MakerShark

Polarderm+ - Polarized smartphone dermoscope attachment & AI skin lesion classifier web-app

Polarderm+ is a skin lesion detection package created by Andrew Pun that includes a 3D-printed smartphone attachment and an AI skin lesion classifier web-app.

Polarderm - 3D-printed smartphone attachment

The 3D-printed smartphone attachment allows physicians to take pictures of skin lesions more clearly with their smartphone camera. Polarderm+ utilizes two orthogonally-oriented polarizing filters and an LED light source to provide polarized illumination, simulating polarized dermoscopic techniques. This reduces glare and reveals sub-surface structures within the targeted lesion, assisting physicians with diagnosis of skin cancers such as melanoma.

Polarderm - AI skin classifier web-app

The smartphone attachment can be used in tandem with the included AI skin lesion classifier web-app. Built using Tensorflow with training data from the open-source Harvard HAM10000 dataset, the AI is able to recognize 7 different categories of skin lesions from just a picture, such as actinic keratosis, basal cell carcinoma, benign keratosis, dermatofibroma, melanocytic nevi, melanoma, and vascular lesions. Experimental data suggests that the AI is especially accurate when analyzing images using the smartphone attachment. The web-app has been configured for use with the Oscar EMR to instantly upload images with their results into a patient's file, but may be modified to use in other EMR systems.

Read more

Read more about Polarderm+ and its beginnings as a science fair project here and here. You can also download the .stl files for Polarderm+ for free on GitHub.