Optical Radiomic Signatures Derived from Optical Coherence Tomography Images Improve Identification of Melanoma

Area of Science:

• Dermatology and biomedical optics.
• Medical imaging and artificial intelligence.

Background:

• Current melanoma diagnosis relies on invasive excisional biopsy, leading to unnecessary procedures for benign lesions.
• Biopsies cause patient pain, scarring, and disfigurement, increasing healthcare burdens.
• Optical coherence tomography (OCT) offers noninvasive, high-resolution imaging for melanoma diagnosis.

Purpose of the Study:

• To develop and validate an image analysis algorithm, optical properties extraction (OPE), to enhance OCT's diagnostic performance for melanoma.
• To improve the sensitivity and specificity of noninvasive melanoma detection using OCT.
• To reduce the need for invasive biopsies in melanoma diagnosis.

Main Methods:

• An "optical properties extraction" (OPE) algorithm was developed to identify unique optical radiomic signatures for melanoma detection.
• The OPE algorithm's performance was evaluated using tissue-mimicking phantoms.
• The algorithm was tested on 69 human subjects for differentiating benign nevi from melanoma.

Main Results:

• The OPE algorithm successfully identified unique optical radiomic signatures for melanoma detection.
• Benign nevi and melanoma were differentiated with 97% sensitivity and 98% specificity.
• The algorithm demonstrated high accuracy in distinguishing between malignant melanoma and benign nevi.

Conclusions:

• The OPE algorithm significantly improves OCT's specificity and sensitivity for melanoma diagnosis.
• This noninvasive method offers a safe, simple, and accurate alternative to traditional biopsies.
• Clinical adoption of the OPE algorithm can enhance melanoma diagnosis and improve patient outcomes.