ARTIFICIAL INTELLIGENCE AND MACHINE-LEARNING IN CRANIO MAXILLO-FACIAL SURGERY: OUR EXPERIENCE
Authors:
Umberto Committeri
Affiliation:
University of Naples Federico II, Naples, Italy
Doi: 10.54936/haoms242p99
ABSTRACT:
Objectives: The practice of surgery requires making rapid and complex decisions while managing the sometimes-uncertain health consequences for patients. In this regard, Artificial Intelligence (AI) have improved the surgeon’s southerly activity in all phases of patient management: screening, diagnosis, surgical procedure, and follow-up. It is specified how AI represents a modern technology with the ability to perform a task reserved for humans: learning, perception, reasoning, recognition. Machine learning (ML) a subdiscipline of AI, then, aims to design computer models capable of performing a task without having been explicitly programmed to do so. This type of learning is regularly employed in healthcare because of the complexity of data and high heterogeneity of patients.
The purpose of our study involved evaluating the effectiveness of ML in maxillofacial surgery.
Methods: ML capabilities were explored in several areas: trauma, oral surgery, and oncology to obtain an automated detection of maxillofacial fractures, to differentiate cystic lesions of the jaws, salivary gland tumors, and identify lymph nodal metastases in early stage of tongue cancers. The parameters analyzed regarded radiomic features extracted from MRI or CT associated then with inflammatory biomarkers obtained from blood sampling. Univariate analyses with nonparametric tests and multivariate analyses with machine learning approaches were used to validate the protocol.
Results: All clinical outcomes variables showed statistically significant differences (p < 0.05) in the Kruskal−Wallis test on median values. The multivariate analysis permitted to identify in each research field the best combinations of features and clinical parameters with a mean of 84% accuracy, 73% sensitivity, and 89% specificity.
Conclusion: In our experience, the AI may represent a valid support system for diagnosis, therapeutic decision, preoperative planning, or prediction of the outcome. However, there are still some challenges particularly in terms of ethics and data protection, to offer an augmented medicine.
KEY WORDS: Artificial Intelligence, Machine-learning, Inflammatory Biomarkers, Traumatology, Head-Neck Oncology, Oral surgery
Authors:
Umberto Committeri
Affiliation:
University of Naples Federico II, Naples, Italy
Doi: 10.54936/haoms242p99
ABSTRACT:
Objectives: The practice of surgery requires making rapid and complex decisions while managing the sometimes-uncertain health consequences for patients. In this regard, Artificial Intelligence (AI) have improved the surgeon’s southerly activity in all phases of patient management: screening, diagnosis, surgical procedure, and follow-up. It is specified how AI represents a modern technology with the ability to perform a task reserved for humans: learning, perception, reasoning, recognition. Machine learning (ML) a subdiscipline of AI, then, aims to design computer models capable of performing a task without having been explicitly programmed to do so. This type of learning is regularly employed in healthcare because of the complexity of data and high heterogeneity of patients.
The purpose of our study involved evaluating the effectiveness of ML in maxillofacial surgery.
Methods: ML capabilities were explored in several areas: trauma, oral surgery, and oncology to obtain an automated detection of maxillofacial fractures, to differentiate cystic lesions of the jaws, salivary gland tumors, and identify lymph nodal metastases in early stage of tongue cancers. The parameters analyzed regarded radiomic features extracted from MRI or CT associated then with inflammatory biomarkers obtained from blood sampling. Univariate analyses with nonparametric tests and multivariate analyses with machine learning approaches were used to validate the protocol.
Results: All clinical outcomes variables showed statistically significant differences (p < 0.05) in the Kruskal−Wallis test on median values. The multivariate analysis permitted to identify in each research field the best combinations of features and clinical parameters with a mean of 84% accuracy, 73% sensitivity, and 89% specificity.
Conclusion: In our experience, the AI may represent a valid support system for diagnosis, therapeutic decision, preoperative planning, or prediction of the outcome. However, there are still some challenges particularly in terms of ethics and data protection, to offer an augmented medicine.
KEY WORDS: Artificial Intelligence, Machine-learning, Inflammatory Biomarkers, Traumatology, Head-Neck Oncology, Oral surgery