Plastic Surgery

AI ultrasound system sharply improves burn depth classification over clinical exam

An AI-based system integrating tissue Doppler elastography and harmonic B‑mode ultrasound classified burn depth with 95% accuracy for third-degree burns in humans. Performance exceeds reported diagnostic accuracy of experienced burn surgeons, particularly for distinguishing deep partial from full-thickness injuries. The model was trained in a pig burn model and then prospectively tested in 30 thermal burn patients, with biopsy confirmation in operative cases. This noninvasive tool could better guide timing and extent of excision, especially for less-experienced burn or acute reconstructive surgeons. 1

Acute rejection remains the rule, not the exception, in VCA despite modern immunosuppression

This systematic review of 136 VCA recipients found acute rejection in about 60% within the first postoperative year. Rejection most commonly presented with Banff grade I–III skin changes such as lesions, erythema, edema, and rash on the graft. Corticosteroids, often methylprednisolone with or without topical tacrolimus, formed the backbone of treatment across centers. Additional agents, including antithymocyte globulin, alemtuzumab, mycophenolate, and rituximab, were used variably, with no clear superior regimen identified. Chronic rejection remains the leading cause of graft failure, underscoring the need for standardized protocols and better long-term strategies. 2

Bilamellar bionic scaffold supports integrated posterior lamellar eyelid regeneration

Investigators developed a heterogeneous bilamellar scaffold (bGA‑ADM) mimicking tarsus and conjunctiva to repair posterior lamellar eyelid defects. The scaffold’s patterned basement membrane and spongy dermal layer supported stratified growth of fibroblasts, endothelial cells, meibomian epithelial, and mucosal cells in vitro. In rabbit ear and eyelid models, bGA‑ADM enhanced vascularization, matrix deposition, and conjunctival resurfacing, approximating native posterior lamellar architecture. It also created an anti-inflammatory microenvironment conducive to meibomian gland regeneration, suggesting potential for functional posterior lamellar reconstruction. 3

AI-enhanced simulators offer objective, adaptive microsurgical training but rest on weak evidence

This systematic review identified 13 studies using AI or machine learning to augment microsurgical training simulations. Most systems used convolutional neural networks for instrument tracking, motion analysis, or real-time coaching, with median accuracy around 84%. AI-enhanced training generally reduced technical errors and improved learning curves compared with traditional methods in simulated tasks. However, studies were small, mostly single-centre, at high risk of bias, with very low-certainty evidence and poor external validation. Clinicians should view current platforms as promising adjuncts for skills acquisition, not yet as validated surrogates for real-case performance. 4