Today’s advanced imaging modalities are acquiring more and more images that must be interpreted in less and less time [1
]. Concerns have been raised that radiologist workloads are becoming so demanding that fatigue and reduced time for interpretation are negatively impacting diagnostic accuracy [9
]. In court, a plaintiff’s attorney has argued that a radiologist missed a breast lesion because he was overworked [9
Although radiologist fatigue has been a concern for years, only recently have dedicated studies been conducted. Some early studies did not examine fatigue or viewing times directly. For example, Oestmann et al. [12
] demonstrated that detection accuracy for lung nodules decreased as viewing time decreased, but fatigue was not examined. Bechtold [11
] found that error rates in the interpretation of abdominal CT more than doubled when radiologists read out more than 20 studies in a day. This retrospective review and classification of errors in clinical cases was not a controlled examination of fatigue.
More recently, studies have examined reader accuracy at different times during the day, with mixed results. Taylor-Phillips et al. [15
] examined data from the UK Breast Screening Programme for nearly 200,000 cases in an attempt to relate accuracy to time of day and reading time. They found that recall rates varied with time of day but not in the same way for the individual readers. Some readers had lower recall rates in the afternoon, while others did not. Recall rates tended to decline with increased reading time (i.e., recall rates were lower around lunch and the end of the day), but again it varied considerably among readers. The sample was too noisy to document anything significant beyond a possible trend. This study did not directly examine fatigue or conduct a controlled study in which readers read a dedicated set of cases before and after a day of clinical reading.
Al-s’adi et al. [16
] also found that breast lesion detection varies with time of day, but that no particular time of day had a significant effect. Radiologists at a national meeting were recruited to read a set of mammograms during one of 4 reading times (7:00 – 10:00; 10:00 – 13:00; 13:00 – 16:00; 16:00 – 20:00). There were no statistically significant differences in sensitivity, specificity or area under the receiver operating characteristic (ROC) curve (AUC) as a function of time of day. Limitation of this study include readers only participating in a single session and that they could choose the time of their participation, possibly choosing a time of higher performance or motivation.
The impact of fatigue was directly studied by Krupinski et al. [17
] using skeletal radiographs with fractures as the detection task. Forty radiologists and residents interpreted a set of 60 patient examinations before and after a day in the reading room. Resting state of accommodation (a.k.a. dark focus) was measured as an indicator of visual workload on oculomotor equilibrium. Subjective measures of physical and visual strain and/or fatigue were collected. The results indicated that diagnostic accuracy was reduced significantly from before to after the day of clinical reading (p < 0.05) and the radiologists and residents became more myopic. Subjective ratings indicating increased lack of energy, physical discomfort, sleepiness, physical exertion, lack of motivation, and eyestrain. In general, the residents exhibited greater effects of fatigue on all measures compared to the attending radiologists. The conclusion was that after a day of clinical reading, radiologists have reduced ability to focus and a reduced ability to detect fractures, and increased symptoms of fatigue and visual strain.
The results of this study probably generalize well to most radiographic modalities. However, there are usually few radiographs per patient and the images are static. Tomographic modalities such as CT, magnetic resonance imaging (MRI) and digital breast tomosynthesis are viewed in fundamentally different way than radiographs.
The sequences of tomographic sections are typically viewed in ciné-animation mode with successive sections presented one after another under the radiologist’s control. The difference between static and dynamic displays places different demands on the visual system. A very basic, yet critical, distinction in the human visual system is between channels processing static stimuli and channels processing moving or changing stimuli [18
]. Briefly, the transient visual channel, with high temporal resolution but poor spatial resolution, serves as an “early warning system” for the sustained visual channel which has poor temporal resolution and high spatial resolution. Things that move or change attract attention and eye movements. That is why people wave when they want to attract attention and why warning signals flash off and on. It is why things that move seem blurry and things that do not move seem to be sharp. These characteristic reflect the sensitivities of the two parts of the visual system handling perception of these stimuli. As the radiologist cycles dynamically through a sequence of CT sections, the sudden onset and offset of a pulmonary nodule captures the viewer’s attention and directs it to the location of change [20
]. With dynamic images, the motion channel of visual processing which directly affects attention comes into play, and the task of guiding the eyes around the changing image in search of lesion becomes more complex. Thus, the impact of fatigue may differ for dynamic and static image interpretation.
The goal of the present study was to measure diagnostic accuracy for pulmonary nodule detection in dynamic CT chest sequences before and after a day (or night) of diagnostic image interpretation. We also investigated measure of visual strain, the resting static of convergence, often referred to as dark vergence.