Publication: Convolutional neural network-based acne detection using reconstructed hyperspectral images
Loading...
Date
2024-01-01
Authors
Ali, Mohammed Ridha
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Acne vulgaris is a common type of skin disease that affects more than 85% of
teenagers and frequently continues even in adulthood. Although it is not a dangerous
skin disease, it can considerably affect quality of life. With recent advancements in
imaging techniques, hyperspectral imaging (HSI), which captures a wide spectrum of
light, has emerged as a tool for the detection and diagnosis of various skin conditions.
However, its practical use in clinical settings is limited due to the high cost of
specialised hyperspectral (HS) cameras. In this research, an acne detection system that
utilises reconstructed HS images from red, green, blue (RGB) images is proposed. The
study consists of two parts. In the first part, a hierarchical spectral reconstruction
algorithm (HRNET) and CNN reconstruction algorithm (HSCNN-D), are proposed to
create a dataset of reconstructed acne HS images. The second part of the research
focuses on the detection of acne regions by using the RetinaNet algorithm with the
ResNet backbone, which is tested by using three datasets, namely, the original HS,
reconstructed HS and RGB images. The proposed model with RetinaNet and ResNet
produces the best precision (66.14%) when the reconstructed HS image dataset is used
and outperforms two existing methods that utilise the Faster R-CNN and F-RCN
algorithms. Results suggest that by using a deep-learning approach, the reconstructed
HS images have high potential to be used to aid in acne diagnosis. In conclusion, this
study has successfully achieved its main objective, which is to develop a reliable
RetinaNet-based detection system for acne using reconstructed HS images.