In this study, we found no difference in the diagnostic performance of the linear-array, curved-array, and phased-array transducer. The diagnostic performance was very good: sensitivity, specificity, PPV, and NPV were all between 96.4 and 98.8%. The observers needed an additional .51 s to reach a final diagnosis with the phased-array transducer compared to the linear-array transducer. A final diagnosis was reached much faster when lung sliding was present regardless of the transducer type.
The image quality scored by the observers on a 1–5 scale was significantly different between all three transducers; the linear-array transducer achieved the best scores, the phased-array transducer the worst. Moreover, the image quality was too bad to reach a diagnosis in eight phased-array transducer clips and two linear-array transducer clips.
These findings suggest that the actual diagnostic performance of the three transducers for pneumothorax is comparable. However, these experienced observers perceived the best image quality and needed the least amount of time when they judged the linear-array transducer clips. Based on these findings, the linear-array transducer might qualify as the preferred transducer for lung ultrasonography. However, the transducer choice may depend on more important factors such as the intended gamut of indications US is used for and whether the machine will be equipped with one or more transducers. In a single transducer setup, the best choice is probably a curved-array or a phased-array transducer to be able to evaluate both the abdomen and pericardium, in addition to the chest.
To our knowledge, there are no studies that have compared US transducers for diagnosing pneumothorax in a similar study design.
In a study with a comparable design, but not focused on pneumothorax, the authors compared a 10–5 MHz and a 14–5 MHz linear-array transducer for a wide array of emergency department point-of-care ultrasound indications [15]. However, lung ultrasonography was discussed only briefly. Overall, their observers most frequently preferred the 10–5 MHz transducer over the 14–5 MHz transducer.
In another study, the investigators compared a 5–10 MHz linear-array and a 1–5 MHz phased-array sector transducer in 55 patients scheduled for a thoracic-computed tomography (CT) scan [16]. The authors evaluated the diagnostic performance for pneumothorax, pleural effusion, consolidation, and interstitial syndrome. In six patients with a pneumothorax, confirmed with CT, they found that the linear-array transducer performed best with a sensitivity and specificity of 83% and 100%, respectively. The phased-array transducer showed a sensitivity and specificity of 67% and 100%. Sensitivity of both auscultation and chest radiography was only 50%. In our study, the gold standard was a thoracoscopically induced and confirmed pneumothorax. Because we assessed 495 ultrasound clips showing pneumothorax, the diagnostic performance we found is more reliable.
We hypothesized that the linear-array transducer would have the best diagnostic performance. This study, however, showed no difference in diagnostic accuracy between the transducers.
The Nijmegen HEMS introduced prehospital ultrasonography to the Netherlands in 2006 and used a phased-array transducer ever since. Only years later, a linear-array transducer was added. A curved-array transducer has never been used. This history might explain the transducer preferences of the observers and the high and equal diagnostic performance between transducers.
Although diagnostic performance is equal, we recommend the linear-array transducer for (prehospital) lung ultrasonography. The diagnosis is made faster and with a better image quality. These are important advantages in the dynamic prehospital environment, HEMS physicians encounter challenges such as time pressure, limited working space, residual clothing, defibrillator pads, and Velcro® straps. Most importantly, the interpretation of US images may be hampered by sunlight or precipitation. When the HEMS physicians have the best possible image quality, they can better deal with these factors and do the best possible for our patients.
Furthermore, when the linear-array transducer is installed as the default transducer, it may have additional advantages. It is the preferred transducer for vascular access and assessment of the airway and endotracheal tube position [15]. These matters often take precedence over detailed assessment of breathing, although it may be of vital importance to be informed about a significant pneumothorax before airway management is commenced.
In addition, the linear-array transducer is superior for ultrasound-guided regional anesthesia (UGRA) in severely injured or trapped extremities and for optic nerve sheath diameter (ONSD) measurements in traumatic brain injury (TBI) patients [17]. For abdominal ultrasound and echocardiography, however, the phased-array or curved-array transducer is still invaluable.
The observers were able to successfully assess the video clips of normal ventilation and pneumothorax without having access to the US machine or the patient. This situation is comparable to a telemedicine setup in which the US operator could be at a different physical location than the observer of the images. Therefore, we agree that lung ultrasound can be successfully used in telemedicine setups [18].
Strengths and limitations
We chose a unique approach to select VATS patients with a freshly induced and visually confirmed pneumothorax as the gold standard. In addition, we included the video clips of the same patients with normal anatomy before surgery. Another unique aspect was the cropping of the video clips thus transducers could not be identified by any text or image or sector shape.
A limitation of this study is that it might be underpowered because we could not demonstrate a difference in diagnostic performance between transducers. It could also mean the difference is close to none.
Another limitation is that we informed the observers that all patients were ventilated similarly and that besides a pneumothorax in half of the video clips, no other pathology was present. This could be an advantage for them judging the clips and might have resulted in an overestimation of the diagnostic performance and time needed. The performance could have been even better when we acquired M-mode clips looking for lung pulse or clips that included the lung point [14].
Conversely, most observers were uncomfortable with the fact that they had to assess video clips and that they were therefore unable to reposition or adjust the transducer, adjust the gain or depth, or compare with the contralateral chest. In addition, it was regarded a disadvantage that no additional clinical parameters were provided. The setting of pulmonary surgery introduced some minor challenges. In some clips, lung sliding was minimal, probably due to the lung-protective ventilator settings. B-lines were still present, obviously. In contrast to the normal clips, the pneumothorax clips were recorded with the transducer wrapped in a sterile cover. In theory, this might result in a slightly degraded US image.
Image quality might be overstated in video clips in which the diagnosis was made fast and perceived to be easy. Those clips might be scored good quality because they were ‘easy’ to assess.
A suggestion for future studies comparing ultrasound transducers might be to include subjects of all BMIs to better represent the general population of critically ill and injured patients.