AI image analysis

As a “second opinion” during your consultation, your doctor might use an artificial intelligence system to analyse your spot.

The DermEngine Visual Search system compares your spot with a database of many others. It retrieves the images that most closely resemble your spot and presents statistics of these images, e.g. how many of these images are skin cancers.

This system is not foolproof and sometimes tends to overdiagnose skin cancers. However it can be a useful backup tool and provides educational information to help you understand why your spot is likely or unlikely to be skin cancer.

What is the evidence?

Results of a trial comparing multiple AI image analysis algorithms with medical professionals (including dermatologists GPs with expertise in dermoscopy) and demonstrated that the AI algorithms:

• Detected skin cancers with better sensitivity than human practitioners.
• Classified skin lesions better than human practitioners.

The DermEngine algorithm outperformed other algorithms in this trial.¹

Limitations

The DermEngine AI algorithm works well on a controlled set of images used in a clinical trial but is not necessarily as effective on real-world images seen in clinical practice. In addition, a doctor examining a skin lesion can also get further information: how long it has been there, whether it has changed, what it feels like on examination and more.

This means the AI analysis should be seen as part of an overall skin lesion assessment in conjunction with your doctor’s clinical assessment and examination.

References

1. Tschandl, P., Argenziano, G., Razmara, M., Yap, J. (2018). Diagnostic accuracy of content‐based dermatoscopic image retrieval with deep classification features. British Journal of Dermatology. 
2. DermEngine team (2019). Study overview | How DermEngine’s AI algorithms are shaping the future of dermatology.