Christopher R. Lamb, MA, VetMB, DACVR, DECVDI, ILTM, MRCVS
In an ideal world, the results of diagnostic tests would always be true. Ideal diagnostic tests would always give a positive result in patients with the disease, and would always be negative in unaffected patients. Unfortunately, the practice of medicine is not perfect and false test results are encountered frequently.
Diagnostic imaging is not immune from the problem of imperfect results!
False negative results may occur when a lesion is not visualized because it is small or in an inaccessible anatomic location.
False positive results may occur if a normal structure is misinterpreted as abnormal or if a measurement of a normal organ is outside the reference range.
Avoiding Radiographic Errors
Make good quality radiographs
Include all relevant anatomy
Search the film carefully
Make good quality radiographs
Know normal anatomical variations
Once good quality radiographs have been made, they must be searched carefully for signs of disease. How to look at a radiograph is not as obvious as you might think. They are major differences of opinion between radiologists about how to search films for abnormalities. Two methods are described:
Directed search pattern, i.e., look at the various structures on the radiographs according to a preconceived sequence in an attempt to avoid concentrating on a central or obvious abnormality at the expense of peripheral or unexpected lesions. Periphery first is a common search pattern.
Hypothesis-driven search, i.e., form a hypothesis about possible diagnosis from the history or from the initial observation of a suspected abnormality, then use this to guide further examination of the radiograph.
It is common practice to teach undergraduates and radiology residents to use a directed search pattern for examining radiographs despite evidence that higher performing students and expert diagnosticians search radiographs according to hypotheses they generate about the patient rather than using a directed search pattern.
This approach reflects the fact that clinical radiographs must always be considered in the context of a specific patient. There are assessments that present candidates with radiographs without any accompanying patient information or history, but this is an artificial situation. In practice, I suggest that we should always know why we have made radiographs, hence the history is very important. Without the history it is not possible to answer these key questions:
What is the aim of this study?
Is the study adequate?
What is the prior probability (prevalence) of disease?
Does the history suggest a diagnosis?
Helps interpret a negative result
Even though the history may be incomplete, there is usually some information available that prompts examination of specific parts of the film. With increasing experience, we become more familiar with the usual location and appearance of all the common veterinary conditions, which means we are more likely to check specific locations on the film, chosen because of our suspicions about that patient, rather than use a traditional "periphery first" directed search pattern. Even if there is no history, our initial observation of the film may reveal a sign that we analyse first and then use to prompt examination of the other parts of the film. There is abundant evidence that experienced radiologists begin to form hypotheses almost immediately they start the information-gathering process, and these hypotheses influence the direction of further searches for information.
To take a simple example: the first observation made by an experienced radiologist when examining thoracic radiographs of a cat with a history of regurgitation might be a thick soft tissue band on the dorsal aspect of the tracheal lumen, i.e., a tracheal stripe sign, which occurs because of air in the oesophagus. This finding will rapidly prompt a search of specific areas of the film for other signs of oesophageal dilatation, oesophageal foreign material, hiatus hernia, and then signs of aspiration pneumonia, a common sequel to regurgitation. The expert may go further, looking for evidence of a mediastinal mass that might be present if thymoma and concurrent myasthenia gravis are the conditions underlying a megaoesophagus. All this is done in the time it takes a novice following a directed search pattern to examine the ribs.
A recent study at The RVC compared the performance in a film-reading quiz of students taught a directed search pattern and students encouraged to use a hypothesis-driven search. There was a trend suggesting a better performance by students taught a hypothesis-driven search, and students taught a hypothesis-driven search were less likely to record a false positive observation than students taught a directed search.
These results are compatible with the results of an earlier study of chiropractic students, which concluded that those who are most skilled at film reading examination:
Are better able to identify key radiographic signs
Are better able to recognise the relationship between multiple abnormalities
Think of possible diagnoses early in their examination of the films
Make a complete examination of the film.
It is known that medical students taught a highly structured, step-wise approach to examining radiographs performed no better than students who did not receive such coaching. So why does the directed search pattern persist? The rationale that underpins the directed search is based largely on the assumption that it maximises the completeness of scrutiny of the radiograph while minimising the potential to miss an unsuspected or subtle abnormality in a patient with an obvious abnormality - the error known as satisfaction of search.
Satisfaction of search occurs because obvious abnormalities capture visual attention and decrease vigilance for more subtle abnormalities, which may be fixated but are not recognised, either because of decreased search time or other mechanisms. It is not clear that use of a directed search pattern reduces satisfaction of search. In a study in which radiologists were interviewed during film interpretation, satisfaction of search was reduced, possibly because the act of describing their focus of attention may have prompted observers to inspect the radiographs in a more deliberate, systematic way; however, there is also evidence that it is difficult to perform a directed search: studies of eye movements in radiologists trained to use a directed search found that most used a free search, including those who believed they were following a directed search. Many radiologists scan the entire radiograph with short excursions to examine in more detail regions that they suspect might be abnormal. This is precisely the approach that the classic description of a directed search advises against.
It is my belief that a hypothesis-based search pattern is more efficient, more accurate and easier to learn that a directed search. Students of radiology should consider adopting it and teachers (and examiners) should be tolerant of this approach.
There will always be radiographic abnormalities that defy diagnosis because they are unexpected and/or subtle. All we can do is try to remain vigilant.