A total of 412,000 persons are expected to be diagnosed with leukemia worldwide, with acute lymphoblastic leukemia accounting for around 12% of all cases. Thus, the early detection of leukemia can save millions of populations. This article mainly focusing on blood cell counting and detection of Leukemia using deep learning mechanisms. Initially, the images are preprocessed using median filters and segmented using RGB-Otsu's approach. Then, the features are extracted using scale-invariant feature transform (SIFT) and these features are applied to recurrent neural network (RNN) for classification. Finally, the proposed method resulted in superior performance as compared to the state of art machine learning approaches.
Brain tumor is the most common kind of cancer, accounting for millions of fatalities worldwide. Early detection and treatment of new dangerous Brain tumor cases are critical to ensuring a low mortality rate as well as a high survival rate. The majority of relevant research has focused on machine learning-based algorithms, but they haven't been able to achieve the highest levels of accuracy and specificity. To address this issue, this work employs a deep learning based convolutional neural network (CNN) classification mechanism based on advanced deep learning. For effective identification of the location of Brain tumor, a k-means clustering based segmentation technique is first applied. Finally, CNN was constructed for Brain tumor classification with grey level co-occurrence matrix (GLCM) based Texture features, decimated wavelet Transform (DWT) based low level features, and Statistical Color features, respectively, to archive the system's maximum efficiency. As a result, the research findings can be utilized to classify benign and malignant based brain tumors. In comparison to state-of-the-art techniques, simulation study demonstrates that the suggested method provides better qualitative and quantitative analyses.
11 May 2025 in International Journal of Electronics
Real-time transmission of biomedical signals is immensely challenging and requires cloud and internet of things (IoT) infrastructure. Security is also an important factor; however, to accomplish this, a reconstruction method is developed in which the entire signal is supplied as an input, the primary portion is considered here, and the signal is further encoded and transmitted to the destination. Electrocardiogram (ECG) compression for the lightweight wireless network is quite challenging for long-term healthcare monitoring. Compressed sensing (CS) involves efficient encoding mechanisms for error rate estimation for reconstruction and energy consumption for wireless transmission of data. We propose a multi-task compressed sensing (MT-CS) reconstruction mechanism in this study for ECG compression of data is most chosen for a wireless network system that has various sensors embedded in it. This model further extracts the essential adaptive features for correlation existing in the ECG signals. The performance of the proposed MT-CS reconstruction mechanism is evaluated on the multiparameter intelligent monitoring in intensive care (MIMIC-II) dataset, which ensures its robustness and generalization. The results obtained upon simulation ensure that the proposed MT-CS based reconstruction approach ensures excellent reconstruction signal with fewer measurements in comparison with the existing state-of-art CS techniques.