Nail Insight: Enhanced Nail Image Analysis for Early Disease Detection
Sayeeda Khanum Pathan, Sravani Jatoth, Paani Narisetty, Sai Venkat Pulari, Ajaysinha Vadithya
Abstract
The human nails serve as vital indicators of overall health, reflecting potential underlying conditions and diseases. Examining their appearance and texture can provide valuable clues to various health disorders, facilitating early diagnosis and prevention. Images depicting these nail conditions were classified using five standard deep Convolutional Neural Network (CNN) architectures: Mobile Net, ResNet50, Dense Net, Google Net, Vgg16, and Yolo v8. Upon observing these standard models, the study proposes a fusion model that combines VGG16 and Google Net through concatenation and weighted average methodologies. Across the standard architectures Mobile Net, ResNet50, Dense Net, Google Net, Vgg16, and Yolov8, the accuracies were 38.83%, 67.14%, 88.42 %, 94.95%, 97.50<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, and 92.35%, respectively. The fusion architecture of VGG16 and Google Net using concatenation and weighted concatenation at the feature level yielded accuracies of 88.11 % and 88.17%, respectively. The fusion architecture of VGG16 and Google Net using concatenation and weighted Average at Decision level yielded accuracies of 84.62% and 89.48%, respectively.