Intelligent algorithm for detection of dengue using mobilenetv2‐based deep features with lymphocyte nucleus

• DOI: 10.1111/exsy.12904
• Publication Date: 2021-12-01T01:15:55Z
• Authors: Hilda Mayrose; Niranjana Sampathila; G. Muralidhar Bairy; Sushma Belurkar; Kavitha Saravu; Akash Basu; Saman Khan
• Keywords: Blood smear

Abstract Dengue is a vector‐borne disease that highly endemic in countries located tropical regions. It can cause severe complications and even lead to death the case of delayed diagnosis. Detection dengue done by manually examining platelets lymphocytes Leishman's stained peripheral blood smear (PBS) images. PBS examination considered gold standard for diagnosing various haematological disorders. However, manual analysis labour‐intensive, tedious, time‐consuming, requiring skilled experienced haematologist. Today, soft computing methods artificial intelligence have made their way into every science technology branch. One such area which has adopted this approach digital pathology, automatically identifying diseases. The main objective work was design an intelligent algorithm classify normal patients with help microscopic A total 94 dengue‐infected PBSs were acquired at magnification 100×. Grey‐level segmentation based on Otsu's thresholding used nucleus lymphocytes. Distinct features from differentiated infected cells extracted using pre‐trained MobileNetV2 network local binary pattern. Significant selected ReliefF algorithm. Subsequently, these fed support vector machine (SVM) classifier. Our proposed system gave accuracy, sensitivity, specificity 95.74%, 98.14%, 92.50%, respectively. Hence, developed model deep hand‐crafted be valuable