Author + information
- Rossella Raso, PhD∗ (, )
- Gennaro Tartarisco, PhD,
- Marco Matucci Cerinic, MD, PhD,
- Giovanni Pioggia, PhD,
- Eugenio Picano, MD, PhD and
- Luna Gargani, MD
- ↵∗CNR, Institute of Clinical Physiology, Via G. Moruzzi, 1, 56124 Pisa, Italy
Lung ultrasound (LUS) has recently been proposed as a new sonographic application to image pulmonary interstitial syndrome through evaluation of B-lines, discrete laser-like vertical hyperechoic reverberation artifacts that arise from the pleural line, extend to the bottom of the screen, and move synchronously with respiration (1). Pulmonary interstitial syndrome includes both pulmonary interstitial edema, such as increased extravascular lung water, and pulmonary interstitial fibrosis (2). LUS does not require the expertise necessary for echocardiographic examination and interpretation (3) and is rapid to perform, portable, repeatable, nonionizing, and independent of cardiac acoustic windows. However, although B-lines are highly sensitive in the detection of pulmonary interstitial syndrome, they cannot be quantified exactly. Our aim was to develop a soft computing–based B-line analysis through a knowledge-based model for an objective, operator-independent, automated, and quantitative classification of the severity of pulmonary interstitial syndrome.
During LUS examinations, 96 short movies were recorded in patients with a diagnosis of heart failure and clinical and radiological signs of pulmonary interstitial edema, 95 short movies were recorded in patients with clinical and radiological signs of pulmonary fibrosis, and 62 short movies were recorded in control subjects. Commercially available 2.5- to 3.5-MHz sector transducers were used. All clips were evaluated by a physician with more than 9 years' experience in LUS (L.G.). She labeled each clip with a number from 0 to 10, according to increasing degree of pulmonary interstitial syndrome. This classification was then divided into 4 degree groups of pulmonary interstitial syndrome: absent (0 to 1 B-line), mild (2 to 4 B-lines), moderate (5 to 7 B-lines), and severe (8 to 10 B-lines). Each movie was analyzed offline with a bio-inspired algorithm to assess B-lines and extract features to train an artificial neural network. The final software was registered as copyright in January 2013 by the Società Italiana degli Autori ed Editori.
The intraobserver and interobserver discordance rates were 5.2% and 8.12%, respectively. A mismatch of ±1 B-line was evaluated as clinically equivalent. The model correctly identified normal subjects in 100% of cases and was able to discriminate the 3 levels of severity of pulmonary interstitial edema and pulmonary fibrosis. At most, 5.1% of cases of edema and 5.3% of fibrosis were misclassified with only 1 degree of disagreement (Figure 1).
Our results show that a soft computing–based B-line analysis was able to objectively classify the degree of pulmonary edema and pulmonary fibrosis with high feasibility and very high agreement over 4 levels of severity. Our results are consistent with those of a recently published paper by Brattain et al. (4), which underlines the need for a computerized diagnostic decision support system based on sonographic video processing to aid nonexpert users. The added value of our findings lies in the use of an artificial neural network pattern recognition procedure to create a new decision support system tool for processing and computer-aided analysis of LUS images. Moreover, we analyzed B-line videos derived from pulmonary congestion and pulmonary fibrosis, with very similar results between the 2 etiologies.
This software could be a robust approach for developing a portable device for the individualized and automatic detection of pulmonary interstitial edema or fibrosis, which could have a high impact on public health. We are moving toward pervasive healthcare systems in human-oriented environments. Because telemedicine is being increasingly used in many different scenarios for patient monitoring and specialist consultations (5), the development of operator-independent, computer-based systems to support clinical judgment contributes to promoting radical changes toward a patient-centric healthcare environment.
Please note: This study was supported by a grant from the Regione Toscana, Italy (Regional Health Research Program).
- American College of Cardiology Foundation
- Barskova T.,
- Gargani L.,
- Guiducci S.,
- et al.
- Brattain L.J.,
- Telfer B.A.,
- Liteplo A.S.,
- Noble V.E.
- Topol E.