Dealing with distribution mismatch in semi-supervised deep learning for COVID-19 detection using chest X-ray images: A novel approach using feature densities

Saúl Calderón Ramírez, Shengxiang Yang, David A. Elizondo, Armaghan Moemeni

Abstract: Highlights•COVID-19 detection using chest X-ray images with deep learning need many labels.•Semi supervised learning can be used to increase model’s accuracy with few labels.•A distribution mismatch might exist between the labelled and unlabelled datasets.•An assessment of this problem in the context of COVID-19 detection is presented.•Two novel methods to score unlabelled data are proposed in this work.