Grid-Based Structural and Dimensional Skin Cancer Classification with Self-Featured Optimized Explainable Deep
Kavita Behara1, Ernest Bhero2, John Terhile Agee2
1Department of Electrical Engineering, Mangosuthu University of Technology, Durban 4031, South Africa.
International Journal of Molecular Sciences
|February 10, 2024
Summary
A new Grid-Based Structural and Dimensional Explainable Deep Convolutional Neural Network improves artificial intelligence (AI) for skin cancer classification. This AI model enhances early detection accuracy and interpretability, outperforming existing methods.
Area of Science:
- Dermatology and Artificial Intelligence
- Medical Image Analysis
- Machine Learning in Healthcare
Background:
- Traditional skin cancer diagnosis is costly, time-consuming, and requires expert medical professionals.
- Artificial intelligence (AI) offers potential for automated skin cancer diagnosis but faces challenges in complexity, reproducibility, and explainability.
- Early detection of skin cancer is crucial for effective treatment and patient outcomes.
Purpose of the Study:
- To develop a novel, accurate, and interpretable AI model for skin cancer classification.
- To address the limitations of existing AI diagnostic tools, specifically complexity, reproducibility, and explainability.
- To improve the accuracy and robustness of AI-assisted early skin cancer detection.
Main Methods:
- Proposed a Grid-Based Structural and Dimensional Explainable Deep Convolutional Neural Network (GSD-EDCNN).
- Utilized adaptive thresholding for region of interest (ROI) extraction and VGG-16 for hierarchical feature extraction.
- Employed an Adaptive Intelligent Coney Optimization (AICO) algorithm for hyperparameter tuning and self-feature selection.
- Trained and validated the model on the ISIC (10,015 images) and MNIST (2,357 images) datasets.
Main Results:
- Achieved high accuracy (0.96) and CSI (0.97) on the ISIC dataset, significantly outperforming various established CNN models.
- Demonstrated minimal false positive rate (FPR) of 0.03 and false negative rate (FNR) of 0.02 with the AICO-optimized model.
- Attained low model loss values (0.09 for ISIC, 0.18 for MNIST), indicating superior performance.
- The model showed improved accuracy, interpretability, and robustness compared to existing techniques.
Conclusions:
- The proposed GSD-EDCNN model offers a significant advancement in AI-driven skin cancer classification.
- The model's enhanced accuracy and interpretability can aid clinicians in earlier and more reliable diagnosis.
- This research contributes to the development of more effective AI tools for dermatological applications.
Related Concept Videos
01:30Skin Cancer
4.1K
Skin cancer is a type of cancer that occurs when there is an abnormal growth of skin cells, usually triggered by damage to the DNA within the skin cells. It is primarily caused by exposure to ultraviolet (UV) radiation from the sun or artificial sources like tanning beds. Skin cancer is the most common type of cancer worldwide, and its incidence continues to rise.
Basal Cell Carcinoma (BCC): BCC is the most common type of skin cancer, accounting for about 80% of cases. It typically develops in...
Basal Cell Carcinoma (BCC): BCC is the most common type of skin cancer, accounting for about 80% of cases. It typically develops in...
4.1K
01:12Renewal of Skin Epidermal Stem Cells
2.5K
The skin is divided into epidermis, dermis, and hypodermis, the skin's outermost, middle, and inner layers. The human epidermal layer regularly undergoes renewal, where old, dead cells are replaced by new cells. Epidermal stem cells or EpiSCs divide and differentiate to restore the lost cells. For the renewal process, some EpiSCs continuously self-renew. In contrast, few others differentiate into transit-amplifying cells, which later form prickle or spinous cells, followed by granular...
2.5K
01:19Clinical Applications of Epidermal Stem Cells
2.7K
Epidermal stem cells (EpiSCs) are mainly located at the basal layer of the epidermis. These cells repair minor injuries of the skin and replace dead skin cells. However, EpiSCs’ cannot heal severe wounds such as major burns or those from diabetes or hereditary disorders. In such cases, culturing the epidermal stem cells from the patient is possible and has yielded successful treatment options, such as laboratory-grown skin grafts. These grafts are synthesized using a patient’s own...
2.7K