Pulmonary Nodule Malignancy Classification Using its Temporal Evolution with Two-Stream 3D Convolutional Neural Networks

TL;DR

This paper introduces a two-stream 3D CNN that analyzes temporal evolution of pulmonary nodules from multiple CT scans to improve malignancy classification accuracy, outperforming single-scan models.

Contribution

The novel two-stream 3D CNN effectively leverages temporal data from multiple scans, enhancing malignancy prediction over existing single-scan approaches.

Findings

Achieved 77% F1-score on test data.

Improved F1-score by 9% over single-time-point models.

Improved F1-score by 12% over single-time-point models.

Abstract

Nodule malignancy assessment is a complex, time-consuming and error-prone task. Current clinical practice requires measuring changes in size and density of the nodule at different time-points. State of the art solutions rely on 3D convolutional neural networks built on pulmonary nodules obtained from single CT scan per patient. In this work, we propose a two-stream 3D convolutional neural network that predicts malignancy by jointly analyzing two pulmonary nodule volumes from the same patient taken at different time-points. Best results achieve 77% of F1-score in test with an increment of 9% and 12% of F1-score with respect to the same network trained with images from a single time-point.