key: cord-0028973-l0a8rsuf authors: Di Serafino, Marco; Iacobellis, Francesca; Schillirò, Maria Laura; D’auria, Divina; Verde, Francesco; Grimaldi, Dario; Dell’Aversano Orabona, Giuseppina; Caruso, Martina; Sabatino, Vittorio; Rinaldo, Chiara; Guerriero, Pasquale; Cantisani, Vito; Vallone, Gianfranco; Romano, Luigia title: Common and Uncommon Errors in Emergency Ultrasound date: 2022-03-04 journal: Diagnostics (Basel) DOI: 10.3390/diagnostics12030631 sha: 49dd4c757261599eb1249b0e4a501721fe4a99ed doc_id: 28973 cord_uid: l0a8rsuf Errors in emergency ultrasound (US) have been representing an increasing problem in recent years thanks to several unique features related to both the inherent characteristics of the discipline and to the latest developments, which every medical operator should be aware of. Because of the subjective nature of the interpretation of emergency US findings, it is more prone to errors than other diagnostic imaging modalities. The misinterpretation of US images should therefore be considered as a serious risk in diagnosis. The etiology of error is multi-factorial: it depends on environmental factors, patients and the technical skills of the operator; it is influenced by intrinsic US artifacts, poor clinical correlation, US-setting errors and anatomical variants; and it is conditioned by the lack of a methodologically correct clinical approach and excessive diagnostic confidence too. In this review, we evaluate the common and uncommon sources of diagnostic errors in emergency US during clinical practice, showing how to recognize and avoid them. Ultrasound (US) is a routine imaging procedure that is frequently the initial course of action for emergency care and plays a critical role in the diagnosis of patients in the emergency room. Because of the subjective nature of the interpretation of emergency US findings, it is more prone to errors than other diagnostic imaging modalities. The misinterpretation of US images should therefore be considered as a serious risk in diagnosis [1] . Furthermore, the emergency room setting is an environment with a particular risk for malpractice claims [1, 2] . US diagnostic errors in the emergency setting can be classified into three groups as follows: Errors dependent on: 1. Environmental factors; 2. Patients; 3. Technical skills of the operator. The accuracy of the US examination is directly related to the operator's skill, training, and experience. The medical operator's responsibilities include fully exploiting the diagnostic capability of US, knowing what to look for, and having the competence to interpret the findings based on understanding the physiology and pathological changes of the examined organs [1, 2] . Furthermore, the interpretation of US examinations requires a subjective judgement that is heavily influenced by the operator and cannot be achieved only on the basis of the images. Unlike other imaging procedures, US examination is a real-time test that cannot be delegated, although in many countries across Europe and North America it is usually performed by a specialized non-medical technical operator such as a sonographer, followed by a delayed report by a physician; in this way it loses its special value-facilitating real-time imaging. The increased number of devices and operators devoted to US influence a large number of examinations; interpretational doubts create the need for further diagnostic imaging confirmations and a consequent increase in healthcare costs, diagnostic delays and medico-legal disputes [3] . US is often wrongly considered as an imaging modality that needs short training, in the mistaken belief that it is, due to the widespread diffusion of the method, a very simple procedure to perform. Errors of ignorance are due to inadequate knowledge, whereas errors of implementation occur during the application of knowledge [3, 4] . Furthermore, knowing the intrinsic US limits also represents at the same time the achievement of a good awareness of US imaging, the so-called "right measure", so as to be able to better combine this imaging modality with others in order to achieve good diagnostic performance, and at the same time, make US a concrete diagnostic opportunity. The training and advanced study of US techniques, thorough knowledge of human anatomy, the study of US artifacts, and ultrasonographic semiotics are all crucial tools, and according to the increase in the supply of US performance and its clinical use in emergency scenarios, the need to implement universally accepted guidelines of US training courses appears imperative. Bad artifacts; 2. Chest artifacts with no clinical context; 3. US-setting errors; 4. Anatomy and anatomical variants. Image artifacts are commonly found with US and can be confusing to the physician who encounters them. Some artifacts may be avoidable and result from an improper scanning technique. Other artifacts are generated by the physical limitations of the mode [5] . In clinical practice, side lobe artifacts, the mirror imaging effect, speed-displacement artifacts, refraction and reverberation artifacts, image-adaptation artifacts, and anisotropy are all common [5] . These are defined as follows: A powerful reflecting surface settled out of the main US beam generates echoes that the transducer may detect (Figure 1 ). Image artifacts are commonly found with US and can be confusing to the ph who encounters them. Some artifacts may be avoidable and result from an improp ning technique. Other artifacts are generated by the physical limitations of the m In clinical practice, side lobe artifacts, the mirror imaging effect, speed-displaceme facts, refraction and reverberation artifacts, image-adaptation artifacts, and anisotr all common [5] . These are defined as follows: A powerful reflecting surface settled out of the main US beam generates echo the transducer may detect (Figure 1 ). The display assumes that the return echoes that come from the main axis are mistakenly understood as coming from the main axis itself, and theref places and duplicates the structure (multiple yellow boxes) in the context of the target imag ified from Feldman MK et al. [5] . An example is given by the "pseudo-mud" which is observed at the bottom bladder or in the gallbladder ( Figure 2 ). In this case, the correct setting of the US the focusing and the use of multiple scans allow the image quality to be improved When the US of the primary incident beam passes through a curved interfa high specular reflection capacity (e.g., the diaphragm), the structure close to this in can be reproduced on the US image both in its real position and in a symmetrical o The display assumes that the return echoes that come from outside the main axis are mistakenly understood as coming from the main axis itself, and therefore misplaces and duplicates the structure (multiple yellow boxes) in the context of the target image. Modified from Feldman MK et al. [5] . An example is given by the "pseudo-mud" which is observed at the bottom of the bladder or in the gallbladder ( Figure 2 ). In this case, the correct setting of the US image, the focusing and the use of multiple scans allow the image quality to be improved [5, 6] . When the US of the primary incident beam passes through a curved interface with high specular reflection capacity (e.g., the diaphragm), the structure close to this interface can be reproduced on the US image both in its real position and in a symmetrical opposite position, beyond the reflecting surface (e.g., mirror image of the liver, which can be mistaken for a parenchymal consolidation, or the aorta, which can produce a "ghost image") ( . Mirror-image artifacts will disappear when the reflector is scanned with oblique orientation [5, 6] . Diagnostics 2022, 12, x FOR PEER REVIEW position, beyond the reflecting surface (e.g., mirror image of the liver, which can taken for a parenchymal consolidation, or the aorta, which can produce a "ghost ( Figures 3-5 ). Mirror-image artifacts will disappear when the reflector is scann oblique orientation [5, 6] . This happens when the ultrasonic beam passes through the propagation medium at a much slower speed than expected (1540 m/s), causing the return echo to take longer to be decoded by the scan converter and to return (Figures 6 and 7) . The echoes thus appear deeper on the image than they really are due to the assumption that the length of time taken for a single round trip of an echo is related only to the distance that it travels [5] . The use of multiple scans can improve the image quality. This happens when the ultrasonic beam passes through the propagation medium at a much slower speed than expected (1540 m/s), causing the return echo to take longer to be decoded by the scan converter and to return (Figures 6 and 7) . The echoes thus appear deeper on the image than they really are due to the assumption that the length of time taken for a single round trip of an echo is related only to the distance that it travels [5] . The use of multiple scans can improve the image quality. When the US beam encounters an interface that attenuates the sound to a greater or lesser extent than in the surrounding tissue, the strength of the beam distal to this structure will be either weaker or stronger than in the surrounding field [3] . This occurs because a focus material might cause the beam to deviate from its rectilinear path at an angle that causes a deflection towards the side with the higher resistance ( Figure 8 ). Because this type of artifact can mimic a traumatic kidney rupture, it is recommended that scans are performed with coronal, axial, and sagittal pictures in order to overcome this interpretation deficiency ( Figure 9 ) [5, 6] . When the US beam encounters an interface that attenuates the sound to a greater or lesser extent than in the surrounding tissue, the strength of the beam distal to this structure will be either weaker or stronger than in the surrounding field [3] . This occurs because a focus material might cause the beam to deviate from its rectilinear path at an angle that causes a deflection towards the side with the higher resistance ( Figure 8 ). Because this type of artifact can mimic a traumatic kidney rupture, it is recommended that scans are performed with coronal, axial, and sagittal pictures in order to overcome this interpretation deficiency (Figure 9 ) [5, 6] . Because the US beam is reflected back and forth several times between two highly reflective parallel surfaces ("reverberates"), the US transducer interprets the reflected echoes as happening at deeper structures because they take longer to return to the transducer (Figures 10 and 11 ) [3] . This very common mistake can be easily solved with a change of the angle of insonation in order to avoid the reverberation between strong parallel reflectors [5, 6] . Because the US beam is reflected back and forth several times between two reflective parallel surfaces ("reverberates"), the US transducer interprets the reflec oes as happening at deeper structures because they take longer to return to the tran (Figures 10 and 11 ) [3] . This very common mistake can be easily solved with a ch the angle of insonation in order to avoid the reverberation between strong paralle tors [5, 6] . Real-time image processing methods have been developed to improve the contrast resolution of US images by focusing on texture details and reducing noise and artifacts [3] . Edge-enhancement filters, on the other hand, can produce an artifactual hypoechoic line parallel to a highly reflecting interface, simulating thin-fluid collection ( Figure 12 ) [5, 6] . In musculoskeletal applications, this is a common occurrence. It occurs when the US beam strikes a structure with a fibrillar structure (e.g., a ligament or tendon), causing the majority of the insonating sound beam to recede from the transducer. The fibrils have aberrant echogenicity, which the transducer interprets as a hypoechoic zone [5] . This anisotropic effect is dependent on the angle of the insonating beam ( Figure 13 ). To overcome this, the insonation angle must always be perpendicular to the structures to be examined [7] . Figure 13 . Normal Achilles tendon (a) and anisotropy-related artifact seen at the same tendon (b) that appears hypoechoic (arrows) due to an incorrect angle of the transducer. Intrinsic or patient-related artifacts (such as subcutaneous emphysema), the coexistence of several pathologies that increase or decrease the sub-pleural air content (such as emphysema and atelectasis, respectively), or existing fibrotic interstitial lung disease, can all be confounding factors in the interpretation of lung US (LUS) findings in the acute setting, and are heavily influenced by the patient's age [6] . LUS has intrinsic limits, and since it is based on the presence or absence of pulmonary artifacts such as A-lines or B-lines, it must be remembered that as well as recognizing them, they must necessarily be contextualized to the clinical data since they can have a dichotomous interpretation ( To avoid errors, a thorough understanding of the functionality and underlying mechanics of the US equipment is required. The goal should always be to produce the highest possible image quality by employing a check-list procedure to configure the correct system settings as well as the correct Doppler parameters, which are critical for interpreting clinical US findings ( Table 2 , Figures 14 and 15) [6, [9] [10] [11] [12] . Table 2 . Steps to achieve optimal settings for B-Mode and Doppler US. Modified from Zander D et al. [9] . 1st Transmission power Depth penetration is improved and scattering is reduced with increasing transmission power. 2nd Gain Adjust as low as possible to avoid overexposure. The adjustment is almost always performed during the execution of an exam to reduce some artifacts and increase the visibility of low-contrast sound formations with respect to the surrounding tissues. By selectively amplifying the echoes from greater depths, using the time-gain compensation (TGC) method or depth-gain compensation (DGC), equal reflectors at unequal depths are displayed as structures of equal brightness on the monitor. 3rd Frequency Pivotal element of ultrasound image quality. Adjustment can be of great help with both convex and linear transducers. This parameter affects the spatial resolution and mainly the axial resolution. Depth penetration Adjust to the structure of interest. The adjustment is almost always carried out during the execution of an exam, as it is important in US to start from the greatest possible spatiality and then move on to the detailed analysis. 5th Focal zone(s) At the level of interest or use focus. This parameter optimizes the lateral resolution. 6th Further settings Only in the case of insufficient image quality: change the pre-set, adjust the dynamic range, grey maps, persistence, and/or frame rate. 1st Pulse repetition frequency (PRF) Optimal PRF proportional to the flow to be studied and in principle must be lowered until it almost reaches the aliasing threshold. 2nd Size and position of the color box The color box must be neither too wide, too small, too long nor too deep. Some anatomical structures and variances might yield difficult-to-interpret pictures, which can lead to errors if not fully understood [6] . The most insidious in emergency situations are pseudo-splenic hematoma (Figures 16 and 17) , pseudo-collections of pleural, pericardial, peritoneal, and retroperitoneal fluids (Figures 18-21) , and pseudo-pneumothorax related to abolished lung sliding due to the lung pulse or to patient apnea [6, 7] . Others, such as hypertrophy diaphragmatic pillar (Figure 22) , bladder pseudo-masses (Figure 23) , inguinal pseudo-hernias (Figure 24) , and the rouleaux phenomena, are generic, random, and highly conditioned (Figure 25 ). Often it is the emergency condition of the clinical context itself as well as the traumatic accident that could influence interpretational doubts or over-diagnosis with the necessity of more diagnostic confirmations, involving more costly complex examinations, an increased waiting time for the final diagnosis and medico-legal disputes [3, 13] . Some anatomical structures and variances might yield difficult-to-interpret pictures, which can lead to errors if not fully understood [6] . The most insidious in emergency situations are pseudo-splenic hematoma (Figures 16 and 17) , pseudo-collections of pleural, pericardial, peritoneal, and retroperitoneal fluids (Figures 18-21) , and pseudo-pneumothorax related to abolished lung sliding due to the lung pulse or to patient apnea [6, 7] . This phenomenon depends on several items, such as the lack of a methodologically correct clinical approach, excessive diagnostic confidence generated by superficiality and/or lack of experience [3] . Due to partial visibility and poor attention or interest, these can damage the retroperitoneum or the spleen especially when carrying out focused assessment with sonography in trauma (FAST) examination ( Figure 26 ). Furthermore, poor image quality, as well as inconclusive and inadequate US reports, can often relate to errors of underestimation ( Figure 27 ) [13] [14] [15] . It is important to document in detail the US findings found, as well as to provide for their archiving safeguarding in case of future medico-legal disputes. In the same way, an accurate and detailed report emphasizing the importance of the description of pathological changes detected should not be ambiguous, but easy to understand and timely with respect to the clinical question [3] . How to Avoid Diagnostic Errors in Ultrasound: Is Artificial Intelligence the Right Way? In this new era of cutting-edge medical advances, artificial intelligence (AI) has emerged as a subset of computer science involved in human-like processes such as learning, adapting, and solving complex problems. The current branches of AI in medical imaging mainly include machine learning (ML) and deep learning (DL), consisting of algorithms that can make predictions or decision tasks without prior explicit programmed rules [15] . ML algorithms use iterative statistical learning methods from "training" data to progressively improve the model performance over time, enabling recognition patterns in large datasets and classification of instances [16] . DL is a subgroup of ML techniques, structured as artificial neural networks which consist of multi-layered networks that automatically extract features without prior labels and perform high-level tasks [17] . The application of AI models coupled to medical imaging has been embraced in US image analysis to empower the clinical-radiological workflow and to reduce US errors derived by different image variation factors such as operator-, scanner-, and patient-dependent. Of note, the development of AI-based US should assist less-experienced users with performing correct examinations, ultimately improving the clinical decision process. In this regard, recent evidence has shown image-quality improvement by using AI algorithms to enhance resolution, making the US scans both more detailed and easier to read [18] . Furthermore, AI systems have been applied to specific US image-based tasks such as disease classification, abnormality detection, image segmentation, and prognosis assessment, and for different organ systems including thyroid, breast, abdomen and pelvis, obstetrics and gynecology, heart, and the musculoskeletal system [19] . Although more clinical evidence is still necessary, the outlook of AI in US imaging remains bright. US quality is operator-dependent and subjective to interpretive error. Following specific rules concerning technique and interpretation and always integrating emergency US findings into the broader clinical context could avoid many misdiagnoses and thus enhance patient safety. Knowledge of common sites of artifacts and their typical/atypical appearances will help to prevent misinterpretation of these findings and may lead to improved diagnostic accuracy. Furthermore, the development of artificial intelligence (AI)-based US represents a useful additional tool of imaging modality. This aids the clinical decision process, making US scans both more detailed and easier to read. This article discussed the most common and uncommon diagnostic errors and mistakes in US examinations. Similarly to other diagnostic tools, it is a fact that errors may occur in the emergency room. Therefore, whether we are talking about common or intrinsic errors rather than uncommon dynamic or variable errors, the risk of a mistake remains. Besides that, the increased complexity in the interpretation of dynamic errors emphasizes the importance of proper training and credentialing for the operator. It is, however, understood that, beyond any technological progress, with a knowledge of the intrinsic limits of US and of its potential "bad" artifacts, as well as a standardized US approach, mistakes can be avoided. This helps to safeguard patient health and prevent potential medico-legal disputes associated with such mistakes. Sources of error in emergency ultrasonography Errors in imaging patients in the emergency setting A critical evaluation in the delivery of the ultrasound practice: The point of view of the radiologist Err Is Human: Building a Safer Health System Emergency ultrasound in trauma patients: Beware of pitfalls and artifacts! Artifacts in musculoskeletal ultrasonography: From physics to clinics The lung ultrasound: Facts or artifacts? In the era of COVID-19 outbreak Ultrasound image optimization ("knobology"): B-mode Ultrasound of scrotal and penile emergency: How, why and when Common pitfalls in point-of-care ultrasound: A practical guide for emergency and critical care physicians Optimizing image quality when evaluating blood flow at doppler US: A tutorial The role of point of care ultrasound in radiology department: Update and prospective. A statement of Italian college ultrasound Diagnostic accuracy of pubic symphysis ultrasound in the detection of unstable pelvis in polytrauma patients during e-FAST: The value of FAST-PLUS protocol. A preliminary experience Current applications and future impact of machine learning in radiology Current trends of artificial intelligence in cancer imaging Deep learning: A primer for radiologists Towards clinical application of artificial intelligence in ultrasound imaging Artificial intelligence in ultrasound Funding: This research received no external funding. Informed Consent Statement: Written informed consent has been obtained from the patients to publish this paper. Data sharing is not applicable. The authors declare no conflict of interest.