Improving Cancer Imaging Diagnosis with Bayesian Networks and Deep Learning: A Bayesian Deep Learning Approach

TL;DR

This work addresses cancer imaging diagnosis under uncertainty by proposing Bayesian Deep Learning (BDL) that merges Deep Learning with Bayesian Networks. It surveys three methodologies—SWA-Gaussian, Deep Ensemble, and Bayesian Neural Networks—to infuse uncertainty estimation into DL models. The analysis highlights that BDL can improve accuracy and provide informative uncertainty, with examples like ARA-CNN maintaining performance under data uncertainty and mislabeled inputs, and reducing reliance on purely deterministic predictions. The study suggests that BDL holds practical potential for clinical decision support in oncology imaging, while acknowledging the need for optimized integration strategies and model selection.

Abstract

With recent advancements in the development of artificial intelligence applications using theories and algorithms in machine learning, many accurate models can be created to train and predict on given datasets. With the realization of the importance of imaging interpretation in cancer diagnosis, this article aims to investigate the theory behind Deep Learning and Bayesian Network prediction models. Based on the advantages and drawbacks of each model, different approaches will be used to construct a Bayesian Deep Learning Model, combining the strengths while minimizing the weaknesses. Finally, the applications and accuracy of the resulting Bayesian Deep Learning approach in the health industry in classifying images will be analyzed.

Figures (5)

• Figure 1: Structure of Deep Learning Model.
• Figure 2: Sample Bayesian Network.
• Figure 3: Algorithmic implementation of SWAG.
• Figure 4: Algorithmic implementation of Deep Ensemble.
• Figure 5: Accuracy of ARA-CNN with mislabeled images.