Plant disease recognition using residual convolutional enlightened Swin transformer networks

Agriculture plays a pivotal role in the economic development of a nation, but, growth of agriculture is affected badly by the many factors one such is plant diseases. Early stage prediction of these disease is crucial role for global health and even for game changers the farmer's life. Recently, adoption of modern technologies, such as the Internet of Things (IoT) and deep learning concepts has given the brighter light of inventing the intelligent machines to predict the plant diseases before it is deep-rooted in the farmlands. But, precise prediction of plant diseases is a complex job due to the presence of noise, changes in the intensities, similar resemblance between healthy and diseased plants and finally dimension of plant leaves. To tackle this problem, high-accurate and intelligently tuned deep learning algorithms are mandatorily needed. In this research article, novel ensemble of Swin transformers and residual convolutional networks are proposed. Swin transformers (ST) are hierarchical structures with linearly scalable computing complexity that offer performance and flexibility at various scales. In order to extract the best deep key-point features, the Swin transformers and residual networks has been combined, followed by Feed forward networks for better prediction. Extended experimentation is conducted using Plant Village Kaggle datasets, and performance metrics, including accuracy, precision, recall, specificity, and F1-rating, are evaluated and analysed. Existing structure along with FCN-8s, CED-Net, SegNet, DeepLabv3, Dense nets, and Central nets are used to demonstrate the superiority of the suggested version. The experimental results show that in terms of accuracy, precision, recall, and F1-rating, the introduced version shown better performances than the other state-of-art hybrid learning models.

A novel machine learning model for breast cancer detection using mammogram images

The most fatal disease affecting women worldwide now is breast cancer. Early detection of breast cancer enhances the likelihood of a full recovery and lowers mortality. Based on medical imaging, researchers from all around the world are developing breast cancer screening technologies. Due to their rapid progress, deep learning algorithms have caught the interest of many in the field of medical imaging. This research proposes a novel method in mammogram image feature extraction with classification and optimization using machine learning in breast cancer detection. The input image has been processed for noise removal, smoothening, and normalization. The input image features were extracted using probabilistic principal component analysis for detecting the presence of tumors in mammogram images. The extracted tumor region is classified using the Naïve Bayes classifier and transfer integrated convolution neural networks. The classified output has been optimized using firefly binary grey optimization and metaheuristic moth flame lion optimization. The experimental analysis has been carried out in terms of different parameters based on datasets. The proposed framework used an ensemble model for breast cancer that made use of the proposed Bayes + FBGO and TCNN + MMFLO classifier and optimizer for diverse mammography image datasets. The INbreast dataset was evaluated using the proposed Bayes + FBGO and TCNN + MMFLO classifiers, which achieved 95% and 98% accuracy, respectively.

ECG Noise Removal Using Modified Distributed Arithmetic Based Finite Impulse Response Filter

ECG monitoring is essential to support human life. During signal acquisition, the signals are contaminated by various noises that occur due to different sources. This paper focuses on Baseline wander and Muscle Artifact noise removal using Distributed Arithmetic (DA) based FIR filters. An area-efficient modified DA based FIR filter consists of LUT-less structure and used for noise removal. The performance of the modified DA based FIR filter is compared with the conventional DA FIR filter. An arbitrary real-time ECG record is taken from MIT-BIH database and Baseline Wander noise, Muscle artifact noises are taken from MIT-BIH noise stress test database. The performance of both filters is evaluated in terms of output Signal to Noise Ratio (SNR) and Mean Square Error (MSE). For Baseline wander noise removal, the modified DA based FIR filter produces high output SNR and also low MSE of 76.6% than the conventional filter. Similarly, for Muscle Artifact noise removal, it produces high SNR, and MSE is reduced to 73.8%. A modified DA based FIR filter is synthesized for the target FPGA device Spartan3E XC3s2000-4fg900 and hardware resource utilization is presented.