Using 3D differential forms to characterize a pigmented lesion in vivo

Area of Science:

• Dermatology
• Medical Imaging
• Computational Geometry

Background:

• Traditional melanoma diagnosis relies on ABCD rules (asymmetry, border, color, diameter).
• Emerging feature extraction methods aim to improve melanoma characterization.
• A new direction is needed, orthogonal to ABCD rules, for enhanced melanoma recognition.

Purpose of the Study:

• To describe a new research direction for automatic melanoma recognition.
• To characterize 3D local disruption of skin surfaces.
• To explore differential forms of skin surfaces for melanoma classification.

Main Methods:

• Acquired 3D skin surface data using photometric stereo.
• Determined 3D differential forms of lesion surfaces to capture geometrical texture.
• Employed a least-squared error-based linear classifier for binary classification.

Main Results:

• Achieved optimal sensitivity of 71.4% and specificity of 86.4% for melanoma classification.
• Utilized 3D data from 35 melanoma and 66 benign lesion samples.
• The receiver operating characteristics curve area was 0.823.

Conclusions:

• 3D differential forms show significant promise for melanoma characterization.
• This geometric approach offers a new perspective beyond ABCD rules.
• Combining 3D features with traditional ones may enhance diagnostic accuracy and reliability.