Analysis: A Step Toward Standardizing Diabetic Foot Images

Abstract

In this issue of Journal of Diabetes Science and Technology Yap and coauthors report on the design and present results from a reliability study of the application called FootSnap that is intended for standardization of the distance and the orientation of the camera relative to the diabetic foot that is photographed. To produce standardized photographs FootSnap uses a straightforward, yet original and useful method. However, other systems have been reported and clinically tested earlier, in which the same parameters of foot images might be standardized and which are free of some limitations of FootSnap. Nevertheless, FootSnap should contribute to increase clarity of documenting changes in the diabetic foot shape, texture color and lesions, and it should facilitate application of automatic image processing algorithms.

In this issue of Journal of Diabetes Science and Technology Yap and coauthors report on the design and present results from a reliability study of the mobile application for an iPad called “FootSnap.” ¹ The app is intended for standardization of the distance and the orientation of the camera relative to the diabetic foot that is photographed. To produce such standardized photographs FootSnap uses a straightforward, yet useful method, that is, two images are created when a particular foot is photographed for the first time—the first photograph is an ordinary plantar foot image, whereas the second one is an outline of the foot called a “ghost image.” The ghost image can be superimposed over a view of the same foot produced by the iPad during consecutive uses of the system, which can help to appropriately align the picture. Photographs are taken using the blue color background and the lightning module with standardized luminosity to further normalize foot images and facilitate operation of the image analysis algorithms.

Using the Jaccard Similarity Index (JSI) as a measure of reproducibility of foot images taken using FootSnap by two operators on two separate occasions, Yap and coauthors demonstrated high reliability of the proposed method with intra- and interoperator mean JSI values of 0.89-0.91 for diabetic feet and 0.93-0.94 for the control feet, respectively.

In the recently published systematic review on prevention of foot ulcers in the at-risk patients with diabetes, where a total of 30 controlled studies (including 19 randomized control trials) and another 44 noncontrolled studies were assessed, authors claim that more clarity is required in description of study populations, interventions, outcomes and outcome assessment. ² Application of a tool such as FootSnap would undoubtedly contribute to increase clarity of documenting changes in the diabetic foot shape, texture color and lesions during the longitudinal clinical trials or the routine clinical monitoring procedures. From this standpoint, FootSnap is a step in the right direction for the standardization of the diabetic foot images. It not only helps to align foot images on photographs taken during consecutive examinations of the same patient but it also tries to standardize background and the lighting conditions.

The approach of the inventors of FootSnap is original. However, other systems have been reported earlier, in which the same parameters of the plantar foot images as in FootSnap might be standardized.^3,4 These systems were manufactured as stand-alone, portable devices, in which either a still image digital camera or a flat scanner was used to acquire foot images. The distance and position of the foot against the camera or scanner, as well as the lighting conditions, remain constant due to the proper construction of these systems. The FootSnap as well as both the above-mentioned systems are restricted to take photographs of the plantar part of the foot. Still, this is not a serious limitation as majority of the lesions in the diabetic foot patients appears in the plantar region. Nevertheless, it seems that FootSnap might be able to document also other parts of the foot, if necessary, whilst the design of the other two systems virtually precludes this. In contrast, these systems are free of some other important limitations of FootSnap. First, in both cases, unlike FootSnap, foot images could be taken by patients without an assistance of any other person. Thanks to this feature as well as a compact design, these devices can be used not only in ambulatory settings but also in patients’ homes as a part of home telecare systems. Second, thanks to the design of these devices, the plantar surface of the foot being photographed is parallel to the surface of the image sensor of the digital camera or scanner, which is not guaranteed in case of FootSnap. This feature makes it possible to minimize errors made during manual or automatic estimation of the surface area of diabetic foot ulcers, while assessing effectiveness of the applied wound healing therapy. It is especially important in light of the recently updated recommendations of the Wound Healing Society, which treats percentage change in wound area of diabetic foot ulcers over 4 weeks of treatment as a predictor of effectiveness of therapy and likelihood of healing, and advise to reevaluate the therapy and to consider other treatments in patients who fail to show a reduction in ulcer size by 50% or more after 4 weeks of the therapy. ⁵ It is noteworthy that both above-mentioned systems have been positively assessed in pilot clinical tests.^3,4,6,7

According to the intention of the inventors of FootSnap, the standardized foot photographs should be analyzed using advanced computer image processing algorithms to help in the detection and monitoring of diabetic foot pathologies. From this point of view, FootSnap is also a step in the right direction, but, most probably, further image preprocessing steps will be necessary, for example, to ensure that plantar surface of the foot in the ghost image is parallel to the surface of the image sensor of the camera or to adjust and equalize color and brightness of foot images.

For several reasons it is not possible to judge, based on the results presented in the article by Yap and coauthors, whether the ghost-image method is effective enough to create ready-to-use images for automatic multitemporal analysis and monitoring of the diabetic foot lesions, for example, healing of the diabetic foot ulcers. The most significant limitations of the reliability study of the FootSnap app concern: relatively small study group, a lack of patients with the diabetic foot ulcers among participants, engagement of only two operators taking pictures on only two separate occasions and, most importantly, the use of only one similarity parameter, that is, JSI. The JSI values in the range of 0.89 to 0.94 appear to be respectable, but based on them it is very difficult to estimate the JSI values for feet with diabetic ulcers. It is also difficult to judge how a particular value of JSI is associated with measures that are clinically more meaningful, such as an accuracy of estimation of the diabetic ulcer surface reduction between two consecutive examinations. Hence, further studies are required to answer these questions. Nonetheless, even if further transformation of foot pictures on two or more photographs is required to register them (ie, to fine tune their spatial alignment) for multitemporal analysis, this transformation should be easier to perform using pictures acquired with FootSnap than those acquired with any standard ‘human-operated’ camera.

I agree with Yap and coauthors that the technology proposed in their paper may “represent the first stage toward a meaningful step forward in the prevention and management of diabetic foot pathologies” as far as the management of diabetic foot pathologies is concerned. As for the prevention, the evidence base to support the use of any intervention to prevent a first foot ulcer in the at-risk patient is almost nonexistent. For the prevention of recurrent foot ulcers, home monitoring of foot temperature, pressure-relieving therapeutic footwear, and certain surgical interventions proved to be effective. ² It is unlikely that monitoring of changes in plantar foot images could alone contribute to prevention of foot ulcers, but such monitoring is certainly an important component of the holistic system of care of the at-risk population.