Detection and Classification of Bronchiectasis Through Convolutional Neural Networks

Abstract

Purpose: Bronchiectasis is a chronic disease characterized by an irreversible dilatation of bronchi leading to chronic infection, airway inflammation, and progressive lung damage. Three specific patterns of bronchiectasis are distinguished in clinical practice: cylindrical, varicose, and cystic. The predominance and the extension of the type of bronchiectasis provide important clinical information. However, characterization is often challenging and is subject to high interobserver variability. The aim of this study is to provide an automatic tool for the detection and classification of bronchiectasis through convolutional neural networks.Materials and Methods: Two distinct approaches were adopted: (i) direct network performing a multilabel classification of 32×32 regions of interest (ROIs) into 4 classes: healthy, cylindrical, cystic, and varicose and (ii) a 2-network serial approach, where the first network performed a binary classification between normal tissue and bronchiectasis and the second one classified the ROIs containing abnormal bronchi into one of the 3 bronchiectasis typologies. Performances of the networks were compared with other architectures presented in the literature.Results: Computed tomography from healthy individuals (n = 9, age = 47 ± 6, FEV 1 %pred = 109 ± 17, FVC%pred = 116 ± 17) and bronchiectasis patients (n = 21, age = 59 ± 15, FEV 1 %pred = 74 ± 25, FVC%pred = 91 ± 22) were collected. A total of 19,059 manually selected ROIs were used for training and testing. The serial approach provided the best results with an accuracy and F1 score average of 0.84, respectively. Slightly lower performances were observed for the direct network (accuracy = 0.81 and F1 score average = 0.82). On the test set, cylindrical bronchiectasis was the subtype classified with highest accuracy, while most of the misclassifications were related to the varicose pattern, mainly to the cylindrical class. Conclusion:The developed networks accurately detect and classify bronchiectasis disease, allowing to collect quantitative information regarding the radiologic severity and the topographical distribution of bronchiectasis subtype.

Do Redox Balance and Inflammatory Events Take Place in Mild Bronchiectasis? A Hint to Clinical Implications

“…Exclusion criteria for all the patients and control subjects included other chronic cardiovascular or respiratory disorders, chronic metabolic diseases, signs of severe bronchial inflammation and/or infection, current or recent invasive mechanical ventilation, long-term oxygen therapy, and poor collaboration. Most of the patients recruited for the purpose of the investigation had a mild-to-moderate disease severity on the basis of lung function impairment, disease scoring using several indices, and radiologic extension [ 22 , 24 , 25 , 26 , 27 ]. All the patients were stable at the time of study entry.…”

Profile of Clinical and Analytical Parameters in Bronchiectasis Patients during the COVID-19 Pandemic: A One-Year Follow-Up Pilot Study

“…Exclusion criteria for all the patients included other chronic cardiovascular or respiratory disorders, acute and chronic respiratory failure, chronic metabolic diseases, signs of severe bronchial inflammation and/or infection, current or recent invasive mechanical ventilation, long-term oxygen therapy, and poor collaboration. The majority of the patients recruited for the purpose of the investigation had a mild-to-moderate disease severity according to lung function impairment [ 21 ], disease severity scores, and radiological extension [ 1 , 22 , 23 , 24 , 25 ]. All the patients were stable: no acute exacerbations in the last four weeks prior to study entry.…”

Diagnosis and Initial Investigation of Bronchiectasis

“…Characterization of bronchiectasis is subject to high interobserver variability and recent literature suggested a possible role of automatic tools for the detection and classification of bronchiectasis through convolutional neural networks. 69 Aliboni et al developed networks able to accurately collect quantitative information regarding the radiologic severity of bronchiectasis and the topographical distribution of bronchiectasis subtype. 69 The implementation of automated image analysis systems in both clinical practice and research requires standardization of CT protocols and of lung volume during chest CT acquisition, as well as age-and sex-references values.…”

“…69 Aliboni et al developed networks able to accurately collect quantitative information regarding the radiologic severity of bronchiectasis and the topographical distribution of bronchiectasis subtype. 69 The implementation of automated image analysis systems in both clinical practice and research requires standardization of CT protocols and of lung volume during chest CT acquisition, as well as age-and sex-references values. 41 Finally, radiology reports do not contain quantitative outcome data.…”