Implementation of a neural-network-based Surgical Workflow Analysis algorithm for surgical educational environments

• C. Jiménez-Torres ^[1] ; C. Guzmán-García ^[1] ; P. Sánchez-González ^[1] ; I. Oropesa ; E.J. Gómez ^[1]
  1. [1] Universidad Politécnica de Madrid

     Universidad Politécnica de Madrid

     Madrid, España

     
• Localización: XXXVIII Congreso Anual de la Sociedad Española de Ingeniería Biomédica. CASEIB 2020: Libro de actas / Roberto Hornero Sánchez (ed. lit.), Jesús Poza Crespo (ed. lit.), Carlos Gómez Peña (ed. lit.), María García Gadañón (ed. lit.), 2020, ISBN 978-84-09-25491-0, págs. 210-214
• Idioma: inglés
• Enlaces
  • Texto Completo Libro
• Resumen

  • The automatic analysis of surgical workflow plays an important role in the modeling of surgical processes. Although current automatic approaches to recognize surgical procedures’ phases achieve accuracies varying from 0.8 and 0.9, they do not usually analyze their performance per phase, which reports relevant data on their strong and weak points. In this study, an innovative algorithm is implemented to classify the different phases of a laparoscopic cholecystectomy, following an approach based on neural networks and probabilistic methods, and analyzing the model performance for each phase. The best results were achieved using a pre-trained convolutional neural network, VGG-16, in combination with a Hidden Markov Model (accuracy = 0.91). The potential of this algorithm may further be explored for the definition of training and assessment systems of surgeons’ cognitive and decision-making skills.