Kiranantawat, Kidakorn F.R.C.S.T.; Sitpahul, Ngamcherd M.D.; Taeprasartsit, Pinyo; Constantinides, Joannis F.R.C.S.(Plast.)(Ed.); Kruavit, Arthi F.R.C.S.T.; Srimuninnimit, Vichai F.R.C.S.T.; Punyahotra, Narong F.R.C.S.T.; Chatdokmaiprai, Chalermpong F.R.C.S.T.; Numhom, Surawej F.R.C.S.T.
Since its introduction in the late 1950s, microsurgical reconstruction using free flaps has evolved to be the method of choice for reconstruction of large or complicated soft-tissue defects, with reported success rates of at least 95 percent in leading centers.1–3 Despite the high success rate, free flap surgery is sometimes associated with technical failure, a catastrophic event. Salvage rates for compromised free flaps are inversely related to the time interval between the onset of vascular occlusion and reexploration.1,4,5 Therefore, effective monitoring and early recognition of free flap ischemia or congestion are essential to increase the possibility of complete flap salvage.
In 1975, Creech and Miller6 described the attributes of the ideal monitoring device for free tissue transfer. These are harmlessness to patients and flaps, rapid responsiveness, accuracy, reliability, and applicability to all types of transferred tissue. Despite many efforts for technical innovation, the mainstay of flap monitoring to date is direct clinical observation involving an experienced health care professional. Clinical parameters being observed for are flap color, temperature, capillary refill, and skin turgor complemented by pinprick testing if needed.7–10 Among these clinical parameters, flap color is recognized to be the most reliable clinical indicator of flap viability.11 A number of monitoring devices are currently in use, using surface temperature measurement, laser Doppler flowmetry, perfusion photoplethysmography, and other modalities. However, in spite of their effectiveness, these are rarely applicable in resource-limited settings because of their high cost and the requirement for staff specialist training to operate. New, effective, and easy-to-use flap monitoring devices are thus needed.
Recently, the advancement of smartphone technology has provided us with an excellent opportunity for the development of new applications that may benefit patients and health care providers globally. In this study, we introduce a novel application operating on the Android (Google, Inc., Mountain View, Calif.) mobile platform, and demonstrate its effectiveness in monitoring tissue perfusion.
PATIENTS AND METHODS
The study was approved by our university ethics committee, and verbal informed consent was given by subjects. The project was divided into two phases: application development, followed by validation.
For us to provide the software with background information on skin color for future use, we designed an experimental model of tissue ischemia in humans and recorded that photographically. Vascular occlusion was induced experimentally in healthy individuals, using the Vascular Diagnosis System (Nicolet Vasoguard P84; Viasys Healthcare, Inc., Waukegan Ill.) to generate transient ischemia of the index finger. The system deploys a circular cuff around the subject’s finger base connected to a regulated pressure generator (Fig. 1). Different pressures were applied to create varying degrees of vascular compromise of the index finger and the associated skin color changes12: 10 mmHg was applied to induce partial occlusion of veins, 50 mmHg to induce complete occlusion of veins, 30 mmHg less than the subject’s systolic blood pressure to induce partial occlusion of arteries, and 30 mmHg higher than the subject’s systolic blood pressure to induce complete occlusion of the index finger’s arteries. Images were taken after applying the above pressures for 3 minutes. The finger was allowed to rest for another 3 minutes between application of different pressures.
Sixty images were acquired from 20 subjects using the Galaxy S2 (Samsung Electronics Co., Seoul, Republic of Korea) smartphone and its camera. The software requires a camera with minimum resolution of 0.5 million pixels, which is common in every smartphone today. To control image quality variables for subsequent analysis and comparison, light intensity, camera-to-object distance, and white balance of all images were predetermined. We used a specifically created white box made of cardboard with one surface (to be the inner one), standard white in color, cut into the shape of an incomplete pyramid to the exact dimension described in Figure 2, left. With the white box, we could standardize camera-to-object distance to 7.8 cm and avoid out-of-focus images. The box also provided us with a chamber to achieve similar light intensity for all subjects (Fig. 2). The design of the box included small white stripes surrounding the area of the tissue to be photographed at the base of the pyramid, which were used for the white balance of images (Fig. 2, right). The phone camera was set to a continuous flash mode. Sample images taken are shown in Figure 3.
The Android application (SilpaRamanitor) was developed using the sample images acquired by the above technique to build an image bank. The application was thus taught to recognize skin characteristics of both compromised and healthy regions from the subjects. The correlation established between the degree of venous and/or arterial occlusion caused by the pressure regulator and the skin color on the images was introduced in the application as a baseline for future comparison. Skin color characteristics were used as features for pattern recognition performed by the k-nearest neighbor algorithm. The application categorized the skin area reviewed into three types: normal, venous outflow occlusion, and arterial inflow occlusion. Based on the severity of occlusion, each type was further subdivided to two grades, partial and complete. To evaluate a skin image for a new subject in the future, the application was asked to identify the set of bank images with the most similarity to the input image. A validation process followed to ensure reliable results. The user interface is shown in Figure 4.
One hundred twenty-six new images were acquired from 42 subjects and their index and middle fingers, using the same experimental setup. Subject age varied from 18 to 50 years. Exclusion criteria included previous injuries to the fingers examined, prior procedures involving the neurovascular structures of the upper limb, and chronic illnesses (particularly peripheral vascular disease or diabetes). Smokers were also excluded.
Three images of the index and middle fingers were taken from each subject. The first image contained both fingers under normal conditions. The second image was taken while pressure was applied around the base of the index finger using the device described above to occlude the digital veins, with the middle finger representing the control. The third image was taken when the arterial inflow to the index finger was occluded. For images 2 and 3, half of the group (21 subjects) was subjected to cuff pressures causing complete vascular occlusion, and the rest (21 subjects) had their vessels partially occluded. Each finger photograph was manually designated to be either the control (normal color of skin) or the test area (altered color of skin caused by various degrees of occlusion). By using the bank images collected during the development stage, the software was able to comment on the vascular status of the test area by declaring it either (1) normal, (2) partial venous occlusion area, (3) complete venous occlusion area, (4) partial arterial occlusion area, or (5) complete arterial occlusion area. (See Video, Supplemental Digital Content 1, which demonstrates use of the application to analyze tissue circulatory status, http://links.lww.com/PRS/B39.)
Video. Supplemental ...Image Tools
The diagnostic performance of the application was assessed for sensitivity, specificity, and accuracy. Sensitivity was calculated by
Equation (Uncited)Image Tools
, where Cv is the number of cases in which venous occlusion was correctly identified and Ca is the number of cases in which arterial occlusion was correctly identified, and Nv and Na are the number of cases where venous and arterial occlusions were actually present, respectively. Specificity was calculated from
Equation (Uncited)Image Tools
, where Cn is the total number of cases when the test area was normally perfused and the software correctly confirmed that, whereas Nn is the total number of cases with normal perfusion. Accuracy was computed from
Equation (Uncited)Image Tools
. Clinical application of the final product was initiated using a test case of a free transverse rectus abdominis musculocutaneous flap, with a positive result, which is also presented.
For the software validation process, as explained above, 42 healthy subjects were enrolled, 21 men and 21 women. All subjects are medical students and nurses at Ramathibodi Hospital. The mean age of the subjects was 33 years (range, 18 to 50 years). Using the formula described above, the software sensitivity to diagnose any degree of venous or arterial occlusion was calculated to be 94 percent (Table 1). The specificity was 98 percent and the accuracy was 95 percent. Thus, it was proven that the application is highly accurate in diagnosing vascular occlusion of the fingers using the color of skin as a reference. In addition, we calculated the ability of the software to diagnose the two different grades of occlusion severity (partial or complete). For the 39 cases correctly identified as having venous occlusion, the application accurately subclassified the occlusion severity in 33 cases (85 percent). Furthermore, for the 40 cases correctly identified as an arterial occlusion problem, the application correctly subclassified the occlusion severity in 33 cases (83 percent). These findings demonstrated that the application could not only very reliably identify the presence of vascular occlusion but also indicate occlusion severity for most of the cases.
The flap monitoring application was initially tested in a clinical setting in our hospital (Fig. 5). A 47-year-old woman underwent muscle-sparing free transverse rectus abdominis musculocutaneous flap surgery for immediate breast reconstruction after left modified radical mastectomy. The recipient vessels were the left internal mammary artery and vein. One artery anastomosis and one vein anastomosis were performed in an end-to-end fashion. The control image was obtained immediately after a flowing anastomosis, and the appearance was of a well-perfused flap following inset. Every hour postoperatively, the application was used to monitor the skin color and comment on vessel patency. The SilpaRamanitor indicated partial venous occlusion using the image obtained 8 hours postoperatively, which at that time was not apparent on direct clinical observation. One hour later, flap color was darker clinically and venous blood was present using the prick test. At this time, the SilpaRamanitor demonstrated complete venous occlusion, just before reexploration of the flap. During the operation, complete venous occlusion was observed, and the flap was successfully salvaged (Fig. 6). It became apparent to us that the application could detect a slight venous problem well ahead of clinical observation. This is very important for successful flap salvage.5
Apart from good preoperative planning and surgical technique, one of the most critical factors in the success of free flap surgery is postoperative monitoring. The earlier the problem is detected, the higher the flap salvage rate that can be achieved.5
To date, the most commonly used method for flap monitoring is direct clinical assessment. Although safe and reliable when expert staff are available, constant human input is required, which can be expensive and time consuming. Interobserver reliability can be an issue, and documentation for accurate communication, later process analysis, and feedback can be inaccurate. When other methods are concerned, only laser Doppler flowmetry, implantable Doppler, and quantitative fluorometry have resulted in a proven improvement of flap salvage rates.13 Other techniques—including microendoscopy, tissue oxygenation, microdialysis, ultrasound, plethysmography, tissue pH, surface temperature measurement, pulse oximetry, and spectrometry—that have been reportedly used to observe flaps in a small number of cases have not demonstrated an effect on salvage rates.7,13–15
From the three methods offering a proven improvement on flap salvage, laser Doppler has gained wider clinical acceptance in the monitoring of skin flaps, although it can suffer from calibration difficulties, high sensitivity to movement, and positioning or probe replacement errors.16 The implantable Doppler probe is an effective tool that improves the salvage rates, particularly for buried flaps; however, it is expensive, invasive, and increases operating time. Initial research demonstrated a 3 percent false-positive rate and a 5 percent false-negative rate for perfusion compromise when the probe was placed on the artery. Furthermore, a delay of up to 5 hours was found between a venous obstruction and the loss of the arterial signal in large muscle flaps. Best results occur if an implantable probe is placed on the vein instead of the artery, allowing the detection of venous obstruction immediately followed by detection of arterial thrombosis.10
Photography is reportedly used for flap monitoring as an adjunct to other modalities. A color image of the flap tissue may be taken after revascularization in the operating room and posted at the patient’s bedside. The nurse may then be instructed to call the house officer on a change in the appearance of the flap tissue, whether the tissue appears dark and bluish in color or pale and mottled.10 Previous studies demonstrated that the usefulness of telemedicine using mobile phones was hampered by the quality of the camera and images taken.17–25 Several factors influence the brightness and color of photographs, such as lighting, camera settings, and technical specifications for each camera. Thus, predetermined camera settings, light control, and postprocessing photograph standardization are crucial. Without these three important steps, the photographs obtained from different places, times, and environments will not be comparable. This might lead to a false interpretation of flap vascular status.
Since 2008, it has been demonstrated that the use of digital images and the Internet allows reconstructive surgeons to have an up-to-date and reliable picture of the state of the free tissue transfer.26 Hwang and Mun27 reported their use of smartphone and mobile messenger application for better and faster communication among health care members during postoperative monitoring following free tissue transfer procedures. Photographs of the flap and Doppler ultrasonography sound clips were uploaded and distributed to all team members using the Internet, creating nearly real-time communication and discussion. Their reported flap survival rates increased from 96.2 percent to 100 percent, flap salvage rates for take-back increased from 50 percent to 100 percent, and response time interval after notification of flap compromise was shortened from 4 hours to 1.4 hours.
Before the development of our application, it was common practice in our hospital for the on-call physicians to photograph the flap during postoperative monitoring for documentation and later review. While reviewing the photographs of failing flaps during our morbidity meeting, we noted color changes to be more obvious compared with direct clinical observation. We theorized that photography could provide us with a way of detecting subtle changes in color several hours earlier than clinical examination and indicate flap vascularity problems, especially if the images were compared electronically. This reflects the potential of visual light to offer adequate information to determine the status of the flap.
The white light control box provides limited, standardized space between the camera and the flap. The white color of the inner surface of the box creates an optimum reflection effect of the light originating from the continuous mode of the smartphone’s flash and ensures enough intensity of light for tissue photography. This results in a faster shutter speed and a sharper and clearer quality of image. The second benefit of using the light box is to prevent environmental light from interfering, a factor that can vary between different places and times. The light box also provides the lens with the optimal distance and angle for a satisfactory focus. With these standards in place, the quality, color, and brightness of photographs taken will be reproducible, regardless of place, time, and the surrounding environment. Furthermore, the white strips, which were folded in from each side of the box, provide great benefit for postprocessing photograph adjustment by the application. As different phones provide different cameras with quality, postprocessing adjustment by the application is crucial. Users need to select the area on the strip that is not contaminated by other color or blood. Then, the application will adjust the white balance and the brightness of the overall image according to the selected area on the white strips. With this technique, photographs taken using different phone cameras can be standardized and reliably compared.
Our team has developed the above smartphone application to assess tissue perfusion based on the analysis of skin color and demonstrate its predictive value from the calculation of method accuracy, sensitivity, and specificity. The sensitivity, specificity, and accuracy were 94, 98, and 95 percent, respectively. The application produced false-negative results for 6 percent of cases and false-positive results for 1 percent. In addition, the accuracy of the invented method in grading occlusion severity was evaluated. Accuracy in differentiating between different grades of venous and arterial obstruction was 85 and 83 percent, respectively. Furthermore, the advantages of this application are ease of use even by inexperienced personnel, ability to transfer data, rapidity, and less dependence of its accuracy on the skill of assessors. As shown in the sample case, early, imperceptible changes in skin color could be detected by this application. In addition, the device could be used to evaluate subtle color changes in patients with darker Fitzpatrick types of skin, which are naturally more difficult to observe. There are strong indications that this application might be more sensitive for detecting vascular compromise compared with the criterion standard clinical monitoring. However, future research using a large series of patients with free tissue transfer is needed to substantiate the above claim.
It should be noted that the accuracy of this method is dependent on the quality of the camera, which needs to be at least equal to the specification of the smartphone that was used in the study. At this stage of application development, the device cannot be used to monitor flaps without a skin paddle such as a muscle flap, buried flap, or intestinal flap, although future modification through the input of the appropriate baseline data should be able to allow this. With rapid progress in mobile phone camera technology and the appropriate software upgrades, this application and its future derivatives promise to offer a solution close to ideal as a monitoring system for patients and surgeons.
Our mobile phone–based diagnostic application allows physicians or allied medical staff to acquire skin images using a smartphone and a custom-made light box to identify whether vascular inflow or outflow obstruction is present within the transferred free tissue. Furthermore, this diagnostic task requires only seconds to perform and can be executed on-site without the use of other computing devices or Internet access. The pilot study executed demonstrated its high diagnostic accuracy and reliability. Together with being easily affordable and potentially superior to plain clinical observation—the aim of a future study—the proposed system offers significant potential for efficient flap monitoring.
The authors would like to acknowledge Stamatis Sapountzis, M.D., for contributing to this study.
1. Smit JM, Acosta R, Zeebregts CJ, Liss AG, Anniko M, Hartman EH. Early reintervention of compromised free flaps improves success rate. Microsurgery. 2007;27:612–616
2. Jones NF, Jarrahy R, Song JI, Kaufman MR, Markowitz B. Postoperative medical complications—not microsurgical complications—negatively influence the morbidity, mortality, and true costs after microsurgical reconstruction for head and neck cancer. Plast Reconstr Surg. 2007;119:2053–2060
3. Nakatsuka T, Harii K, Asato H, et al. Analytic review of 2372 free flap transfers for head and neck reconstruction following cancer resection. J Reconstr Microsurg. 2003;19:363–368; discussion 369
4. Siemionow M, Arslan E. Ischemia/reperfusion injury: A review in relation to free tissue transfers. Microsurgery. 2004;24:468–475
5. Chen KT, Mardini S, Chuang DC, et al. Timing of presentation of the first signs of vascular compromise dictates the salvage outcome of free flap transfers. Plast Reconstr Surg. 2007;120:187–195
6. Creech B, Miller SGrabb WC, Myers MB. Evaluation of circulation in skin flaps. Skin Flaps. 19751st ed Boston Little, Brown:21–38
7. Jallali N, Ridha H, Butler PE. Postoperative monitoring of free flaps in UK plastic surgery units. Microsurgery. 2005;25:469–472
8. Haddock NT, Gobble RM, Levine JP. More consistent postoperative care and monitoring can reduce costs following microvascular free flap reconstruction. J Reconstr Microsurg. 2010;26:435–439
9. Smit JM, Zeebregts CJ, Acosta R, Werker PM. Advancements in free flap monitoring in the last decade: A critical review. Plast Reconstr Surg. 2010;125:177–185
10. Salgado CJ, Moran SL, Mardini S. Flap monitoring and patient management. Plast Reconstr Surg. 2009;124(Suppl):e295–e302
11. Hentz VR, Mathes SJ Plastic Surgery. 2005 London, United Kingdom Saunders;
12. Korhonen P, Syvänen K, Aarnio P. Surrogates of large artery versus small artery stiffness and ankle-brachial index. Int J Angiol. 2011;20:167–172
13. Pratt GF, Rozen WM, Chubb D, et al. Modern adjuncts and technologies in microsurgery: An historical and evidence-based review. Microsurgery. 2010;30:657–666
14. Whitaker IS, Gulati V, Ross GL, Menon A, Ong TK. Variations in the postoperative management of free tissue transfers to the head and neck in the United Kingdom. Br J Oral Maxillofac Surg. 2007;45:16–18
15. Luu Q, Farwell DG. Advances in free flap monitoring: Have we gone too far? Curr Opin Otolaryngol Head Neck Surg. 2009;17:267–269
16. Payette JR, Kohlenberg E, Leonardi L, et al. Assessment of skin flaps using optically based methods for measuring blood flow and oxygenation. Plast Reconstr Surg. 2005;115:539–546
17. Chu Y, Ganz A. A mobile teletrauma system using 3G networks. IEEE Trans Inf Technol Biomed. 2004;8:456–462
18. Chu Y, Ganz A. A mobile teletrauma system for rural trauma care. Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society IEEE Engineering in Medicine and Biology Society.September 1–4, 2004San Francisco, Calif
19. Belala Y, Issa O, Gregoire JC, Wong J. A secure mobile multimedia system to assist emergency response teams. Telemed J E Health. 2008;14:560–569
20. Rohrich RJ. The Web and your cosmetic surgery practice. Plast Reconstr Surg. 2001;107:1253–1254
21. Hsieh CH, Tsai HH, Yin JW, Chen CY, Yang JC, Jeng SF. Teleconsultation with the mobile camera-phone in digital soft-tissue injury: A feasibility study. Plast Reconstr Surg. 2004;114:1776–1782
22. Tsai HH, Pong YP, Liang CC, Lin PY, Hsieh CH. Teleconsultation by using the mobile camera phone for remote management of the extremity wound: A pilot study. Ann Plast Surg. 2004;53:584–587
23. Kim DK, Yoo SK, Kim SH. Instant wireless transmission of radiological images using a personal digital assistant phone for emergency teleconsultation. J Telemed Telecare. 2005;11(Suppl 2):S58–S61
24. Chandhanayingyong C, Tangtrakulwanich B, Kiriratnikom T. Teleconsultation for emergency orthopaedic patients using the multimedia messaging service via mobile phones. J Telemed Telecare. 2007;13:193–196
25. Kim DK, Yoo SK, Park JJ, Kim SH. PDA-phone-based instant transmission of radiological images over a CDMA network by combining the PACS screen with a Bluetooth-interfaced local wireless link. J Digit Imaging. 2007;20:131–139
26. Varkey P, Tan NC, Girotto R, Tang WR, Liu YT, Chen HC. A picture speaks a thousand words: The use of digital photography and the Internet as a cost-effective tool in monitoring free flaps. Ann Plast Surg. 2008;60:45–48
27. Hwang JH, Mun GH. An evolution of communication in postoperative free flap monitoring: Using a smartphone and mobile messenger application. Plast Reconstr Surg. 2012;130:125–129