PACS displays influence brain ischemia detection

The specifications of LCD monitors -- specifically spatial resolution and luminance -- play an important role in the ability to detect brain ischemia on CT studies, particularly for less experienced radiologists, Italian researchers reported at last week's RSNA 2016 meeting in Chicago.

As part of a multicenter study evaluating the influence of spatial resolution and luminance characteristics on a number of different modalities and clinical scenarios, the investigators found that a 3-megapixel monitor with calibrated maximum luminance of 400 cd/m2 yielded the best reproducibility and diagnostic performance for identifying brain ischemia. Junior radiologists were the most susceptible to a drop-off from using monitors with lower spatial resolution and maximum luminance levels.

"The experience of the radiologist can only partially compensate for the disadvantage represented by a poorly performing monitor that, in the case of less-experienced radiologists, can heavily affect the reporting," said presenter Dr. Calogero Cicero of General Hospital "San Bassiano" in Bassano del Grappa, Vicenza province.

Measuring luminance

The 97/43 EURATOM directive of 30 June 1997 defined the calibration, safety, quality assurance, and quality control requirements of equipment used on patients. In addition, the 2012 Practice Guideline for Digital Radiography published by the American College of Radiology (ACR), American Association of Physicists in Medicine (AAPM), and the Society for Imaging Informatics in Medicine (SIIM) defined the equipment guidelines, specifications of data manipulation and management, and quality control for the use of digital image data that should result in high-quality radiological care.

These three organizations recommended that the resolution (pixel pitch) of monitors used in diagnostic interpretation should be about 0.2 mm and not less than 0.21 mm. In addition, the maximum luminance of diagnostic monitors should be at least 350 cd/m2, Cicero said.

As part of their multicenter study, the researchers sought to evaluate the impact of spatial resolution and luminance in identifying cerebral infarction on CT. A senior radiologist selected 32 studies, 16 without cerebral infarction and 16 with cerebral infarction. The cerebral infarction cases included six that were very difficult to interpret, five studies that were deemed to have an intermediate difficulty level to interpret, and five that were considered to be simple to interpret, he explained.

The team utilized four LCD monochrome displays provided by NEC Display Solutions:

• Monitor A: MD211C3 -- A 3-megapixel monitor with a pixel pitch of 0.212 mm, a calibrated maximum luminance of 400 cd/m2, and a calibrated minimum luminance of 0.5 cd/m2
• Monitor B: MD211C2 -- A 2-megapixel monitor with a pixel pitch of 0.27 mm, a calibrated maximum luminance of 200 cd/m2, and a calibrated minimum luminance of 0.5 cd/m2
• Monitor C: MD211C3 -- A 3-megapixel monitor with a pixel pitch of 0.212 mm, a calibrated maximum luminance of 200 cd/m2, and a calibrated minimum luminance of 0.5 cd/m2
• Monitor D: MD211C2 -- A 2-megapixel monitor with a pixel pitch of 0.27 mm, a calibrated maximum luminance of 400 cd/m2, and a calibrated minimum luminance of 0.5 cd/m2

After the CT studies were anonymized and coded according to a randomization scheme, three radiologists blindly evaluated the exams on the selected displays with a period of two weeks in between reading sessions. The readers included a junior radiologist with four years of experience, a junior radiologist with six years of experience, and a senior radiologist with 14 years of experience. Each of the radiologists completed an evaluation form for every study, rating the presence or absence of ischemia on a five-point percentage confidence level: ≤ 20%, 20% to 40%, 40% to 60%, 60% to 80%, and > 80%. They also indicated the area of the brain with the suspected lesion.

All reading sessions were performed under standardized conditions with ambient room light of less than 20 lux to minimize reflections, Cicero said.

The researchers computed the Cohen's kappa statistic to assess reproducibility and also quantified observer diagnostic performance using a trapezoidal area under the empirical observer-weighted jackknife alternative free-response receiver operating characteristic (wJAFROC) curve.

They found that monitor A produced the highest level of agreement for all radiologists.

Reproducibility vs. reference evaluation
	Junior radiologist 1 (second read)	Junior radiologist 2 (second read)	Experienced radiologist (second read)
Monitor A (3MP, 400 cd/m²)	0.738	0.764	0.955
Monitor B (2MP, 200 cd/m²)	0.420	0.575	0.814
Monitor C (3MP, 200 cd/m²)	0.653	0.527	0.909
Monitor D (2MP, 400 cd/m²)	0.574	0.655	0.727

The junior radiologists had substantial agreement on monitor A, while the senior radiologist had almost perfect agreement for all of the monitors, he continued.

Diagnostic performance

Monitor A also yielded significantly higher diagnostic performance than monitor B (p = 0.017).

	Monitor A (3MP, 400 cd/m²)	Monitor B (2MP, 200 cd/m²)	Monitor C (3MP, 200 cd/m²)	Monitor D (2MP, 400 cd/m²)
Junior radiologist 1	0.795	0.605	0.750	0.711
Junior radiologist 2	0.844	0.730	0.717	0.781
Experienced radiologist	0.969	0.875	0.938	0.813
Average performance	0.869	0.737	0.801	0.768

The differences between A and the other two monitors did not reach statistical significance. Overall, monitor A had sensitivity of 81.3% and specificity of 93.8%, while monitor B had sensitivity of 58.3% and specificity of 83.5%.

The findings show LCD spatial resolution and luminance are important factors for detecting brain ischemia, particularly for less-experienced radiologists, according to the group.

"This would allow us not only to obtain a more optimized procedure for the clinical reporting, but also to rationalize the purchase of new radiology workstations," Cicero said.