Home // BIOTECHNO 2024, The Sixteenth International Conference on Bioinformatics, Biocomputational Systems and Biotechnologies // View article

Automated Lung Segmentation: Assessing LungQuant's Efficacy and Robustness

Authors:
Arman Zafaranchi
Francesca Lizzi
Alessandra Retico
Camilla Scapicchio
Maria Evelina Fantacci

Keywords: LungQuant; Lung Segmentation; COVID-19; Lesion Segmentation

Abstract:
Nowadays, automatic deep learning-based systems in medical fields play pivotal role in aiding experts with diagnosing diseases through the analysis of medical images and signals. LungQuant, a sophisticated automatic deep learningbased system, specializes in the segmentation of lung structures and lesions. The algorithm comprises two main phases of lung segmentation and lesion segmentation of COVID-19 cases. This study meticulously evaluates the performance of the LungQuant algorithm in its initial phase, leveraging diverse datasets, namely Luna-16 and COVID-19 CT scans. LungQuant demonstrates robust performance in lung segmentation with a Dice Coefficient Similarity (DCS) average of 90% and 88% with Luna-16 and COVID-19 ct datasets, respectively. The study explores various pre-processing techniques, including Anisotropic diffusion and Contrast Limited Adaptive Histogram Equalization (CLAHE), to understand their impact on the algorithm’s initial phase. While the outcomes show promising results, future enhancements may include advanced deep networks and further refinement of preprocessing methodologies. The study contributes insights into the practical considerations of noise reduction and edge preservation, underscoring the importance of a stable initial phase in ensuring overall algorithmic efficacy.

Pages: 1 to 2

Copyright: Copyright (c) IARIA, 2024

Publication date: March 10, 2024

Published in: conference

ISSN: 2308-4383

ISBN: 978-1-68558-137-4

Location: Athens, Greece

Dates: from March 10, 2024 to March 14, 2024