Lung ultrasound (LUS) is a relatively novel application of ultrasound technology, which is increasingly expanding since the nineties. However, contrary to standard ultrasound imaging, which was primarily developed for imaging noninvasively the anatomy of internal body parts, LUS is mainly based on the visual interpretation of imaging artifacts. Among which, the so-called vertical artifacts are particularly important as they correlate with various pathologies. The main limitations associated with this type of pattern analysis remain its subjectivity and limited reproducibility. Moreover, the understanding and exploitation of the mechanisms behind the genesis of vertical artifacts are just in their beginnings. In this context, the most diffused and utilized strategies in LUS analyses are the so-called semi-quantitative techniques, which are based on the visual interpretation of LUS patterns, where a score is assigned based on visual interpretation of LUS patterns, which correlate with the state of lung. However, these techniques are strongly operator dependent. To this end, the use of artificial intelligence (AI) to automatically score LUS data could be instrumental to reduce the subjectivity in the evaluation of LUS patterns. For this reason, as a first novel contribution, we proposed a technique to automatically classify LUS videos by means of an aggregation strategy based on a deep learning (DL) frame-based classification. However, given the strong subjectivity of the task, it is not reliable to expect levels of agreement between AI and human operators at video or frame level around 90-100%. Indeed, the use of AI algorithms could lead to more reproducible analyses but cannot completely avoid subjectivity, as AI training remains based on the subjective labeling performed by clinicians. Another important aspect to be considered in semi-quantitative techniques is the proper definition and standardization of acquisition protocols. As an example, the number and spatial distribution of areas of the chest to be scanned are often defined arbitrarily and not following an evidence-based approach. For this reason, after having proposed (in a study of March 2020 I coauthored) a standardized imaging protocol specifically designed for the coronavirus disease 2019 (COVID-19) patients based on 14 scanning areas, we evaluated the impact of changing the scanning areas on the evaluation of COVID-19 (second novel contribution) and post-COVID-19 (third novel contribution) patients. In addition, to properly define the imaging settings to be used in a LUS examination and, in parallel, to develop quantitative LUS techniques specifically designed for lung, the mechanisms behind the genesis of vertical artifacts (whose comprehension is just in its beginnings) should be deeply understood. To better understand the dependence of vertical artifacts on imaging parameters, we performed two experimental studies (fourth and fifth novel contributions), where we assessed the dependence of vertical artifacts' intensity on different imaging parameters. On one hand, the presented results showed how there exist different confounding factors (e.g., focal point and angle of incidence of ultrasound beam) that should be reduced when developing a LUS approach. On the other hand, the results showed how a frequency characterization of vertical artifacts could be exploited to develop a LUS quantitative approach, as these artifacts seemed to be associated with specific resonance phenomena. Specifically, the acoustic traps' theory suggested that vertical artifacts originate from multiple reflections of ultrasound waves trapped within channels that can form between alveoli when lung tissue becomes pathological. By exploiting this concept, the frequency characterization of these artifacts could be used to indirectly estimate the size of acoustic channels (or traps). To further evaluate the possibility to estimate these channels' size with a multi-frequency approach, we performed a numerical study with the k-wave MATLAB toolbox (sixth novel contribution). The main advantage of in silico studies consists of the possibility to control the disposition of alveoli, which can be located at precise distances between each other. Therefore, with this kind of studies, it is possible to look for a correlation between the vertical artifacts' intensity as a function of frequency and the alveolar disposition. In the final novel contribution of this thesis, we performed a clinical study in humans showing the potentiality to exploit a quantitative multi-frequency approach to differentiate patients affected by pulmonary fibrosis (PF) from patients with other lung pathologies. Specifically, the frequency characterization of vertical artifacts along with their intensity was able to differentiate patients with PF with a specificity and sensitivity of 92%. In conclusion, quantitative approaches represent the future of LUS, as they could provide a physical metric able to characterize the lung surface by applying an acquisition technique specifically designed for the lung. Nevertheless, to develop these techniques, the genesis of vertical artifacts needs to be more deeply investigated and understood by means of controlled in vitro and in silico studies. In the meantime, semi-quantitative approaches based on image analysis techniques should be exploited to estimate the state of lungs by detecting and recognizing specific LUS patterns that do signal different levels of aeration. However, to reduce the impact of confounding factors, the standardization of the imaging protocols and scoring systems is essential.

Development of Lung Ultrasound Quantitative Approaches and Automatic Semi-Quantitative Strategies: In Silico, In Vitro, and Clinical Studies / Mento, Federico. - (2022 Nov 02), pp. 1-137.

Development of Lung Ultrasound Quantitative Approaches and Automatic Semi-Quantitative Strategies: In Silico, In Vitro, and Clinical Studies

Mento, Federico
2022-11-02

Abstract

Lung ultrasound (LUS) is a relatively novel application of ultrasound technology, which is increasingly expanding since the nineties. However, contrary to standard ultrasound imaging, which was primarily developed for imaging noninvasively the anatomy of internal body parts, LUS is mainly based on the visual interpretation of imaging artifacts. Among which, the so-called vertical artifacts are particularly important as they correlate with various pathologies. The main limitations associated with this type of pattern analysis remain its subjectivity and limited reproducibility. Moreover, the understanding and exploitation of the mechanisms behind the genesis of vertical artifacts are just in their beginnings. In this context, the most diffused and utilized strategies in LUS analyses are the so-called semi-quantitative techniques, which are based on the visual interpretation of LUS patterns, where a score is assigned based on visual interpretation of LUS patterns, which correlate with the state of lung. However, these techniques are strongly operator dependent. To this end, the use of artificial intelligence (AI) to automatically score LUS data could be instrumental to reduce the subjectivity in the evaluation of LUS patterns. For this reason, as a first novel contribution, we proposed a technique to automatically classify LUS videos by means of an aggregation strategy based on a deep learning (DL) frame-based classification. However, given the strong subjectivity of the task, it is not reliable to expect levels of agreement between AI and human operators at video or frame level around 90-100%. Indeed, the use of AI algorithms could lead to more reproducible analyses but cannot completely avoid subjectivity, as AI training remains based on the subjective labeling performed by clinicians. Another important aspect to be considered in semi-quantitative techniques is the proper definition and standardization of acquisition protocols. As an example, the number and spatial distribution of areas of the chest to be scanned are often defined arbitrarily and not following an evidence-based approach. For this reason, after having proposed (in a study of March 2020 I coauthored) a standardized imaging protocol specifically designed for the coronavirus disease 2019 (COVID-19) patients based on 14 scanning areas, we evaluated the impact of changing the scanning areas on the evaluation of COVID-19 (second novel contribution) and post-COVID-19 (third novel contribution) patients. In addition, to properly define the imaging settings to be used in a LUS examination and, in parallel, to develop quantitative LUS techniques specifically designed for lung, the mechanisms behind the genesis of vertical artifacts (whose comprehension is just in its beginnings) should be deeply understood. To better understand the dependence of vertical artifacts on imaging parameters, we performed two experimental studies (fourth and fifth novel contributions), where we assessed the dependence of vertical artifacts' intensity on different imaging parameters. On one hand, the presented results showed how there exist different confounding factors (e.g., focal point and angle of incidence of ultrasound beam) that should be reduced when developing a LUS approach. On the other hand, the results showed how a frequency characterization of vertical artifacts could be exploited to develop a LUS quantitative approach, as these artifacts seemed to be associated with specific resonance phenomena. Specifically, the acoustic traps' theory suggested that vertical artifacts originate from multiple reflections of ultrasound waves trapped within channels that can form between alveoli when lung tissue becomes pathological. By exploiting this concept, the frequency characterization of these artifacts could be used to indirectly estimate the size of acoustic channels (or traps). To further evaluate the possibility to estimate these channels' size with a multi-frequency approach, we performed a numerical study with the k-wave MATLAB toolbox (sixth novel contribution). The main advantage of in silico studies consists of the possibility to control the disposition of alveoli, which can be located at precise distances between each other. Therefore, with this kind of studies, it is possible to look for a correlation between the vertical artifacts' intensity as a function of frequency and the alveolar disposition. In the final novel contribution of this thesis, we performed a clinical study in humans showing the potentiality to exploit a quantitative multi-frequency approach to differentiate patients affected by pulmonary fibrosis (PF) from patients with other lung pathologies. Specifically, the frequency characterization of vertical artifacts along with their intensity was able to differentiate patients with PF with a specificity and sensitivity of 92%. In conclusion, quantitative approaches represent the future of LUS, as they could provide a physical metric able to characterize the lung surface by applying an acquisition technique specifically designed for the lung. Nevertheless, to develop these techniques, the genesis of vertical artifacts needs to be more deeply investigated and understood by means of controlled in vitro and in silico studies. In the meantime, semi-quantitative approaches based on image analysis techniques should be exploited to estimate the state of lungs by detecting and recognizing specific LUS patterns that do signal different levels of aeration. However, to reduce the impact of confounding factors, the standardization of the imaging protocols and scoring systems is essential.
XXXV
2021-2022
Ingegneria e scienza dell'Informaz (29/10/12-)
Information and Communication Technology
Demi, Libertario
no
Inglese
File in questo prodotto:
File Dimensione Formato  
Mento_Federico_Thesis.pdf

embargo fino al 01/08/2024

Tipologia: Tesi di dottorato (Doctoral Thesis)
Licenza: Tutti i diritti riservati (All rights reserved)
Dimensione 19.37 MB
Formato Adobe PDF
19.37 MB Adobe PDF   Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11572/354762
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact