Artificial Intelligence in Plastic Surgery of the Future – Role of Big Data, Artificial Neural Networks, and Machine Learning.

Big data refers to an extraordinarily large amount of data that needs to be processed and analyzed. Recent technological advances have made it possible for researchers to collect massive amounts of data that can be used to find patterns and make predictions. In recent times, this has become a major focus of AI (artificial intelligence) research. Big Data healthcare research focuses on large sets of data collected by researchers and physicians, as well as the big aggregates of smaller datasets in the electronic database of health sciences. The United States is trying to improve the quality of all surgeries across the nation, including plastic surgeries, by collecting data from national databases. One of these databases is CosmetAssure, which collects information about the results of adult plastic surgery and cosmetic procedures. This database can be analyzed efficiently with artificial intelligence models such as ANNs (artificial neural networks). Big data analysis with AI models such as ANNs can help researchers efficiently analyze the huge amount of information contained in these databases to answer clinically significant questions. Researchers used ML (machine learning) to determine relationships between method terminology and postoperative complications. They formed a supporting vector machine to determine procedural complexity and risk using data from the National Surgical Quality Improvement Project. The researchers compared their supporting vector machine score to other known measures of procedural complexity, which provide insight into how surgical complications may be addressed in future research. Machine learning applications in plastic surgery include predictive models that leverage ANN's pattern recognition capabilities to help surgeons make preoperative and postoperative decisions. The early 2000s saw some researchers develop a model that used data from a portable reflective spectrophotometer to determine healing depth and time with an accuracy of 86%. ANN was able to predict whether a burn would heal in 14 days or not with 96% accuracy. More recently, an application based on the camera of the Samsung Galaxy smartphone was used to evaluate the viability of a free flap based on its skin color. It was accurate in 75% of the patients, and the results were higher than physician estimates which were around 26%. In another study, applications were used on patients having varying degrees of venous and arterial occlusion, to see if the results were accurate or not with the new Predictive ML model. So they used different methods of testing on subjects and found that the predictions were fairly accurate. The ML model was able to accurately assess the vascular status of new subjects with a sensitivity and specificity of 94% and 98% respectively. Another predictive model of Machine Learning has been developed to predict the outcome of various nerve dispositions with an accuracy exceeding 90%. This indicates the potential benefits of ML models in predicting nerve dispositions. In a different study, researchers trained an Artificial Neural Network with several categories of variables identified from experimental records of nerve graft studies in rats. Many of these variables were complicated and included biomaterials, extracellular matrix proteins, growth factors, and their receptors, cytokine levels, disease state, and its severity, time-dependent changes, etc. In the future, Machine Learning systems should allow early diagnosis of a wide range of conditions by streamlining the analysis of clinical data. In another recent study, researchers developed an ML algorithm that is designed to diagnose different types of craniosynostosis based on CT imaging. The algorithm had a sensitivity of 92.7% and a specificity of 98.9%. Mental imagery is subjective, so surgeons used the ML model to help characterize dysmorphology before correcting these malformations. The ML model assisted in extremely effective surgical planning by being more accurate than our traditional approaches. Another supervised ML model has been shown to support surgical planning through automated diagnosis and simulation. Researchers created a 3D morphable model, a statistical model of facial shape and quality using databases containing 10,000 three-dimensional facial scans of healthy volunteers and orthognathic surgical patients. Three separate models were created: a global model with faces of healthy volunteers and preoperative patients, a preoperative patient model, and a postoperative patient model. The models were able to distinguish orthognathic patients from healthy volunteers and predict the patient-specific postoperative facial shape by regression analysis. Thus, these predictive models have the power to help the operating plastic surgeons in the objective evaluation of preoperative and postoperative aesthetics while improving patient education, thus resulting in much better surgical planning, execution, and results.