DSCC_Net: Multi-Classification Deep Learning Models for Diagnosing of Skin Cancer Using Dermoscopic Images
Maryam Tahir1, Ahmad Naeem2, Hassaan Malik1,2
1Department of Computer Science, National College of Business Administration & Economics Lahore, Multan Sub Campus, Multan 60000, Pakistan.
Cancers
|April 13, 2023
Summary
This study introduces DSCC_Net, a deep learning model for accurate skin cancer detection. The novel network significantly outperforms existing models in classifying melanoma, basal cell carcinoma, squamous cell carcinoma, and melanocytic nevi.
Area of Science:
- Dermatology and Artificial Intelligence
- Medical Image Analysis
- Computational Pathology
Background:
- Skin cancer diagnosis is critical for patient survival, yet current methods are time-consuming.
- Early detection significantly improves recovery prospects for skin cancer patients.
- Deep learning (DL) shows promise for enhancing the speed and accuracy of skin cancer identification.
Purpose of the Study:
- To develop a multi-classification deep learning model for diagnosing four types of skin cancer: melanoma (MEL), basal cell carcinoma (BCC), squamous cell carcinoma (SCC), and melanocytic nevi (MN).
- To introduce a novel deep learning-based skin cancer classification network (DSCC_Net) for improved diagnostic performance.
- To evaluate DSCC_Net against established deep learning models using public benchmark datasets.
Main Methods:
- Proposed a novel deep learning convolutional neural network (CNN) architecture named DSCC_Net.
- Evaluated DSCC_Net on three benchmark datasets: ISIC 2020, HAM10000, and DermIS.
- Employed SMOTE Tomek technique to address class imbalance issues in the datasets.
- Compared DSCC_Net performance against six baseline models: ResNet-152, Vgg-16, Vgg-19, Inception-V3, EfficientNet-B0, and MobileNet.
Main Results:
- DSCC_Net achieved a high classification performance with 99.43% AUC, 94.17% accuracy, 93.76% recall, 94.28% precision, and 93.93% F1-score.
- The proposed DSCC_Net significantly outperformed all six baseline models in classifying the four skin cancer types.
- Baseline models achieved accuracies ranging from 89.32% to 92.51%.
Conclusions:
- The novel DSCC_Net demonstrates superior performance in classifying various skin cancers compared to existing deep learning models.
- DSCC_Net offers a promising tool to aid dermatologists and healthcare professionals in the early and accurate diagnosis of skin cancer.
- The developed model has the potential to enhance patient outcomes through improved diagnostic efficiency.
