| Abstrakt: |
CKD is a medical condition that affects people all around the world and is a major health concern. It increases the risk of developing cardiovascular and cerebrovascular diseases, which can lead to serious illness and even death. Ultrasound imaging is typically the initial and most widely used diagnostic technique for individuals at risk for CKD. The existing methods are restricted by features with high dimensions, computational hurdles, and extended processing times. To address these issues, this article proposes the development of an enhanced deep-learning model with an optimum selection of features for the accurate diagnosis of CKD. The proposed technique begins with pre-processing, involving image filtering and contrast enhancement. Then, presented an improved Otsu's algorithm for segmenting kidney masses, followed by a stage of feature extraction. A hybrid Lion Optimization Algorithm (LOA) and Moth flame Optimization (MFO) is a novel contributions to improve the convergence rate of the MFO algorithm. The hybrid FS algorithms select the optimal subset of features for disease classification. Finally, a novel Long-Term Recurrent Convolutional Network (LRCN) is introduced for detecting kidney impairment. The models are developed and validated utilizing a database of ultrasonography (US) images including four classes of Chronic kidney images: stone, cyst, tumor, and normal. The efficiency of the framework is assessed based on its of accuracy, precision, recall, F1-score, and specificity of 98.7%, 96.6%, 96.4%, 97.9%, and 96.2% respectively. In addition, testing results show that the framework obtains the best overall performance when compared to existing methods for the classification of ultrasound images. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink