Kyocera Corporation is partnering with the University of Tsukuba to develop Artificial Intelligence (AI)-based image recognition for eHealth applications to diagnose melanoma and other skin diseases by analyzing digital images of a patient’s skin.

They are targeting 2020 for commercialization.

Recent developments in AI, image recognition, and IT infrastructure are facilitating great advances in the ability to analyze digital images. Diagnosing skin diseases from digital images using AI will offer great advantages over conventional practices, which now often depend on the knowledge and experience of a physician.

Kyocera subsidiary Kyocera Communication Systems Co., Ltd. (KCCS) is working with Professor Manabu Fujimoto and Assistant Professor Yasuhiro Fujisawa (both of the Department of Dermatology, Faculty of Medicine, University of Tsukuba) to develop an image-recognition system accurate enough to distinguish several types of skin malignancies, including melanoma.

KCCS and the University of Tsukuba will conduct joint research from March 2017 through March 2018, aiming toward a commercial application in the fiscal year ending March 2020.

The next phase of their project will aim for image-based diagnostic support of any skin disease. They plan to develop a system capable of identifying more than 2,000 different skin diseases from digital images by combining their respective resources and expertise in the future.

In addition to helping dermatology specialists, AI-based image recognition could allow accurate diagnoses in rural and remote areas lacking a local clinician, using pictures from smartphones or digital cameras to greatly improve healthcare outcomes.

The project benefits from a database of more than 20,000 clinical images accumulated over 20 years by the University of Tsukuba Hospital’s Department of Dermatology. The University’s experiences and knowledge including these images will be instrumental in assessing image-based diagnostic accuracy in real-world conditions.

Additionally, KCCS will bring AI-based image-processing expertise accumulated through Labellio — a cloud-based web service that allows any user to create a simple “drag-and-drop” image classifier powered by deep learning.