitivity and also provides a maximum F1 score of 72.73 %.

At the system level, cancer is viewed as the malfunctioning of proteins synthesized by the mutated genes. One of the effective ways of cancer treatment is personalized combinational drug therapy, where a best-suited mixture of drugs is given to the patient. Finding the most effective combination requires the study of mutations which is an exponentially complex task. In this work, we considered case study of melanoma. Signal transduction pathways for melanoma are mapped into Boolean networks (BNs), the plausible mutation sights in the biological pathway are marked as stuck-At faults, and drugs are marked as inhibitory inputs. Finding mutations or stuck-At faults in the BN is an NP-complete problem, which needs heuristic algorithms to determine the solutions. We used the Boolean satisfiability (or SAT) algorithm MiniSAT2.2 to determine such possible fault locations and consequently, the optimal drugs. In this work, we modified the previously published algorithm for SAT-based drugs therapy to obtain faster results and freshly applied the methods on melanoma pathways. We expect the best therapy in minimum time, which is a crucial factor in the rapidly growing disease like cancer.

Applications of unmanned aerial vehicles (UAVs) are increasing enormously. The control and stabilization of those UAVs are challenging, especially in case of applications like tracking. We developed a similar UAV (referred to as 'a quadcopter' in the manuscript) application such that it tracks the movement of the ball thrown in the air and then hits the ball when it reaches a certain altitude. Our quadcopter is based on the YMFC-AL project and is cost-effective while yet providing good flight performance. The experiment is performed indoors, and the ball tracking is done with the help of an overhead camera. The data from the camera is sent to the quadcopter with Wi-Fi for taking the necessary tracking action. A PID controller is used for the stabilization of the quadcopter at the desired location and smooth tracking. The extended version of this project can be instrumental in applications like suspect tracking in the crowd, where many live tracking and control actions are required.

The detection of a stuck-at fault in the digital circuit requires a specific set of inputs, commonly known as test inputs. The fault detection problem is NP-complete, with a complexity of at least O(2n) for n faults. With an increase in the size of the circuit, and consequently the possibility of the number of faults, the generation of test inputs becomes computationally challenging. We have used the Boolean satisfiability (SAT) algorithm to deal with such complexity. MiniSAT 2.2 is a well-known SAT algorithm with a computational complexity of less than O(1.3n). The result of this SAT problem was obtained with MiniSAT 2.2 algorithm and python platform. The results obtained were analyzed to determine the test sets for combinational circuits. Another advantage of this method is that it is helpful for the detection of single as well as multiple-simultaneous stuck-at faults, which are even harder to detect. The method has been developed by keeping in mind its possible application in biological circuits, like detecting genetic mutations.

Dynamic biological systems can be modelled to an equivalent modular structure using Boolean networks (BNs) due to their simple construction and relative ease of integration. The chemotaxis network of the bacterium Escherichia coli (E. coli) is one of the most investigated biological systems. In this study, the authors developed a multi-bit Boolean approach to model the drifting behaviour of the E. coli chemotaxis system. Their approach, which is slightly different than the conventional BNs, is designed to provide finer resolution to mimic high-level functional behaviour. Using this approach, they simulated the transient and steady-state responses of the chemoreceptor sensory module. Furthermore, they estimated the drift velocity under conditions of the exponential nutrient gradient. Their predictions on chemotactic drifting are in good agreement with the experimental measurements under similar input conditions. Taken together, by simulating chemotactic drifting, they propose that multi-bit Boolean methodology can be used for modelling complex biological networks. Application of the method towards designing bio-inspired systems such as nano-bots is discussed.

Random somatic mutations disrupt homeostasis of the cell resulting in various undesirable phenotypes including proliferation. One of the most important questions in systems medicine research is the therapeutic intervention design, which requires the knowledge of these mutations. A single or multiple mutations can occur in the diseases like cancer. These mutations have been successfully modeled as stuck-at faults in the Boolean network model of the underlying regulatory system. Identification of these fault types for multiple stuck-at faults is a non-trivial problem and requires some system theoretic introspection. This manuscript addresses the dual problem of the fault identification and the therapeutic intervention. Both the problems are mapped to the Boolean satisfiability (SAT) problem. The underlying problems are solved using a fast SAT solver. The synthetic and biological examples elucidate the effectiveness of the mapping.

A new linear approach is proposed to model the dynamics of asynchronous Boolean networks (ABNs). The continuous-time discrete-state model definition makes the ABN a suitable formulation to describe gene regulatory networks. The semi-tensor product of matrices is used as the algebraic tool for the linear formulation. The results on the dynamics of the ABN are elaborated with synthetic and biological examples.

Diagnostics and therapeutic interventions in complex systemic diseases like cancer can be mapped to fault identification and control problem. A complete linear framework has been developed in this manuscript to comprehend different classes of faults and controllability of the output for a class of homeostatic inputs in the Boolean network framework. The problems undertaken in this manuscript are non-trivial. Interesting results have been derived for fault identification and control in the linearised Boolean network model of a proliferating cellular system.

Bacterium such as Escherichia coli (E. coli) show biased Brownian motion in different chemical concentration gradients. This chemical sensitive motility or chemotaxis has gained considerable interest among scientists for some remarkable features such as chemo-sensory dynamic range, adaptation, diffusion and drift. A Boolean model of the whole chemotaxis process has been developed in this manuscript. The response of the circuit is in accordance with the experimental results available in the literature, providing indirect validation of the model. This simple Boolean network (BN) can be easily integrated into the paradigm of modular whole cell modelling. Another crucial application is in designing bio-inspired micro-robots to detect certain spatio-temporal chemical signatures.