On the impact of pixel resolution on automated scoring of lung ultrasound images from coronavirus disease 2019 patients

• DOI: 10.1121/10.0010819
• Publication Date: 2022-05-09T18:08:52Z
• Authors: Umair Khan; Federico Mento; Lucrezia N. Giacomaz; Riccardo Trevisan; Andrea Smargiassi; Riccardo Inchingolo; Tiziano Perrone; Libertario Demi
• Keywords: Resampling

With the outbreak of COVID-19 pandemic, remote diagnosis, patient monitoring, collection, and transmission health data from electronic devices are rapidly taking its share in sector. These are, however, limited on resources like energy, memory, processing power. Consequently, it is highly relevant to investigate how minimize size data, keeping intact information content. The objective this study to, thus, observe impact pixel resolution automated scoring by DL algorithms for LUS videos. First, 448 videos 20 patients were normalized a common resolution, i.e., largest found over dataset (841 pixels/cm2). Next, was further reduced factor 2 resampling 210 pixels/cm2. Original, re-sampled evaluated using algorithm [Roy et al., IEEE Trans. Med. Imaging 39, 2676–2687 (2020)]. At frame level, resampled videos, level agreement results with original 93.2% 86.6%, respectively. Similar performance at video 95.75% 85.93%, showed that significant reduction low variation observed.