Nobleo developed a custom multispectral camera prototype for the assessment of pigmented lesions in the human skin. Prior to the development an extensive patent screening and IP landscaping has been done for the particular application.

Next to the hardware, we have developed the deep learning architectures for the processing of multispectral images. The algorithms were developed for assessment of skin lesions in the visible and near-infrared wavelength range (l = 350 – 900 nm). However, both the hardware and software/deep learning algorithms developed in this project can be used for various inspections in agri and food such as fruits and the leaves of plants.

The absorption spectra of skin chromophores have been used to calculate the amount of local presence of blood and pigment underneath the skin. A multitude of machine learning architectures was tested to evaluate the added value of the chromophore maps for the assessment of pigmented lesions.

Chromophore maps for the assessment of pigmented lesions

The classification results showed that the inclusion of the chromophore maps together with the original RGB images delivered a better classification accuracy.

 Different deep learning architectures to classify pigmented lesions.

Discover more and watch this video:

Please reach out to us if you are interested in:

• competitor analysis & IP landscaping for product development
• design of a custom sensor solution (either custom made or with off-the-shelf components)
• a technology provider in the field of machine learning for sensing systems
• skin sensing
• agri and food inspection solutions
• deep learning / algorithm development
• sensor systems and signal processing know-how
• development or enhancement of multispectral imaging platforms