General-Purpose Devices.

A common way for nurses to take wound photographs is via dedicated digital cameras [8]. After taking photos, a nurse transfers images to a secure, password-protected computer, and deletes the images from the camera. Since digital cameras are not networked, a nurse would then transfer the images via a physical memory card. At Stanford Hospital & Clinics, nurses insert a memory card in a printer to make physical prints. He or she then places each print on a separate piece of paper and fills in relevant information with a pen. Each day, couriers transport a stack of these documents offsite for scanning. It requires up to two days for the images to be uploaded to a patent’s EMR.

Epic Systems (Verona, WI) is a leading provider of EMR software [9]. The Epic Haiku mobile application allows nurses to take photographs using a general-purpose smartphone and immediately upload images to a patient's EMR, where other clinicians may view the images [10]. Photos taken within the Epic Haiku app are not stored on the phone's memory, thereby ensuring compliance with security and privacy protocols [10]. To use Epic Haiku, a nurse typically holds the smartphone in both hands while another nurse holds the paper ruler and positions the patient. Alternatively, one nurse can hold the paper ruler in one hand and the smartphone in the other, in which case pressing the button to take a photograph is often challenging. After taking each image, the nurse must manually enter a text description, which requires one’s full attention and creates a barrier between the nurse and patient. After the patient encounter, the nurse meticulously wipes down all smartphone surfaces to disinfect the device.

MOWA (Mobile Wound Analyzer) is an application for Android smartphones and tablets that takes photos and uses software-based technologies to calculate wound area and identify three types of tissues in the bed of the lesion [11]. Presently, however, MOWA does not integrate with EMRs, thus limiting its utility.

Wearable and Special-Purpose Devices.

In addition to hand-held digital cameras, a range of studies has demonstrated the use of head-mounted cameras (e.g. GoPro Hero3) for clinical use and education [12]. A major advantage of wearable head-mounted cameras, “is that the user is not restricted in his or her manual and oculomotor actions by camera operation tasks” [13]. This allows the performance of actions in a natural way with increased mobility. Although often WiFi-enabled for remote viewing, wearable cameras typically lack patient identification and EMR integration capabilities for routine clinical care.

Optical head-mounted displays, such as Google Glass (Google, Mountain View, CA), can reflect projected images while allowing the user to see through the device’s eyepiece. Although not widely distributed for consumer use (as of November 2014), a number of proof-of-concept projects and studies using Glass have occurred in the medical field. Medical applications for Glass include remote collaboration between doctors, heads-up viewing of vital signs, medications, and lab reports, as well as live streaming of operations to medical students [14]. Several reports document the live two-way broadcasts of actual patient surgeries by doctors wearing Glass in the operating room [15, 16].

Wearable augmented reality (AR) systems have also been demonstrated in a range of clinical application areas, such as image-guided surgery, pre-operative imaging training, health-related behavior change, and clinical education [17, 18, 19]. Although studies have examined the use of AR systems for both in-patient and remote medical procedures [15, 16, 18], there is a growing need for the development of integrated systems that enable users to seamlessly capture and annotate visual images in a hands-free manner, and integrate this information with a patient’s EMR.

Special-purpose devices usually include sophisticated wound-measurement methods, in addition to photographic capabilities. These devices achieve increased accuracy through vision-based technologies such as stereophotography or structured lighting. For example, MAVIS III (Perry Baromedical, Riviera Beach, FL) is a stereoscopic, hand-held camera that measures the area, volume, and circumference of chronic wounds [20]. Traditionally, special-purpose devices have suffered from one or more of the following drawbacks: they may be expensive, cumbersome to use in a clinical setting, or require significant training time [21].

SnapCap System.

The SnapCap System includes both a Google Glass application (known as “Glassware”) and an Android smartphone application initially designed to be used by nurses for chronic wound photography [22]. Google Glass [23] was selected as an initial platform for research, since it has an optical head-mounted display that is capable of taking pictures and recording videos using an integrated camera. The device also has the ability to communicate wirelessly via WiFi and Bluetooth. Additional sensors integrated into the device enable the development of augmented reality-based applications. We developed the SnapCap Glassware using the Glass Development Kit (GDK) for Android 4.0.4 (API Level 15, Ice Cream Sandwich), and the smartphone app for Android 4.3 (API Level 18, Jelly Bean).

Working with Google’s new GDK placed constraints on system development. Since interaction via voice commands was limited, we implemented alternative methods to support the hands-free paradigm through gesture-based control. Similarly, working with the camera proved challenging. To preview images before taking them, it was necessary for the system to maintain a continuous video feed to the eyepiece while also allowing for digital zoom, neither of which are standard Glass features. We overcame problems with overheating hardware and an unstable image. Lastly, we had to develop a protocol to transmit image files via Bluetooth, since this was not natively available in the GDK.

In the current implementation, an Android smartphone serves as a hypothetical EMR that includes a medical database for six fictional patients. The smartphone application communicates with Glass via Bluetooth. Presently, the Bluetooth link is unidirectional, from Glass to the smartphone application, because the app needs to store the images taken with the SnapCap Glassware as well as the associated tags. Speech-to-text conversion takes place on Google servers (Fig 2).

SnapCap Smartphone Application.

The SnapCap smartphone application implements relevant features of the EMR, including a patient list, access to a patient database, and image storage in a media file. The password-protected application allows a clinician to select the name of the relevant patient prior to each patient encounter. The clinician can choose to either (i) take a clinical image, whereby the application establishes Bluetooth communication with Google Glass, or (ii) view the patient’s media file. Fig 3 illustrates the application flow.

After choosing to take a clinical image, the nurse is free to place the Android smartphone in his or her pocket and put on Google Glass for the duration of the patient encounter. The photographs taken using Glass, along with the associated tags and voice annotations, immediately transfer via Bluetooth to the smartphone and are stored in the patient’s media file, where they are available for viewing at any time.

SnapCap Glassware.

The SnapCap Glassware guides a clinician through the steps of sending photographs to the companion smartphone app. The application is launched via the voice command, “ok glass, take clinical image.” A live camera preview appears in the Glass display, and the application prompts the clinician to look at a barcode encoded with the patient’s MRN. If this MRN matches the MRN of the patient selected by the clinician, the Glass display briefly shows the patient’s name and returns to the live camera preview; otherwise, it displays a warning message. All subsequent images taken are tagged with the patient’s data (e.g. name, MRN, and date of birth).

Once in live camera preview mode, the nurse can zoom in by tilting her head to the right or zoom out by tilting her head to the left. She can take a picture by double-blinking. The picture taken remains on the display for three seconds. Unless the nurse discards the photo (by tilting her head forward), the nurse has the option to add a voice annotation. Lastly, she can tilt her head back to send the image and corresponding data to the hypothetical EMR (the smartphone). In this case, the smartphone app compresses the picture, stores it in the media file, and displays a confirmation message. Otherwise, the nurse can wait three seconds for the live camera preview to return and then take another picture, effectively deleting the previous picture without sending it. The nurse can take as many consecutive pictures as necessary to document the wound.

In summary, the novel features of the SnapCap Glassware include (i) barcode scanning using the Glass camera, and tagging subsequent images with a patient’s personal identification information that is embedded in the barcode, (ii) capturing a live video preview in the Glass eyepiece before a photo is taken, (iii) using a double-blinking gesture to take photographs, utilizing Glass’ IR sensor, and (iv) using a head-tilt gesture (while in the preview mode) to zoom in and out of an image, and a head-tilt gesture to send images to a patient’s EMR, through the use of Glass’ internal measurement unit (IMU) sensor.

Google Glass Sensors.

Google Glass has four sensors that we have used to implement the SnapCap System for hands-free wound photography: a camera, an inertial measurement unit (IMU), an infrared sensor facing toward the user’s right eye, and a microphone.

Camera with Digital Zoom. Albrecht et al. [24] concluded that Glass was efficient for acquiring images, but for photo documentation in forensic medicine, the image quality was inferior to those from a Digital Single-Lens Reflex (DSLR) camera. This finding is perhaps not surprising, considering that Google Glass uses a standard 5-megapixel (2528x1956) smartphone camera with a fixed focal length of 3mm. The Glass camera has a wide-angle lens that captures a field of view too large for most typical medical applications and illustrates the need for zoom [25].

Because Glass has a fixed focus, all zooming is digital and does not lead to a gain in optical resolution. In theory, a photo could be cropped in post-processing to obtain an identical image as one taken after digital zooming, but our preliminary interviews showed that nurses appreciated having the ability to zoom before taking the picture. In practice, we discovered one drawback with zooming before picture taking. Usually, Glass uses burst-mode photography (capturing a rapid sequence of shots to improve dynamic range or reduce noise) when instructed to take a picture [26]. Burst-mode is lost when taking a picture after a zoom, resulting in lower quality photos.

Barcode reader. Hospital patients wear ID wristbands printed with a medical record number and corresponding barcode. We implemented a barcode reader based on the open source, cross-platform library ZBar [27]. ZBar works very well with newer smartphones, like the iPhone 5, that have auto-focusing cameras. Since the Google Glass built-in camera has a fixed focus, we used trial and error to find a barcode size that ZBar could read at arm’s length, where the image was focused (we found that objects closer than approximately 30cm would not be in focus). Furthermore, we had to use a flat barcode, rather than the curved barcode on wristbands used at most hospitals.

Microphone. The Android 4.0.4 GDK version allows developers to designate a custom voice command to launch an application. SnapCap is launched with the command, “ok Glass, take clinical image.” We also used the microphone for wound annotation purposes, whereby our app sent audio to the Google servers for speech-to-text conversion. Previous investigators were surprised to find that Glass recognized complex medical terms such as “Microvillous Inclusion Disease” about half the time, but cautioned that for reliable medical use, however, the error rate had to decrease substantially” [25].

IMU and Infrared Sensor. Ishimaru et al. [28], who argued that the inertial measurement unit (IMU) and blink detection are the most characteristic features of the Google Glass platform, studied combining head motion with blink frequency for activity recognition. For the purposes of SnapCap, the IMU and infrared sensor each provided a separate input for hands-free interaction. We used the IMU (gyroscope and accelerometer) to measure head tilt and the infrared sensor to detect double blinks.

Part 1a: SnapCap Feature Evaluation and Application Comparison.

For part one of the evaluation, we brought each nurse into a hospital room where we had placed two pelvis-only mannequins and an identifying barcode on a table (Fig 4). The mannequins are normally used to train students in the assessment and treatment of pressure ulcer wounds. Two to three researchers were present at each user session. The research team provided each participant with a copy of the study protocol at the start of each session, and asked for each nurse’s consent to participate in the study. The study protocol and user questionnaire (S2 Fig and S3 Fig) are provided as Supporting Information. A researcher explained the purpose of the study, the time breakdown per user session, the Glass navigation basics, and the experimental set-up. A researcher explained the use of both SnapCap and Epic Haiku, and asked each nurse to photograph a wound on a mannequin using (i) SnapCap (running on Glass and a Galaxy Nexus smartphone) and (ii) Epic Haiku (running on an iPhone 4S). The photographs taken by nurses (using SnapCap) were saved in the patient’s hypothetical EMR to examine the quality of each photo taken. For wound annotation, each nurse was asked to use text-based documentation in Epic (per standard documentation practices) and to annotate the wound in a 10-second video recording using the SnapCap system. Fig 5 shows the step-by-step flow in each session. The research team took manual notes and recorded each session using a digital recording device. Photographs and brief videos of nurses using the SnapCap System and Epic Haiku application were also captured. Each user session, consisting of the SnapCap evaluation, application comparison, and a post-task questionnaire, lasted 30–45 minutes per participant, and each subject was given a $15 gift card for his or her study participation.

Part 1b: Post-task Questionnaire.

Following the use of each system, participants completed a pen-and-paper questionnaire (see S3 Fig) that was segmented into five areas. The questionnaire first asked users to provide their number of years of wound care experience and smart phone experience, and to list mobile applications that they commonly use. The questionnaire then asked participants to share their preferences for (i) current SnapCap system features, (ii) application preferences for SnapCap versus Epic Haiku, and (iii) preferences for the implementation of future SnapCap features. The research team saved the questionnaire data in an Excel spreadsheet for data analysis.

(i) Glass Feature Preferences. Specifically, in the first part of the post-task survey, participants were asked to indicate their preference for the following SnapCap features:

1. Barcode scanning for patient identification
2. Voice-based documentation via a brief video recording
3. Double-blinking gesture to take photographs
4. Head-tilt gesture for zooming in and out of an image

Response options were “prefer,” “do not prefer,” or “recommended improvements” (with space to provide improvement recommendations). For each question, a response was required and multiple response options could be selected (e.g., prefer and recommended improvements).

(ii) Application Preferences. In the second part of the post-task survey, participants were asked to indicate their preference between SnapCap and Epic Haiku for the following performance dimensions:

1. Considerations for sterile wound image-capture technique
2. Photo-capture capability
3. Image quality
4. Overall Ease of Use

Response options were “Epic Haiku,” “Google Glass,” “no difference,” and “neither.” For each question, a single response was required.

(iii) Preferences for Future SnapCap Features. In the third part of the post-task survey, participants were asked to indicate their preferences regarding the perceived benefit of future SnapCap features:

1. The use of a head-mounted display to share and discuss images with colleagues
2. Historical image retrieval for data recall and sharing
3. The use of a dynamic digital ruler inside the Glass eyepiece

Response options were “yes,” “no,” and “no preference.” For each question, a single response was required.

Part 2: Speech-to-text Wound Annotation.

Following the initial user session, we conducted a 10–15 minute follow-up session at Stanford Hospital with the original sixteen nurse participants, to obtain quantitative data on the performance of Google’s speech-to-text (STT) engine. In these sessions, we asked each nurse to read aloud a wound annotation, consisting of the following two fictitious descriptions, while wearing the Glass head-mounted display:

1. “This is a stage 3 pressure ulcer, measuring 11 centimeters by 6 centimeters by 3.4 centimeters deep, full-thickness ulceration which probes to bone, circumferential undermining, 2.5 centimeters at 12 o’clock and 0.2 centimeters at 6 o’clock.”
2. “Wound bed is approximately 70 percent necrotic tissue, 30 percent non-granulating pale pink wound bed with moderate amount of malodourous brownish red drainage. Wound edges are rolled, closed, non-proliferative. Periwound tissue is intact.”

Results of the transcription were stored locally in Glass’ built-in memory for analysis.

Data Analysis.

For the SnapCap feature evaluation and application comparison (Part 1 of the pilot study), we used the Wilcoxon Signed-ranks test, a non-parametric statistical hypothesis test, to evaluate differences in mean ranks between two response options. The analysis was conducted using IBM SPSS statistical software. For the SnapCap feature preference analysis, in which users could select prefer, do not prefer, and/or recommended improvements, we excluded responses in which users only selected “recommended improvements,” since this option did not indicate a specific preference. Instead, the analysis focused on the “prefer” and “do not prefer” options (which may or may not have included recommended improvements) and converted these preferences to ranks for the two possible options. If the “prefer” option was selected, we assigned the rank of 1 to this option, and the rank of 2 to the “do not prefer” option, and vice versa.

Similarly, for an evaluation of the perceived benefit of future SnapCap features, the same analysis method was applied, with the options of yes, no, or no preference. The “no preference” option was excluded, since it did not indicate a specific preference. If “yes” was selected, we assigned the rank of 1 to this option, and the rank of 2 to “no,” and vice versa. For each SnapCap feature preference evaluation (for current and future features), we reported the Wilcoxon p-value and Z-score. For statistically significant results, we reported the effect size using Pearson’s correlation coefficient (r). We plotted nurses’ preferences for various hands-free interaction tasks as the normalized difference between the sum of ranks for feature preferences (based on the SPSS analysis results provided in S4 Fig).

For the application preference analysis, in which users had the choice of Epic Haiku, Google Glass, no difference, or neither, we excluded responses that entailed only the “neither” option because they did not indicate a specific preference. We converted the other three options (Epic Haiku, Google Glass, and no difference) to ranks for the three possible preferences. If the Epic Haiku option was selected, we assigned the rank of 1 to “Epic Haiku” and the rank of 2 to the “Google Glass” option, and vice versa. If the no difference option was selected, we assigned the average rank of 1.5 to both the “Epic Haiku” and the “Google Glass” options. For each application preference comparison, we reported the Wilcoxon p-value, Z-score, and, whenever the results were statistically significant, the effect size using Pearson’s correlation coefficient (r). We graphed nurses’ preferences for SnapCap versus Epic Haiku for wound photography as the normalized difference between the sum of ranks for application preferences (based on the SPSS analysis results provided in S4 Fig).

For both the feature evaluation and application comparison, the difference between sum of ranks is W⁺ − W^-, where W⁺ is the positive sum of ranks ("prefer" and "Google Glass") and W^- is the negative sum of ranks ("do not prefer" and "Epic Haiku"). The normalized difference between the sum of ranks is (W⁺ − W^-) / W_max, where W_max = n * (n + 1) / 2 = 16 * 17 / 2 = 136.

To examine image quality, we retrieved the images taken by nurses using the SnapCap System and segmented images into the categories: (i) acceptable for clinical use, (ii) blurred, (iii) tilted, (iv) improperly framed, (v) under/over-exposed, and calculated the number and percentage of images in each category.

To examine the qualitative data captured at each user session, the research team reviewed the recommendations for system improvements that were noted by nurses on each questionnaire and added this information to the Excel spreadsheet, by the corresponding system feature. Two team members reviewed the notes and digital recordings captured by researchers at each user session and added comments and observations to the spreadsheet, corresponding to each application’s feature or attribute.

To analyze the Google speech-to-text (STT) data, captured in the follow-up user sessions, we used Word Error Rate (WER) as a standard measure for examining transcription accuracy. WER is defined as the edit distance between the reference word sequence and the sequence emitted by the transcriber. WER = (S + D + I) / N, where S, D, and I are the number of substitutions, deletions, and insertions, respectively, and N is the number of words in the reference [29]. Table 2 illustrates each of these error types.

Barcode Scanning for Patient Identification.

A Wilcoxon Signed-ranks test indicated that there was a statistically significant preference for hands-free barcode scanning for routine clinical care, Z(15) = -3.873, p < 0.001, r = 0.71. All sixteen nurses successfully used the SnapCap Glassware to read the patient barcode within four seconds. One nurse indicated that, “Google Glass has the ability to barcode patient identity to reduce errors.”

Voice-based Documentation through Video.

The data analysis indicated that voice-based documentation through a brief video recording was strongly preferred by nurses, Z(15) = -2.84, p = 0.005, r = 0.52. The qualitative data also indicated that nurses favored the use of voice commands for launching the video recording and image-capture features of the SnapCap System. However, the data revealed that current voice-commands were challenging for participants to use. Only two of sixteen (12.5%) nurses succeeded in using voice commands to launch the video recording feature on their first try. Diverse failure modes included saying an incorrect phrase (3 nurses), saying a phrase too quickly (1 nurse), or saying a phrase too softly (1 nurse). While documenting a wound through a verbally annotated video, nurses frequently commented that the 10-second duration was too short, and that additional time was required.

Double Blinking to Take Photographs.

Nurses took a photograph (made the camera shutter open and close) by double blinking. Overall, double blinking was well received, with a statistically significant preference for this feature, Z(13) = -3.606, p < 0.001, r = 0.71. One nurse took five or six photographs inadvertently because Google Glass registered her natural eye-blinking rate as double blinking. Another cautioned that double blinking “may be too variable depending on the individual.”

Head Tilt for Zooming.

Nurses achieved camera zoom in and out by tilting their head to the right and left, respectively. Overall, this was the least preferred hands-free interaction method, Z(10) = -1.897, p = 0.058. One nurse commented that the zoom was slow to respond, and three more indicated that this feature might interfere with their ability to assess wounds or pressure ulcers (for example on the back or the heel), which usually required them to tilt their heads from side to side. They felt that seeing the image zoom in and out on the eyepiece display could potentially interfere with their ability to take a high quality photograph. One nurse had recently been in a car accident and felt slight pain when tilting her head from side to side.

In its present form, the head tilt for zooming interaction suffers from the “Midas Touch” Problem [30]. At first, it seems useful to tilt one’s head from side to side and have the preview zoom in and out. Before long, especially in real-world situations, the constant zooming in and out becomes difficult as users unconsciously move their head to achieve a better view of an object of interest.

Speech-to-Text Annotation.

When we presented nurses with the possibility of annotating a wound through speech-to-text transcription, they were generally excited about this feature, but anticipated having to access the text for review and editing if necessary.

From data gathered during the follow-up user session, we evaluated the performance of the Google speech-to-text (STT) engine for transcribing wound annotations. Due to technical issues, two of the sixteen annotations had missing data—in one case, the first seventeen words were missing from the transcript; in the other, the last four words were absent. Additionally, we noticed three instances when a nurse uttered an incorrect word, so we removed these individual extraneous words from the data set.

The overall WER of Google’s STT for the fictitious wound annotations was 21.0%. For individual nurses, the WER ranged from 7% to 38% (SD = 10.0%). The word with the highest Individual Word Error Rate, or IWER [31], was “malodorous” (181.3%), which the Google STT reported as “my daughter is” (three instances), “mail order is” (two instances), and “mellow Doris” (one instance). The IWER can be greater than 100% because transcribing “malodorous” to “my daughter is” results in three errors: one substitution and two insertions. “Periwound” had the second highest IWER (140.0%), and Google’s STT had difficulty transcribing the utterance, “periwound tissue,” and generated results such as “Perry won T-shirt” and “Harry born to shoot.” Surprisingly, the word “wound” had the third highest IWER (66.7%), with Google’s STT suggesting “one” instead of “wound” in thirteen instances.

The following subjective comments were worth noting: Five nurses expressed a concern about the performance of the system due to their accents. Three nurses uttered the punctuation—“comma” and “period”—when reading, which the Google STT correctly interpreted. Three nurses thought that the period of silence after which the Google STT considers the utterance complete was too short. One nurse suggested that we load a specialized grammar feature with wound-related terminology to improve speech recognition performance.

From a technical standpoint, the Google STT performed well, in conjunction with the Glass hardware, in a majority of the tests. Fourteen out of sixteen (87.5%) nurses successfully saved the two wound annotations in four attempts or less (although examination of the text files revealed a small amount of missing data on two of the fourteen data files). Two of sixteen nurses (12.5%), however, required more than ten attempts to successfully save the two wound annotations. In both cases, the problem with saving annotations were due to poor network connectivity, which caused the program to hang-up in mid-transcription. The solution required physically moving to another part of the hospital with better wireless reception and recording the annotations in smaller text segments (one sentence at a time) to minimize the chance of malfunction.

Sterile Image-Capture Technique.

A Wilcoxon Signed-ranks test indicated that there was a statistically significant preference for the Glass SnapCap System versus Epic Haiku in regard to sterile image-capture technique when photographing wounds, Z(16) = -3.873, p < 0.001, r = 0.68. Comments relating to the SnapCap system included: “Much better experience,” and “Not keen on touching the Glass frame with gloved hands during a consult.”

Photo Capture Capability.

The results of the data analysis were not statistically significant, and thus insufficient to suggest a difference between the SnapCap System and Epic Haiku for photo capture capability, Z(16) = -1.667, p = 0.096. Also, the data indicated no statistical difference in preferences between the two applications for previewing images after capturing a photograph and before transmitting to an EMR, Z(16) = -0.832, p = 0.405. However, the qualitative data indicated that nurses favored the ability to preview images on a relatively large iPhone screen, as opposed to the smaller Glass display. One nurse commented, “Google Glass has the ability to barcode patient identity to reduce errors, but needs better pixel definition.”

Image Quality.

Similarly for image quality, the results were not statistically significant, indicating a lack of evidence to support a difference in preferences between the image quality of the SnapCap System and Epic Haiku, Z(15) = -0.816, p = 0.414. Nurses agreed that the quality of photographs taken should be sufficient for making clinical decisions, and qualitatively perceived image quality as a weakness of SnapCap. Using SnapCap, the sixteen nurses transmitted thirty wound photographs from Glass to the smartphone app, of which we considered seventeen of thirty (56.7%) to be of acceptable quality for clinical use. Challenges included blurred (7/30, 23.3%), tilted (3/30, 10%), and improperly framed (3/30, 10%) photos (Fig 8). Blurred shots resulted when the participant took the photo too close to the subject or moved during the shot. The most likely cause for tilted shots was taking a photo while still zooming using head tilt, and improperly framed shots may have been due to inadvertent double-blinking. Interestingly, these challenges are similar to those faced by blind users trying to take photographs when seeking help in obtaining visual information [32]. There were no under/overexposed shots, as the study took place in a hospital room with adequate lighting.

Overall Ease of Use.

In terms of overall ease of use, the results were not statistically significant; there was insufficient evidence to claim a difference in preferences between the SnapCap system and Epic Haiku for this attribute, Z(14) = -1.732, p = 0.083. Given the lack of familiarity that the nurses had with head-mounted displays, we see the limited difference in ease of use preferences as an encouraging result. One nurse commented that, “Epic Haiku is more ‘normal’ for today's registered nurses, but Google Glass has some great possibilities.” Another nurse called using SnapCap for wound photography, “a much better experience” then Epic Haiku.

Use of a Head-Mounted Display to Share and Discuss Images.

A Wilcoxon Signed-ranks test indicated that there was a statistically significant preference for the future use of head-mounted displays for sharing and discussing wound images among colleagues, in order to obtain real-time feedback on a diagnosis or to aid in clinical decision-making, Z(13) = -3.606, p < 0.001, r = 0.71. One nurse commented that she felt such a system was, “an interesting idea” and another mentioned, “it is a definite possibility.”

Historical Image Retrieval for Data Recall and Sharing.

The data illustrate a significant preference for historical image retrieval, to achieve time-lapse image recall in a head-mounted display after taking a series of photographs, Z(14) = -3.742, p < 0.001, r = 0.71. Suggested uses for this feature included the ability to see changes in wound margins over time, and to track a wound’s staging and healing progression (for stage 4 to stage 1 pressure ulcers). One nurse commented that it “would show progression or deterioration of the wound.” Another mentioned, “this is very important for staging pressure ulcers—one must know the previous stage so one does not downstage the ulcer.”

Dynamic Digital Ruler inside the Glass Eyepiece.

The data revealed a significant preference for a digital ruler inside the eyepiece of a head-mounted display, to replace a hand-held paper ruler, Z(13) = -3.051, p = 0.002, r = 0.60. Comments included: “Need to work on accuracy, but yes, that would be great so that you wouldn’t have to hold or dispose of the ruler,” and “Huge help!”