The gold standard in diagnostic imaging has advanced from reviewing images on film to computer-based display workstations.Reference Dreyer, Hirschorn, Thrall and Mehta 1 Images are available immediately, they can be reviewed from different workstations, and a computer’s functionality provides reviewers with the tools to magnify, rotate and calculate measurements on images.Reference Huang 2 Guidelines have been created for digital workstations, with respect to such things as screen resolution, pixel counts, luminance ratings and ambient light.Reference Krupinski, Williams, Andriole, Strauss, Applegate and Wyatt 3
A key disadvantage of digital workstations, commonly situated in rooms with low ambient lighting, thus creating ideal conditions to interpret images,Reference Siegel and Reiner 4 is lack of portability. Accessibility is limited, obtaining a second opinion becomes time-consuming, and teaching of clinical trainees and even patients and families is less available in real time. A solution to this problem is to use a handheld device.Reference Johnson, Zimmerman, Heath, Eng, Horton and Scott 5 - Reference Mitchell, Sharma, Modi, Simpson, Thomas and Hill 7 We sought to determine if the quality of image review on a handheld device, such as an iPhone or iPod touch, is sufficient to make a clinical diagnosis by comparing it to image review on a conventional picture archiving and communication systems (PACS) digital workstation.
Three neuroradiologists rated images on the following criteria: image contrast resolution, sharpness of the display, display of artefacts and whether the image is of acceptable quality for clinical use. CT and MR studies were reviewed on a handheld device running ResolutionMD Mobile in both high and low ambient light. The same images were then reviewed in typical clinical review conditions on a digital workstation in low ambient light.
Images were selected to represent common neurological conditions that would be encountered in a tertiary care setting. Selection of cases was made by an independent neuroradiologist (MG). Six CT or CT angiography studies and six MR or MR angiography studies were chosen; CT/CTA studies were assessed as a group followed by the six MR/MRA scans.
Three neuroradiologists (all U.S. board certified and Fellows of the Royal College of Physicians and Surgeons of Canada) independently reviewed and dictated a report on each study. Each was provided with a standard handheld device. They each sequentially and independently reviewed all scans on the handheld device, first in a high ambient light environment (usual hospital hallway conditions), then in an idealized low ambient light environment, and finally on a standard diagnostic imaging workstation in a low ambient light environment (ideal reading conditions). Cases were reviewed during one three-hour period, in random order and under each lighting condition.
Each neuroradiologist dictated a report for each case. They were asked to comment on the imaging diagnosis and particular features of the case, as they would have done normally. On the second and third reviews of each case, they were additionally asked if the change in reviewing conditions allowed them to observe any additional diagnostic features. Finally, they were asked to comment qualitatively on: (a) image contrast for differentiation of subtle tissue density; (b) sharpness of edges and tissue borders; (c) artefacts due to image display; (d) overall clinical image display quality; and (e) whether they considered the image display to be of sufficiently acceptable quality for routine use and decision-making.
The portable handheld system consisted of an iPod touch (Apple Inc., Cupertino, CA), client image viewing software (ResolutionMD Mobile, Calgary Scientific Inc., Calgary, Canada) and a visualization server. The iPod touch had 8 GB of flash memory and a 3.5-inch diagonal screen with 320×480 pixels (pixel pitch=0.15 mm, luminance=500 cd/m2). It ran the ResolutionMD Mobile client program, which handled user input, communication with the visualization server and display of transmitted images. The client software allows for 2D and 3D visualization with interactive window/level, translation, rotation and zoom functions. In 3D mode, the user can also select from a range of tissue-rendering modes. The visualization server had a 2.4-GHz Intel Core 2 Quad central processing unit, 8GB RAM and 2 NVidia GeForce 8800 (512MB) graphics cards. It ran the application ResolutionMD Enterprise (Calgary Scientific Inc., Calgary, Canada), which allows remote 2D/3D visualization of digital imaging and communications in medicine (DICOM) images through a web browser or a mobile device (iOS and Android platforms). The visualization server first reformats a series of 2D DICOM images into a 3D volume. It then performs a rendering operation on the 3D volume to produce a 2D image for interpretation. This 2D image is converted into streaming joint photographic experts group (JPEG) format and then transmitted to the client program running on the mobile handheld device. User interaction on the device, such as a touch or drag event, causes a new rendering operation on the server and transmission of a new image. Communication between the visualization server and the handheld device occurs over a secure wireless network (Wi-Fi 802.11g). This system is capable of delivering and displaying up to 14 images per second on a mobile handheld device. In our experience, a single visualization server can accommodate 10 or more simultaneous handheld device users. Importantly, all patient data remain on the visualization server with this system. The stream of rendered images is not saved on the device, ensuring data security in compliance with health data privacy legislation.
Gold-standard diagnosis occurred on a radiology workstation (IMPAX 22.214.171.12415, Agfa Healthcare, Belgium) connected to a medical-grade 21-inch liquid crystal display (MD21GS-3MP, NEC). This display has a resolution of 2048×1536 pixels (pixel pitch=0.21 mm) and a luminance of 400 cd/m2.
Data are reported using standard descriptive statistics. Qualitative reporting is provided descriptively.
The 12 cases were reviewed in two groups, separated based on their modality (Table 1) and in various lighting conditions (Table 2), resulting in 108 distinct interpretations. Additional findings were identified in low ambient light handheld review in 1 instance (5.6%) for the CT/CTA case group and in 1 instance (5.6%) for the MR/MRA case group. Similarly, additional findings were identified in a low ambient light PACS workstation review in 1 instance (5.6%) for the CT/CTA case group and in 3 instances (16.7%) for the MR/MRA case group. Confidence in the findings was reported more often in the low ambient light conditions, both for the handheld and the PACS workstation review (Tables 3 and 4).
* p<0.001 (ANOVA) for all comparisons to iPhone Hallway adjusted for multiplicity (Scheffé test). Light intensity was not different between iPhone dark and PACS dark (p=0.982).
A=no changes/additional findings; B=seeing findings clearer/more confidently; C=an additional finding.
A=no changes/additional findings; B=seeing findings clearer/more confidently; C=an additional finding.
Combining all cases where an additional finding was seen on second or third review, only one out of six contributed to the initial diagnosis, but none were diagnostically relevant. The increased clarity and confidence of findings was judged to be helpful but clinically insignificant. Qualitatively, all reviewers felt comfortable making a preliminary diagnosis for CT/CTA and MR/MRA studies on a handheld device, and all initial diagnoses made on the handheld device in high ambient light were clinically correct (Table 5).
We found that interpretation of neuroimaging studies using a portable handheld system is reliable and accurate for principal imaging diagnosis. We have also shown that low ambient lighting is qualitatively relevant for the completeness of and confidence in imaging interpretation. Although some findings were seen with more confidence in low ambient light conditions, they did not contribute further to the clinical diagnoses made in high ambient lighting. Each neuroradiologist felt comfortable using a handheld device to make preliminary principal diagnoses in lighting conditions typical of a hospital hallway. However, detailed review on a gold-standard workstation should be completed to ensure that ancillary imaging findings may be identified. This information expands on previous studies of imaging interpretation using mobile devices, which placed an emphasis on user experience with a mobile device rather than on diagnostic capabilities.Reference Raman, Raman, Raman and Beaulieu 8 , Reference Choudhri and Radvany 9
Compared to a conventional workstation, the mobility of a handheld device poses a potential security risk related to sensitive patient data. To address this, the ResolutionMD Mobile system has been designed with important data security features.Reference Mitchell, Sharma, Modi, Simpson, Thomas and Hill 7 First, confidential patient information is not stored on the device, where it could potentially be carried outside the hospital network firewall—a feature that is not integrated into all mobile DICOM viewers.Reference Choudhri, Carr, Ho, Stone, Gay and Lambert 10 , Reference Choudhri, Norton, Carr, Stone, Hagspiel and Dake 11 Instead, this information resides remotely on a secure visualization server. An added benefit of this design is that the remote visualization server can rapidly load, and render, large medical image datasets containing several hundred DICOM images. This allows diagnostic interpretation to begin almost immediately, even from remote locations. Second, when the ResolutionMD Mobile software is closed, the network connection to the server is automatically terminated. 12
Although reviewers felt comfortable assessing CT/CTA and MR/MRA studies on the handheld iPod touch, there are advantages of using a larger screen: studies can be viewed side by side; the user can link various views and scroll through image series simultaneously; important patient demographic information is available to aid in accurate diagnoses; and the larger screen has a greater conspicuity compared to a smaller one. Use of tablet or web platform versions of viewing software or newer-generation handheld devices with larger screens might resolve these limitations. 12
Acknowledgements and Funding
We would like to thank Calgary Scientific Incorporated for providing the device support (both software and hardware) that allowed this study to be completed.
This was an unfunded study. MH, MG and WH own stock in Calgary Scientific Inc., the company that markets the software used for image display.
Statement of Authorship
PR wrote the first draft of the manuscript. MG chose cases for review. WM, JL and WH reviewed cases. MH designed the study. All authors participated in data analysis and critical review of the manuscript.