Related Experiment Video
Updated: Jul 17, 2026
09:37
Combining Reflectance Confocal Microscopy with Optical Coherence Tomography for Noninvasive Diagnosis of Skin Cancers via Image Acquisition
Published on: August 18, 2022
Combination of features from skin pattern and ABCD analysis for lesion classification.
Zhishun She1, Y Liu, A Damatoa
1Faculty of Technology and Computer Science, NEWI, University of Wales, Wrexham, UK. z.she@newi.ac.uk
Summary
Combining standard ABCD features with skin pattern analysis significantly improves the classification of skin lesions. This novel approach enhances the ability to distinguish malignant melanoma from benign lesions, offering a promising diagnostic tool.
Area of Science:
- Dermatology
- Medical Imaging
- Computational Pathology
Background:
- Standard skin lesion classification relies on ABCD features: asymmetry, border irregularity, color variegation, and diameter.
- Malignant lesions disrupt skin patterns, unlike benign ones, suggesting pattern disruption as a diagnostic indicator.
- Previous studies showed promise in using skin line direction and intensity but haven't combined them with ABCD features.
Purpose of the Study:
- To explore combining skin pattern features with ABCD features for enhanced skin lesion classification.
- To investigate the diagnostic potential of skin pattern disruption measurements in white light optical images.
- To improve the accuracy of distinguishing malignant melanoma from benign skin lesions.
Main Methods:
- Extracted skin line direction and intensity from local tensor matrices.
- Conducted ABCD analysis to generate six features: asymmetry, border irregularity, color (RGB) variegation, and diameter.
- Combined the eight features using Principal Component Analysis (PCA) to derive two dominant features for classification.
Main Results:
- Processing of images containing malignant melanoma (MM) and benign naevi showed excellent separation in the two-dimensional dominant feature space.
- A Receiver Operating Characteristic (ROC) plot demonstrated an area under the curve of 0.94, indicating high classification performance.
- Individual features showed limited discrimination, while combined features proved promising.
Conclusions:
- Combined features from skin pattern analysis and ABCD criteria offer superior discrimination capability compared to individual features.
- The proposed method shows significant promise for distinguishing malignant melanoma from benign lesions.
- Integrating diverse feature sets enhances diagnostic accuracy in dermatological imaging.
