Work

An efficient 3D color-texture feature and neural network technique for melanoma detection

Year: 2019

Type: article

Abstract: Malignant melanoma is the deadliest form of skin cancer, but can be more readily treated successfully if detected in its early stages. Due to the increasing incidence of melanoma, research in the field of autonomous melanoma detection has accelerated. In this paper, a new method for feature extraction from dermoscopic images, termed multi-direction 3D color-texture feature (CTF), is proposed, and detection is performed using a back propagation multilayer neural network (NN) classifier. The proposed method is tested on the PH 2 dataset (publicly available) in terms of accuracy, sensitivity, and specificity. The extracted combined CTF is fairly discriminative. When it is input and tested in a neural network classifier that is provided, encouraging results are obtained, i.e. accuracy = 97.5%, sensitivity = 98.1% and specificity = 93.84%. Comparative result analyses with other methods are also discussed, and the results are also improved over benchmarking results for the PH2 dataset.

Source: Informatics in Medicine Unlocked

DOI: 10.1016/j.imu.2019.100176

Authors: Firoz Warsi, Ruqaiya Khanam, Suraj Kamya, Carmen Paz Suárez-Araujo

Institutions: Galgotias University, IIMT College of Engineering, Greater Noida, Universidad de Las Palmas de Gran Canaria

Language: English

Topic: Cutaneous Melanoma Detection and Management

Subfield: Oncology

Field: Medicine

Domain: Health Sciences

SDG: Good health and well-being

Open Access status: gold

Estimated APC paid: $1,400