PLASTIC SURGERY III

Identification of wound margins using a radial gradientapproach to automated edge detectionMichael Januszyk MD, Kristine C Rustad, Jason P Glotzbach MD,Victor W Wong MD, Jerry S Chen MD,Michael T Longaker MD, MBA FACS,Geoffrey C Gurtner MD, FACSStanford University, Stanford, CA

INTRODUCTION: Assessment of gross wound closure representsa critical endpoint for both clinical and preclinical investigations;however variation among observer measurements frequently leadsto poor precision and reduced statistical power. To address this,we have developed an automated computer method to identifywound margins.

METHODS: Twenty-five mice (C57BL/6) were subjected to astented excisional model of wound healing and photographed atregular intervals (total � 350). Images were converted to 8-bit vec-tors and linear discriminant analysis employed to circumscribe eachwound mask, which was then automatically divided into an array of16 � 16-pixel regions of interest (ROIs) and classified according tooverlap with “gold standard” margins outlined by a senior plasticsurgeon. A set of 36 texture-based features was computed in eachROI and distilled to a single predictive weight based on trainingacross a two-layer neural network.

RESULTS: 75% of the images were randomly assigned to the “train-ing set” and the remaining 25% used to test the method. For each testimage, a radial-based gradient descent algorithm was employed atpi/32 radian intervals from the original wound center, and lesionmargins were identified as those pixel arcs with greatest change inscalar field transformation (as weighted above).

CONCLUSIONS: When applied to the training set, the methodachieved a sensitivity of 96.7% at a specificity of 98.1% (AUC �0.99). The trained program was then applied to the test set, andachieved an average difference in wound area of 1.7%, representing asignificant improvement over that of blinded surgical residents underidentical conditions (8.7%).

CD4� T-cells are necessary for fibrosis and lymphaticdysfunction in lymphedema and inhibitlymphangiogenesisAlan Yan MD, Jamie C Zampell MD, Tomer Avraham MD,Seth Z Aschen, Babak J Mehrara MD, FACSMemorial Sloan-Kettering Cancer Center, New York, NY

INTRODUCTION: We have previously shown that CD4� T-cellinflammation is associated with fibrosis and lymphatic dysfunctionresulting from chronic lymphatic stasis. The purpose of these exper-iments was to determine the direct effects of adaptive immune re-sponse on fibrosis and lymphatic regeneration.

METHODS: Chronic lymphatic stasis was produced by circumfer-ential skin excision and microsurgical ligation of the deep lymphaticsystem of adult female RAG mice (deficient in T- and B-cells), CD4knockout (KO), or wild-type littermates. In addition, wild-type micewere depleted of CD4�, CD8�, or CD25� cells using neutralizingantibodies. Control animals were treated with non-specific isotype

antibodies. Lymphatic function and lymphangiogenesis were evalu-ated 6 weeks after surgery.

RESULTS: CD4KO mice demonstrated minimal tail edema andfibrosis. In contrast, RAG mice had tail edema (2x � CD4KO;p�0.01) but no fibrosis as compared to wild-type controls. BothCD4KO and RAG mice demonstrated decreased expression of pro-fibrotic cytokines (IL13, IL4) as compared to controls. CD4KOmice or mice depleted of CD4� cells (but not CD8 or CD25) hadnearly 5-fold increased lymph node uptake of technetium-labeledsulfur colloid as compared to RAG or control mice (p�0.01). Inaddition, CD4KO mice or mice depleted of CD4� cells demon-strated markedly increased lymphangiogenesis (�2x) as compared tocontrols (p�0.01).

CONCLUSIONS: We have shown that CD4� T-cells are necessaryfor fibrosis and lymphatic dysfunction by regulating the expression ofprofibrotic cytokines. In addition, CD4 cells directly contribute tolymphatic function by inhibiting lymphangiogenesis. These findingshave important clinical implications because they may enable iden-tification of targeted treatments for lymphedema.

Epithelial control of dermal remodeling: Keratinocyte-specific deletion of focal adhesion kinase induces matrixmetalloproteinase activity post-injuryVictor W Wong MD, Michael Sorkin MD, Satoshi Akaishi MD, PhD,Kristine C Rustad, Jason P Glotzbach MD,Michael T Longaker MD, MBA, FACS, Geoffrey C Gurtner MD, FACSStanford University, Stanford, CA

INTRODUCTION: Deletion of keratinocyte-specific focal adhe-sion kinase (FAK) produces dramatic functional deficits in vitro,yet phenotypic changes have not been detected during in vivowound repair. However, previous studies utilized non-stentedmurine excisional wounds that heal through contraction ratherthan re-epithelialization/granulation tissue formation. We hypothe-size that use of a stented excisional model will reveal importantwound healing impairments in keratinocyte-specific FAK KO mice.

METHODS: Keratinocyte-restricted FAK KO mice were gener-ated using Cre/loxP technology. FAK deletion was confirmedusing PCR, immunoblot, and immunohistochemistry. Excisionalwound models were applied and tissue harvested at day 5 and 10.Matrix and basement membrane properties were assessed usingtrichrome, polarized light, and transmission electron microscopy.qPCR/immunohistochemistry were performed to analyze gene/protein expression of Mmp2, Mmp9, Timp1, and Timp2. Gelatinzymography was used to confirm matrix metalloproteinase activity.

RESULTS: FAK KO mice demonstrated a significant delay instented excisional wound closure. Dermal thickness was reduced by38% (p�0.001) in day 10 wounds in FAK KO mice. Expression ofMmp2, Mmp9, and Timp2 were significantly elevated and MMP2and MMP9 immunolocalization were stronger in FAK KO wounds,consistent with increased dermal remodeling. Further, gelatin zy-mography demonstrated significantly increased MMP2 and MMP9activity with keratinocyte-specific FAK deletion.

S96© 2011 by the American College of Surgeons ISSN 1072-7515/11/$36.00Published by Elsevier Inc. doi:10.1016/j.jamcollsurg.2011.06.000