Study Results from Visvesvaraya National Institute of Technology (VNIT) Provide New Insights into Machine Learning (Comparative Analysis of Machine Learning Frameworks for Automatic Polyp Characterization).

A study conducted by researchers at Visvesvaraya National Institute of Technology (VNIT) in Nagpur, India, explores the use of machine learning algorithms for the automatic characterization of polyps during colonoscopy procedures. The study compares the performance of two machine learning frameworks, training from scratch and transfer learning, using three different architectures. The results show that the training from scratch framework outperformed the transfer learning framework in terms of accuracy, AUC, and NPV. This study is the first to focus solely on polyp classification using Narrow Band Imaging (NBI) polyp images. [Extracted from the article]