Analyzing domain shift when using additional data for the MICCAI KiTS23 Challenge

TL;DR

This paper investigates how to effectively incorporate additional data in medical image segmentation tasks by addressing domain shift, demonstrating that histogram matching improves data integration over simple normalization.

Contribution

The study introduces and evaluates techniques, especially histogram matching, to mitigate domain shift when combining diverse datasets for improved model training.

Findings

Histogram matching outperforms simple normalization in reducing domain shift.

Transforming additional data improves segmentation performance.

Effective data preprocessing enhances model generalization across different data sources.

Abstract

Using additional training data is known to improve the results, especially for medical image 3D segmentation where there is a lack of training material and the model needs to generalize well from few available data. However, the new data could have been acquired using other instruments and preprocessed such its distribution is significantly different from the original training data. Therefore, we study techniques which ameliorate domain shift during training so that the additional data becomes better usable for preprocessing and training together with the original data. Our results show that transforming the additional data using histogram matching has better results than using simple normalization.