How does SkinVision’s algorithm detect skin cancer?

Curious about how SkinVision's technology works? Let's break it down

The Power of Learning:

Just like humans learn from experience, SkinVision’s algorithm improves its accuracy over time. It's fueled by machine learning, which means it analyzes vast amounts of data to identify patterns. These patterns help predict the risk level of skin cancer in a skin spot.

Training the Algorithm:

Our journey begins with data—over 100,000 skin spot images selected from 2.9 million photos assessed by our dermatologists. This diverse database covers all skin types and conditions, providing a solid foundation for training. By learning from these dermatologist-reviewed images, the algorithm becomes adept at distinguishing between different skin spot photos and their associated risk levels.

Accuracy and Reliability:

The true test of our algorithm's quality comes from comparing its assessments against the golden standard—biopsy-confirmed skin cancer cases. Here, our algorithm has shown impressive sensitivity, detecting skin cancer correctly 95% of the time. This achievement underscores the tool's reliability, supported by clinical tests and research published in reputable scientific journals.

Preventing Unnecessary Worries:

With a specificity rate of 80.1% for clearly benign lesions, SkinVision also excels in identifying harmless spots. This means many users are reassured about spots that don't require a doctor's visit, helping to prevent unnecessary worry and consultations.

For more detailed clinical evidence and to understand the depth of our algorithm’s accuracy, you can explore the following resources:

https://pubmed.ncbi.nlm.nih.gov/31494983/

https://www.karger.com/Article/FullText/520474

References:

1. Udrea, A. et al. (2020), Accuracy of a smartphone application for triage of skin lesions based on machine learning algorithms. J Eur Acad Dermatol Venereol, 34: 648-655.
2. Sangers, T. et al. (2022). Validation of a Market-Approved Artificial Intelligence Mobile Health App for Skin Cancer Screening: A Prospective Multicenter Diagnostic Accurancy Study. Dermatology, 238: 649-656.