This Article Abstract Full Text Full Text (PDF) Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Manning, D J Articles by Donovan, T Search for Related Content PubMed PubMed Citation Articles by Manning, D J Articles by Donovan, T

Detection or decision errors? Missed lung cancer from the posteroanterior chest radiograph

Department of Radiography and Imaging Sciences, St Martin's College, Lancaster LA1 3JD, UK

View larger version (65K): [in a new window]   Figure 1. The lesion distribution in the test bank of images. The nodule locations marked • were those used in the eye-tracking procedure. All locations shown were used for the alternate free response operating characteristic scoring.

View larger version (119K): [in a new window]   Figure 2. The boundary of each lesion was outlined by eye from the image appearance and this was then confirmed by using the % grey level reduction from its peak value as shown in Figures 3 and 4. The four profiles were taken as a minimum number of samples required to represent heterogeneity of nodule internal structure as well as its background.

View larger version (28K): [in a new window]   Figure 3. The mean values of grey level for nodule and background and the greatest value of from four profiles were used in the calculation of the conspicuity index of each of the 81 nodules.

View larger version (17K): [in a new window]   Figure 4. The four profiles for each nodule frequently displayed a wide variation in the edge sharpness.

View larger version (33K): [in a new window]   Figure 5. The correlation between the conspicuity index of the nodules and the probability of them being called diagnostically significant in the alternate free response operating characteristic (AFROC) scoring was weak (R2=0.0054).

View larger version (29K): [in a new window]   Figure 6. The seven radiologists detected 59 (72.8%) of the 81 lesions. There was no significant difference in the physical characteristics of the missed and detected lesions measured by their edge sharpness as the mean value of tan (-1), signal to noise ratio (SNR), length in cross section of derived conspicuity index (p=0.78).

View larger version (29K): [in a new window]   Figure 7. The eye movements of four of the radiologists were tracked. Their mean dwell-time of gaze duration for detected nodules was twice that of those missed. However, the missed lesions attracted average gaze duration of 3.1 s.

View larger version (28K): [in a new window]   Figure 8. The percentage of nodules holding visual attention over a 15 s time interval for the four possible decision outcomes of true positive (TP), true negative (TN), false positive (FP), and false negative (FN). Note that nearly 70% of the missed lesions (FN) held a gaze duration of greater than 1 s. Deciding correctly that a chest zone contained no nodules was rapid (TN curve) but all positive decisions followed extended periods of attention (TP, FP).