A low-cost smartphone attachment for providing repeatable imaging and structured lighting, for 3D texture analysis of pigmented lesions for early detection of skin cancer

Skin cancer is an increasingly serious health issue for which there is currently a lack of devices available to PCPs for objective and repeatable measurement of 2D/3D cancer-related mole features. We have manufactured such a device in the form of a low-cost self-contained lighting ring, which fits onto a smartphone. The device interfaces with the smartphone to capture six high resolution images taken under controlled illumination. This eliminates the drawback of existing smartphone apps for monitoring moles, where the viewing position and lighting characteristics may vary significantly between images, resulting in a lack of objectivity/ repeatability in image capture that limits the reliability/utility of the medical data generated. The six images from our device are combined using Photometric Stereo (PS) to generate 2D (albedo) and 3D (bump map) lesion data. Our research has indicated that inclusion of 3D data increases specificity of pigmented lesions from 25.7 to 76.4%. Combining 3D and 2D PS data, with deep learning analysis, offers potential to further increase performance levels, towards 100% accuracy for on-the-phone 2D/3D mole analysis. Additionally, the inclusion of the 3D data in synthetically rendered visualizations facilitates increased realism/interaction in remote lesion analysis, which is potentially transformational for tele-dermatology. The device demonstrates the critical importance of structured lighting, 3D data modelling and deep learning for accurate and repeatable skin feature identification and analysis. Being low-cost and easy to use, it can be widely employed to provide early identification of suspicious lesions, with their earlier treatment/excision providing an opportunity to save lives.