Vanderbilt University Biomedical Engineering Department Senior Design 2011 The Digital Craniometer Application An iOS application to aid in the screening of plagiocephaly in infants Chris Heelan – Cody Hall – Jorge Perez Robert Galloway (Advisor) Contents Abstract ......................................................................................................................................................... 1 Introduction ................................................................................................................................................... 3 Current Methods for Cranial Measurement .......................................................................................... 3 Operational Scenario ................................................................................................................................. 5 Initial Design Concept .............................................................................................................................. 6 iOS Concept for Image Based Craniometer .............................................................................................. 7 Methodology ................................................................................................................................................. 9 Results ......................................................................................................................................................... 14 Conclusion .................................................................................................................................................. 16 Recommendations ....................................................................................................................................... 17 Table of Figures/Tables Figure 1. Malformations of Infant Cranium.................................................................................................. 3 Figure 2. Normal (left) vs. Supported Sleeping positions ............................................................................. 4 Figure 3. Plagiocephaly Diagnosis vs. Age for treatment with positional modification .............................. 4 Figure 4. Aggressive treatment method using corrective band ..................................................................... 5 Figure 5. Initial Design Concept Digital Caliper .......................................................................................... 6 Figure 6. Requirements Hierarchy ................................................................................................................ 7 Figure 7. Offset marker ............................................................................................................................... 13 Figure 8. Splash Screen and Menu Screen .................................................................................................. 14 Figure 9. Measurement Screen.................................................................................................................... 15 Table 1. Collected Accuracy Data .............................................................................................................. 16 Abstract The purpose of this senior design project was to provide an easy, accurate and fast solution for measuring critical measurements of an infant’s head. Plagiocephaly is a condition where an infant has a deformed head due to either gestational or positional pressures on the skull. 70% of all infants will be screened for plagiocephaly in the first two years after birth. Given the percentage of children screened and the ability to better treat malformations at early stages, it is beneficial to have a simple and accurate method to diagnose plagiocephaly early in a child’s development. The result of our project was a fully functional manual iOS application built to iPhone 4 specifications that can determine critical diagonal measurement ratios and angle measurements useful for early detection of plagiocephaly in newborns. The project could be expanded in the future to include image processing and a more automated application by incorporating edge detection and absolute measurement options. Introduction Current Methods for Cranial Measurement Our team interviewed two pediatricians to determine what current methods are applied to measuring baby heads. We discovered that purely conventional means, such as tape measure, are employed each day to determine critical lengths. Measurements are then hand calculated on paper or with a calculator. This scenario required multiple steps to complete and sometimes became redundant with room for error according to the initial feedback from our surveys. Figure 1. Malformations of Infant Cranium Current methods also include plastic calipers that help speed up the process of diagnosing plagiocephaly. Two sets of measurements are useful for determining the extent of plagiocephaly and other deformations. Figure 1 shows an example of plagiocephaly on the left.5 Not only is there a difference in diagonal measurements but there is also an angle between the midline of the head and the ears that is not orthogonal. This is also useful for a specialist to use for verifying the correct treatment option based on the degree of malformation. Treatment options include positional correction by relieving pressure on the back of the head during sleep or by a more dramatic means of a corrective helmet.4 Figure 2. Normal (left) vs. Supported Sleeping positions Figure 3. Plagiocephaly Diagnosis vs. Age for treatment with positional modification Figure 4. Aggressive treatment method using corrective band Operational Scenario The purpose of the project was to measure the necessary dimensions of an infant's head in order to determine if they are developing plagiocephaly. These measurements included two linear distance measurements (combined to calculate a ratio) and an angular measurement (indicating alignment of the ears relative to the mid-line of the head). If the values of these measurements pass certain thresholds, the pediatrician will recommend the infant to a specialist. To accomplish this goal, we created an iPhone application to be used by pediatricians. The device utilized the camera API, allowing the user to take a new picture or load a previously taken picture. Once the picture was loaded, markers appeared giving the user the ability to drag these markers to the positions on the head that the doctor would like to measure (placement differs depending on whether the ratio or angle calculation is needed). In ‘Ratio-mode’, the application measured the ratios of the diagonal distances across the head along with the angle from the mid line. A ratio was calculated and displayed as an output to the doctor. Likewise, in ‘Angle-mode’ the application calculated the angle between the lines designating the mid-line and the ear-line. The angle will be displayed as an output. With the ratio and angle values, the pediatrician can determine if the infant should seek a specialist’s help. Initial Design Concept At the onset of the project our team wanted to create a physical device to measure the required distances in a more automated fashion. We called this concept ‘digital craniometer calipers.’ Our team designed a hat that could be placed on a child’s head that would read out critical measurements on an LCD screen. Figure 5. Initial Design Concept Digital Caliper This concept drawing in Figure 5 shows the inner workings of the digital caliper. This would be sewn into a hat and fit over a child’s head. The drawing on the left shows the arms that would be finely positioned after the hat is on. The drawing to the right shows the top view, where an LCD screen could be read while the hat is on the baby. iOS Concept for Image Based Craniometer In order to utilize existing equipment and distribution available to over half of healthcare professionals today1, an image based solution was considered the best solution for a semiautomated measurement system. The iPhone was chosen as the platform for our new solution due to the high rate of adoption within healthcare professionals and the success of the Apple App Store. An image guided approach proved to be the strongest alternative to a physical device because of the ability to extend normal diagnosis to future applications available on network connected devices running iOS. In a certain scenario, a physician can upload the photograph of a patient to an electronic medical record and this photo would document disease state and be available to share with specialist should they need to be consulted. Figure 6. Requirements Hierarchy Figure 6 displays the requirement hierarchy for the iOS image based craniometer. This shows each of the necessary considerations for building our device on an image based platform. Methodology The project’s initial approach involved designing a physical device that would be placed on the child’s head to gather the needed measurements. The basic concept involved anchoring two sets of calipers to a central electronics housing. Each set of calipers contained a digital angle sensor thereby providing the needed data (along with the dimensions of the calipers) to calculate the linear distance of the desired diagonals of the infant’s head thereby providing the values needed to determine the infant’s plagiocephaly ratio. Additionally, a digital angle sensor connected between the two sets of calipers allowed the user to find the midline to ear-line angle by placing one set of calipers along the midline and the other along the ear-line. This design required mechanical design and fabrication, electronic hardware selection, PCB design, and microcontroller coding in order to attain a final product. After presenting this idea to our advisors and their graduate students, a much simpler approach than a physical device was realized. Since the required ratio and angle measurements were both relative (as opposed to absolute) measurements, the values could be determined using arbitrary units such as the pixels of a digital image. Furthermore, the image analysis required to find the ratio and angle values was simple enough that the calculations could be computed with much less processing power than is available on most desktop and laptop computers. In fact, the calculations could be performed by most smart phones which also usually include a fairly high resolution camera. By designing an application for a smart phone instead of creating a physical device, the complexity of the project was reduced from hardware and software development to solely software. This transition also removed any physical contact with the child (safer for the infant and easier for the pediatrician) as well as limiting the cost of the project since the hardware platform would already be owned by the pediatrician. Due to the recent release of the iPhone 4 and the iOS 4 operating system, this phone was chosen as the platform for the digital craniometer. Further research showed that about 41% of pediatricians currently use an iPhone indicating that our application would have an initial market base.1 The 5 megapixel camera, 960x640 touchscreen at 326ppi, and an 800MHz A4 processor indicated that the hardware of the iPhone 4 was most likely more than adequate for the application’s required precision (approximately 1mm) and calculations.2 This assumption would later be verified during application testing. Finally, the App Store offered a simple and quick method for mass distribution of the application both domestically and internationally. While all members of the digital craniometer team had coding experience, none had ever designed an application for a smart phone. Therefore, the first stage of developing the application involved heavily researching not only iOS application development but also object-oriented programming in general. This was accomplished using a video lecture series found on iTunes University from Stanford University along with numerous textbooks. Once familiar with Apple’s software development kit (SDK), an initial application was created that included a splash screen, menu screen with two buttons (‘Load Image’ and ‘Camera’), and a measurement screen consisting of a pre-loaded image of an infant’s head (navigated to by pressing the ‘Load Image’ button). This application was then expanded by adding markers to the measurement screen that could be moved around the screen by the user’s touch gestures. A ‘Calc. Angle’ button was then added to the measurement screen which displayed the angle between the midline (designated by a white pair of markers) and ear-line (designated by a blue pair of markers) in a message box. A description of how this angle was calculated is shown below: 1) Calculate the midline and ear-line vector by subtracting the x and y-values of the white and blue markers respectively 2) Calculate the angle between the vectors using the following equation: 𝐴𝑛𝑔𝑙𝑒 = 𝑎𝑡𝑎𝑛2 (𝑚𝑖𝑑𝑙𝑖𝑛𝑒𝑦 , 𝑚𝑖𝑑𝑙𝑖𝑛𝑒𝑥 ) − 𝑎𝑡𝑎𝑛2(𝑒𝑎𝑟 − 𝑙𝑖𝑛𝑒𝑦 , 𝑒𝑎𝑟 − 𝑙𝑖𝑛𝑒𝑥 ) eq. #1 Likewise, a ‘Calc. Ratio’ button was added to the measurement screen which initiated the process of finding the desired diagonals and calculating their ratio. A description of this process is shown below: 1) Calculate the slope of the midline using the x and y-values of the white markers 2) Calculate the slope of the longest diagonal using the x and y-values of the blue markers 3) Calculate the y-intercept of the midline line using the slope and one of the white marker coordinates 4) Calculate the y-intercept of the longest diagonal line using the slope and one of the blue marker coordinates 5) Calculate the angle between the midline and the longest diagonal using eq. #1 6) Using the angle found above, rotate the white markers this angle from the midline using the following equations: 𝑥𝑛𝑒𝑤 = (𝑥𝑜𝑙𝑑 − 𝑖𝑛𝑡𝑒𝑟𝑠𝑒𝑐𝑡𝑖𝑜𝑛𝑃𝑜𝑖𝑛𝑡𝑥 ) ∗ cos(𝛼) − (𝑦𝑜𝑙𝑑 − 𝑖𝑛𝑡𝑒𝑟𝑠𝑒𝑐𝑡𝑖𝑜𝑛𝑃𝑜𝑖𝑛𝑡𝑦 ) ∗ sin(𝛼) + 𝑖𝑛𝑡𝑒𝑟𝑠𝑒𝑐𝑡𝑖𝑜𝑛𝑃𝑜𝑖𝑛𝑡𝑥 eq. #2 𝑦𝑛𝑒𝑤 = (𝑥𝑜𝑙𝑑 − 𝑖𝑛𝑡𝑒𝑟𝑠𝑒𝑐𝑡𝑖𝑜𝑛𝑃𝑜𝑖𝑛𝑡𝑥 ) ∗ sin(𝛼) + (𝑦𝑜𝑙𝑑 − 𝑖𝑛𝑡𝑒𝑟𝑠𝑒𝑐𝑡𝑖𝑜𝑛𝑃𝑜𝑖𝑛𝑡𝑦 ) ∗ cos(𝛼) + 𝑖𝑛𝑡𝑒𝑟𝑠𝑒𝑐𝑡𝑖𝑜𝑛𝑃𝑜𝑖𝑛𝑡𝑦 eq. #3 7) Calculate the slope and y-intercept of the new diagonal using the x and y-values of the white markers 8) Lock the movement of the blue markers to prevent accidental relocation 9) Restrict the movement of the white markers to only in the y-direction (user changes yvalue of white marker and x-value is automatically calculated using the slope and yintercept found in step #7) 10) Calculate the ratio of the diagonals by dividing the length of one (in pixels) by the length of the other (in pixels) Up to this point, the application had only been tested using the iPhone simulator included in Apple’s SDK. In order to implement the true functionality of the menu screen’s ‘Load Image’ and ‘Camera’ buttons, the application had to be loaded onto a physical iPhone because the camera application programming interface (API) could not be accessed from within the simulator. After purchasing a developer’s package from Apple that provided us with the necessary permissions to develop an application on an iPhone, the camera API was fully implemented in the craniometer application. Testing the application on an iPhone verified that the user could load a previously saved image or capture a new one by clicking the appropriate button on the menu screen. Once the application’s functionality had been tested, the team’s focus turned towards its usability. Creating new offset markers shown in Figure #7 allowed the user to see the marker’s point even while their finger was still on the screen. Figure 7. The revised offset marker allowed the user to precisely place the point of the marker without blocking the point with their finger. This provided for more precise placement of the markers. Certain marker locations were also locked during the ratio determination process thereby preventing the user from moving markers that would result in an inaccurate measurement. Finally, a triple-tap function was added that allowed the user to return to the menu screen by triple-tapping anywhere on the measurement screen. Using a gesture (triple-tap) to accomplish this prevented the measurement screen from becoming cluttered with buttons. Once the usability of the application seemed adequate to the design team, testing began to both verify the application’s accuracy and gather input about its usability. Ten people were given instructions on how to use the application. They were provided with a standardized photo with marked points for ratio measurements and angle measurements. The design team calculated the true ratios and angles by hand which provided true values to compare the collected data against. The collected data was averaged across subjects, and percent error was calculated to quantify the accuracy of the digital craniometer application. Lastly, usability suggestions were collected from the test subjects. Results The efforts described in the “Methodology” section resulted in an iPhone 4 application that was simple to use yet extremely effective at its purpose. When the application is launched, the user is shown a splash screen and menu screen as shown in Figure #8. Figure 8. The splash screen (left) and menu screen (right) are the first screens shown to the user after launching the application. After selecting the source of the image (the user’s photo library or camera), the image is loaded onto the measurement screen with two sets of markers (white and blue) as shown in Figure #9. Figure 9. The splash screen (left) and menu screen (right) are the first screens shown to the user after launching the application. The following process outlines how to find the ratio of the longest diagonals once an image is loaded: 1) Place white markers on midline and blue markers on longest diagonal 2) Press the “Calc. Ratio” button 3) Adjust the white markers to the edge of the infant’s head. Note that the x-value of the marker will be automatically calculated from the provided y-value in order to maintain the proper angle of the diagonal 4) Press the “Calc. Ratio” button 5) Result is displayed. Press the “Go Back” button to return to the measurement screen Similarly, the angle between the midline and ear-line can be found by following the procedure described below: 1) Place white markers on midline and blue markers on ear-line 2) Press the “Calc. Angle” button 3) Result is displayed. Press the “Go Back” button to return to the measurement screen At any time during either of these processes, the user may triple-tap the screen to return to the menu screen and start over. The data collected from the ten individuals showed that the application was extremely accurate while determining the ratio of the longest diagonals. Table #1 shows a summary of the collected data indicating that, on average, the application has a percent error of only 1.036%. Length A (cm) Length B (cm) Ratio Average User Results Percent Error 14 13.3 12.6 11.9 11.2 14 14 14 14 14 1 0.95 0.9 0.85 0.8 0.99731 0.94128 0.8834 0.84107 0.79119 0.27% 0.92% 1.84% 1.05% 1.10% Table #1. A summary of the collected accuracy data. Please note the very low percent error values. Unfortunately, a coding error made the midline to ear-line angle results invalid. This problem was quickly remedied once discovered and further testing suggested a similar accuracy as the longest diagonal data. This was as expected since the accuracy of the application has mostly to do with the resolution of the touchscreen. The most prominent usability suggestions included removing buttons from the measurement screen and automating the ratio and angle processes using edge-detection algorithms. While the latter suggestion was explored throughout the semester, a shortage of time prohibited the development of a fully automated application using image processing algorithms. However, the test subjects unanimously agreed that the application was simple to use and fairly robust in its operation. Conclusion In conclusion, the Digital Craniometer for iOS devices was a success. In regards to the actual functionality of the device, we were able to obtain testing that verified the accuracy of the device to calculate ratios between selected lengths on the screen. In the testing, we were able to obtain an average of slightly more than 1% error in the accuracy of the actual device. In regards to the accuracy of the test in determining the presence of plagiocephaly, it becomes dependent on the ability of the user to determine the correct segments to measure. Given correct usage of the device by either a trained professional or a user who is capable of following the directions, the device does have the ability to accurately determine the symmetrical integrity of the child’s head along with the angle of the ear-line. The subjectivity of the test itself comes into question when regarding this accuracy. The test, as given now, requires the use of a caliper (plastic, fork shaped device that is used to physically measure the linear distance). The user will estimate the longest diagonal distance across the child’s head and then measure the distance of the diagonal at the same angle to the midline. The issue with this is that it falls on the user to determine the longest diagonal along with accurately mirroring the angle across the midline. The goal with our application was to remove all of the physical aspects of the test (i.e. touching the child at all); however the subjectivity described above remains. While we did meet our specifications for a manual version of our application, we were unable, due to our time constraint, to create a fully automatic version of the application. With the help of edge detection image processing algorithms the application could become fully automated thereby eliminating even more error and increasing the speed and ease of the test. Recommendations This application has incredible potential in terms of revolutionizing the way initial plagiocephaly detections are done. Eliminating the need to physically contact the infant removes a removes any possible safety concerns while also making the test easier to perform on a squirming child. This alone is reason enough for most doctors to download and use our application for detection in their office. Looking forward, the design team hopes to release this application in the Apple App Store, possibly with the help of the tech transfer office at Vanderbilt. Submitting the application to Apple actually requires a much greater effort than if this application was released for the Android platform. While we will need to ensure our application follows all the guidelines for successful submission, the team is capable of fixing these bugs. In regards to the actual market, our application has enormous potential. As mentioned before, the devices now used to detect plagiocephaly in children are calipers, which cost an average of about $15. These devices are made of plastic and require manufacturing to produce them. One of the most important factors in the success of our product is the fact that the design team did not create a device; it developed software for existing devices. Therefore, in regards to the project as a viable product, it requires no overhead, manufacturing costs, or distribution costs. In addition, due to the chaotic environment of a pediatrician’s office, every room in the office would require a physical device such as a set of calipers. With our application, each doctor would only have to purchase the application once. And, at a cost of $5.99 per download, the application will cost a third of the price of one caliper. Not only would an office save 2/3 the cost of the device, they would also need to purchase less copies. According to research done by a California based consulting company, there is a rise in the popularity of smartphones among physicians. More specifically, it was reported that 94% of physicians use smartphones, with 44% of them choosing the iPhone over other models.1 With nearly 75,000 pediatricians in the US alone3, that leaves just over 31,000 pediatricians who currently own iPhones. This leaves an untapped, and essentially unchallenged (with the numerous benefits of our device over the current caliper) market with an earning potential of nearly $185,000. Sources Cited 1) http://www.knowabouthealth.com/smartphone-use-among-us-physicians-acceleratingrapidly/4502/ 2) http://www.apple.com/iphone/specs.html 3) http://www.aap.org/workforce/ 4) JA, P. (2008). MOC-PS(SM) CME article: managemetn consideratins in the treatment of craniosynostosis. Plast Reconstr Surg, 1-11. 5) McGarry, D. G. (2008). Head shape measurement standards and cranial orthoses in the treatment of infants with deformational plagiocephaly. Dev Med Child Neurol, 568-76.