Alleviating Hyperparameter-Tuning Burden in SVM Classifiers for Pulmonary Nodules Diagnosis with Multi-Task Bayesian Optimization

TL;DR

This study examines the feasibility of employing multi-task Bayesian optimization to accelerate the hyperparameters search for classifying benign and malignant pulmonary nodules using RBF SVM and suggests that multi-task Bayesian optimization significantly accelerates the search for hyperparameters in comparison to a single-task approach.

Abstract

In the field of non-invasive medical imaging, radiomic features are utilized to measure tumor characteristics. However, these features can be affected by the techniques used to discretize the images, ultimately impacting the accuracy of diagnosis. To investigate the influence of various image discretization methods on diagnosis, it is common practice to evaluate multiple discretization strategies individually. This approach often leads to redundant and time-consuming tasks such as training predictive models and fine-tuning hyperparameters separately. This study examines the feasibility of employing multi-task Bayesian optimization to accelerate the hyperparameters search for classifying benign and malignant pulmonary nodules using RBF SVM. Our findings suggest that multi-task Bayesian optimization significantly accelerates the search for hyperparameters in comparison to a single-task approach. To the best of our knowledge, this is the first investigation to utilize multi-task Bayesian optimization in a critical medical context.

Figures (4)

• Figure 1: Workflow of the proposed MTBO based SVMs hyperparameter-tuning method on pulmonary nodule classification. The collected whole body CT scans are first preprocessed to delineate the pulmonary nodules. Then, the segmented nodules are discretized via multiple strategies. After that, for the result of each discretization strategy, various types of features are extracted as the input to a specific SVM. Finally, a multi-task Gaussian process is established through the correlation between tasks, the query point is selected by maximizing the acquisition function, and the optimal hyperparameters are determined by iterative query.
• Figure 2: Segmentation results under different views. Left: axial; top right: coronal; bottom right: sagittal.
• Figure 3: Classification loss obtained by 10-fold cross-validation using STBO and MTBO respectively.
• Figure 4: The landscape for loss function on hyperparameter space. (N = 64, mean ± 3SD)