A first-generation iPad slightly underperformed medical-grade liquid crystal displays (LCDs) for spotting cerebral infarction on brain CT studies, but the effect is reduced when more experienced readers are using the iPad, according to research published in the June issue of the American Journal of Roentgenology.
Japanese researchers compared the original iPad with medical-grade LCDs calibrated to two different methods: One was calibrated to the DICOM grayscale standard display function and the other to the gamma (γ) 2.2 approach. No significant difference was found between the iPad and the LCD calibrated with the DICOM grayscale standard display function, but there was a small but statistically significant difference between the iPad and the LCD calibrated to γ 2.2.
In addition, a notable finding was that readers in the study with more than 10 years of experience had a smaller difference in performance among the display types than less-experienced observers (AJR, June 2013, Vol. 200:6, pp.1304-1309).
"The iPad has the potential to be used as a diagnostic imaging monitor when clinicians diagnose brain infarction, but further evaluation is needed," lead author Kumiko Yoshimura, of Nagoya University Graduate School of Medicine, told AuntMinnie.com.
Display face-off
Following up on previous work that found a notebook PC could be used to view brain parenchyma on CT images, the researchers sought to evaluate whether the iPad could be employed to diagnose cerebral infarction, according to Yoshimura. Because that prior study also determined that the notebook LCD calibrated using a γ 2.2 setting had performed better than a medical-grade LCD calibrated with the DICOM grayscale standard display function, the team also wanted to directly compare the two calibration approaches on the same display type.
In addition to the first-generation iPad (1,024 x 768 resolution), the researchers used two 21-inch RX211 LCD medical-grade monitors (Eizo Nanao Technologies) with 1,200 x 1,600-pixel resolution. The one monitor was calibrated to the DICOM grayscale standard display function using version 3.01 of the Dr. Kal deluxe calibration software (Data Ray), while the other was fitted to the γ 2.2 curve using RadiCS calibration software and a UX1 external-motion sensor (both from Eizo Nanao).
Two experienced radiologists selected a total of 97 patients with brain CT images from Nagoya University Hospital's reporting system and electronic medical charts. The population included 47 patients with proven cerebral infarction, of which 14 were acute lesions and 33 were chronic lesions. The remaining 50 cases were normal controls.
Six general radiologists and three neuroradiologists with a mean experience level of 7.3 years (range, 3-17 years) then read all images in three separate sessions, one on each display type. Reading sessions were separated by a month. Although blinded to clinical information, the observers were informed that half of the images were of healthy control cases and the other half had brain infarction, some of which were acute stroke.
In addition to diagnosing each case, the radiologists were asked to mark their confidence level on a scale of 0 (definite absence of cerebral infarction) to 100 (definite presence of cerebral infarction). Reader observer performance was assessed using receiver operator characteristics (ROC) analysis.
Readers used STWS-005 DV image viewing software (Toshiba Medical Systems) on both LCDs and version 1.1.4 of the OsiriX platform for iPhone software on the iPad.
Small but significant difference
The LCD calibrated to the γ 2.2 standard had an area under the ROC curve (AUC) value of 0.884, while the one calibrated to the grayscale standard display function had an AUC of 0.875. The iPad had an AUC of 0.839.
The difference in AUC values between the iPad and the γ 2.2 LCD was statistically significant (p < 0.05), but the difference was not significant for the grayscale standard display function LCD. The difference between the two LCDs also was not significant.
"Our results show a small difference between medical-grade LCD monitors and iPad, and this is acceptable compared with past reports," the authors wrote.
Reader experience level also appeared to have an effect. There were no statistical differences by display type for observers with 10 or more years of interpretation experience and those with fewer than 10 years; however, the differences between the three display types for detecting infarction were smaller in observers with more than 10 years of experience, the authors noted.
"The present result showed that the iPad was acceptable for reading brain CT, especially when used by well-trained observers," they wrote.
The iPad took the longest for image interpretation, requiring a mean reading time of 83.1 minutes. In contrast, the γ 2.2 LCD and grayscale display function LCD had a mean reading time of 70.9 minutes and 75.2 minutes, respectively.
In other findings, the best display luminance response was turned in by the γ 2.2 LCD, followed by the iPad and then the grayscale standard display function LCD, the researchers noted. That contrasted with the observer performance results in which the grayscale standard display function LCD had finished second, ahead of the iPad.
"Image interpretation is not necessarily correlated only with luminance measurement and is considered to be more complex because of various factors, including both the characteristics of monitors and the different experience of image interpreters," the authors wrote.
Is newer better?
Of course, the iPad has experienced considerable technology advances since its first iteration, which the authors acknowledged as a limitation of their study. The current new iPad offers 2,048 x 1,536-pixel resolution, while the first-generation iPad had a pixel matrix of 1,024 x 768.
Indeed, the researchers plan to study the new iPad for diagnosing acute cerebral infarction, Yoshimura said.
