Submitted version, 1.218Mb
This file wil be publicly accessible from 2022-09-22
Request a copy
Request a copy
Rodrigues M., Mayo M., Patros P. (2021) Interpretable Deep Learning for Surgical Tool Management. In: Reyes M. et al. (eds) Interpretability of Machine Intelligence in Medical Image Computing, and Topological Data Analysis and Its Applications for Medical Data. IMIMIC 2021, TDA4MedicalData 2021. Lecture Notes in Computer Science, vol 12929. Springer, Cham. https://doi.org/10.1007/978-3-030-87444-5_1
Permanent Research Commons link: https://hdl.handle.net/10289/14577
This paper presents a novel convolutional neural network framework for multi-level classification of surgical tools. Our classifications are obtained from multiple levels of the model, and high accuracy is obtained by adjusting the depth of layers selected for predictions. Our framework enhances the interpretability of the overall predictions by providing a comprehensive set of classifications for each tool. This allows users to make rational decisions about whether to trust the model based on multiple pieces of information, and the predictions can be evaluated against each other for consistency and error-checking. The multi-level prediction framework achieves promising results on a novel surgery tool dataset and surgery knowledge base, which are important contributions of our work. This framework provides a viable solution for intelligent management of surgical tools in a hospital, potentially leading to significant cost savings and increased efficiencies.
© Springer Nature Switzerland AG 2021.This is the author's submitted version. The final publication is available at Springer via dx.doi.org/10.1007/978-3-030-87444-5_1