Know Your Enemy--Errors in Veterinary Radiology
World Small Animal Veterinary Association World Congress Proceedings, 2008
John P. Graham, MVB, MSc, DVR, MRCVS, DACVR, DECVDI
Maitland, FL, USA

Introduction

Interpreting diagnostic images from any modality requires a combination of skill and experience. The skill component can be obtained by training and study while experience can only be obtained over time. Regardless of the diagnostic imaging modality, a consistent systematic approach to interpretation is essential to accurate analysis and reliable diagnoses.

Radiological interpretation is based upon detecting changes or alternations from normal. Changes in size, shape, margin, position, number, symmetry and opacity are evaluated. The same process may be applied to any imaging modality merely substituting opacity with echogenicity in ultrasound, attenuation in computed tomography, signal intensity in magnetic resonance imaging and radiopharmaceutical uptake in nuclear medicine. A thorough analysis of the images presented is essential to an accurate diagnosis. The most reliable approach is to use a combination of two techniques. Firstly, one should evaluate all the body parts or components in the image in sequence. The observer should develop a checklist and always follow the list. This approach is most useful for complex images such as thoracic radiographs and abdominal ultrasound. Secondly, one should consider what diseases may affect a particular body part and evaluate the image for evidence of any lesions. This is most useful in evaluating the musculoskeletal system, where structures are less complex. However, one must be careful not to develop tunnel vision or search for what one thinks is the most likely diagnosis.

Lesion Detection

The process of lesion detection can be divided into three phases: fixation, recognition and diagnosis. As the eye scans an image, a vast amount of information is presented to the brain. Fixation occurs when the eye focuses on a portion of the image. Factors that influence how this data is perceived by the brain include lighting conditions, image sharpness and contrast. Many optical illusions are based on the tendency of the brain to distort data presented to it by the eyes. In other words, what the eye sees is not always what the brain sees and this occurs frequently when viewing radiographs. Once an image has been presented to the brain the second phase, lesion recognition, depends upon a process of comparing this image with the expected normal appearance. This phase is affected by the knowledge and experience of the observer, in effect the quality and quantity of stored normal images in the brain. Other factors which may influence this include bias or prejudice on the part of the observer and clinical data. If the brain recognizes a lesion, the third phase is interpretation. In this phase, the brain determines whether the lesion is real or not and what, if any, significance to ascribe to it. The interpretation phase may be affected by factors such as clinical suspicions or prejudices and observer experience.

Technique

Unfortunately this is a common source of error in veterinary imaging. Errors arise from incomplete studies or from studies which are not of diagnostic quality. Such errors may result in both the failure to detect lesions or over interpretation leading to false positive diagnoses.

Complacency

Errors of this type may be the result of a lazy or lackadaisical approach to the case but may have many other causes. A common form of this class of error in veterinary imaging is the observer failing to detect the absence of a normal structure. Disease is expected to produce lesions which one must find, recognise and assess, when the effect of the pathology is to remove something rather than create something new, it is all too easy to overlook. A real abnormality may be determined to be the cause of the patient's signs even though it is not. Similarly, normal anatomic or physiologic variations may be interpreted as pathological. Bias or prejudice on the part of the reader may lead to misdiagnosis. This is particularly hard to avoid for the general practitioner who will have formed an opinion of a case based on history and physical examination finding before reviewing radiographs. An erroneous diagnosis may also occur if one has seen several cases of a particular type recently and is expecting to see more. Many diseases have specific geographic distributions which may lead to over-diagnosis because the disease is perceived to be common. A diagnosis may also be missed because a disease is 'exotic' and does not occur or is rare in a geographical area. Sometimes, for reasons that can defy explanation, an obvious lesion is overlooked or dismissed on initial review of radiographs. Possible causes include dismissing the lesion because the diagnosis could not be that simple or concentrating on searching for small lesions and simply missing the big one.

Faulty Reasoning

When one evaluates images, other clinical data are used to reach a conclusion. Errors in evaluating a lesion can occur because one has been supplied with erroneous or misleading information. Too much significance can be attached to historical or clinical data resulting in over-reading of imaging lesions. This may also occur if the patient's history or physical examination is incomplete or when a colleague communicates only a part of the information pertaining to a patient. Errors can occur when one uses only the data which supports the diagnosis being considered while other data is neglected. A previous diagnosis may be erroneously assumed to be the cause of current clinical signs, while a new diagnosis is overlooked. Diagnosis by the use or rules, such as numerical formulas to determine normalcy, may also lead to errors. One should be wary of diagnostic criteria or tests which incorporate the terms 'always' or 'never'.

Lack of Knowledge

Limited knowledge may result in failure to recognise a lesion as such. This may be the result of lack of experience. Errors of this type also occur with uncommon lesions which are not known to the observer. In veterinary imaging this type of error is especially common when one is learning a new imaging modality such as ultrasound. The body of knowledge in veterinary medicine expands continuously and one cannot hope to know everything. Conversely, as time progresses, so does forgetfulness leading to mistakes.

Under-Reading

This class of error may occur when a lesion is not detected or not ascribed appropriate significance. A common phenomenon is to cease searching for abnormalities having found one, two or more major lesions. In such cases, smaller, and perhaps more significant, lesions are not found. This is sometimes referred to as 'satisfaction of search'. If a diagnosis may have severe consequences for the patient, one is likely to under-read the images if one is not completely confident. Lesions are frequently overlooked when assessing complex body parts such as the thorax. Similarly, lesions may be overlooked in the relatively overexposed parts of a film, unless these parts are examined using a hot light. These errors are frequently propagated when a patient has serial imaging studies. If a lesion is overlooked on the first study, later studies are evaluated for progression or resolution of the abnormalities seen on the first study and the undetected abnormalities are ignored. Alternatively, the earlier films or report may not be read and the lesion(s) missed. Distraction is another cause for under-reading and perhaps one of the more common in busy practices.

Poor Communication

Errors of this nature most commonly occur in reports prepared by radiologists for other specialists or general practitioners. Terms such as 'consistent with', 'most likely', 'possible', 'probable', 'suspect' etc. have quite different meanings and convey quite different levels of confidence for different individuals. Incorrect or vague anatomical descriptions are a common source of error and confusion. Such errors also occur in medical records in general practice. Indeed, the legibility, or rather illegibility, of hand written records lends itself to a multitude of potential errors. Even when carefully written, reports can incorporate errors such as incorrect counts (especially ribs or vertebrae) or body parts (right and left, transposing front and hind limb bones).

Speaker Information
(click the speaker's name to view other papers and abstracts submitted by this speaker)

John P. Graham, MVB, MSc, DVR, MRCVS, DACVR, DECVDI
Maitland, Florida, USA


MAIN : Diagnostic Imaging : Errors in Radiology
Powered By VIN
SAID=27