Rethinking low-cost microscopy workflow: Image enhancement using deep based Extended Depth of Field methods

• DOI: 10.1016/j.iswa.2022.200170
• Publication Date: 2023-01-03T18:14:53Z
• Authors: Tomé Albuquerque; Luís Rosado; Ricardo Cruz; Maria João M. Vas...; Tiago Oliveira; Jaime S. Cardoso
• Keywords: RGB color model; Depth of field

Microscopic techniques in low-to-middle income countries are constrained by the lack of adequate equipment and trained operators. Since light microscopy delivers crucial methods for diagnosis screening numerous diseases, several efforts have been made scientific community to develop low-cost devices such as 3D-printed portable microscopes. Nevertheless, these present some drawbacks that directly affect image quality: capture samples is done via mobile phones; more affordable lenses usually used, leading poorer physical properties images with lower depth field; misalignments microscopic set-up regarding optical, mechanical, illumination components frequent, causing distortions chromatic aberrations. This work investigates pre-processing tackle presented issues proposed a new workflow microscopy. Additionally, two deep learning models based on Convolutional Neural Networks also (EDoF-CNN-Fast EDoF-CNN-Pairwise) generate Extended Depth Field (EDoF) images, compared against state-of-the-art approaches. The were tested using different datasets cytology images: public Cervix93 dataset has publicly available containing captured μSmartScope. Experimental results demonstrate can achieve performance when generating EDoF from