Semi-supervised deep learning for uterus and bladder segmentation on female pelvic floor magnetic resonance imaging with limited labeled data.

BACKGROUND

Accurately outlining ("segmenting") pelvic organs from magnetic resonance imaging scans is crucial for studying pelvic organ prolapse. The labor-intensive process of segmentation that identifies which pixels correspond to a particular organ in magnetic resonance imaging datasets imposes a substantial bottleneck on training artificial intelligence to do automated segmentation techniques, underscoring a need for methods that can operate effectively with minimal prelabeled data.

OBJECTIVE

The aim of this study is to introduce a novel semi-supervised learning process that uses limited data annotation in pelvic magnetic resonance imaging to improve automated segmentation. By effectively using both labeled and unlabeled magnetic resonance imaging data, our approach seeks to improve the accuracy and efficiency of pelvic organ segmentation, thereby reducing the reliance on extensive labeled datasets for artificial intelligence model training.

STUDY

DESIGN

The study used a semi-supervised deep learning framework for uterus and bladder segmentation, in which a model is trained using both a small number of expert-outlined structures and a large number of unlabeled scans, leveraging the information from the labeled data to guide the model and improve its predictions on the unlabeled data. It involved 4103 magnetic resonance images from 48 female subjects. This approach included self-supervised learning of image restoration tasks for feature extraction and pseudo-label generation, followed by combined supervised learning on labeled images and unsupervised training on unlabeled images. The method's performance was evaluated quantitatively using the Dice Similarity Coefficient, Average Surface Distance, and 95% Hausdorff Distance. For statistical analysis, 2-tailed paired t-tests were conducted for comparison.

RESULTS

This framework demonstrated the capacity to achieve segmentation accuracy comparable to traditional methods while requiring only about 60% of the typically necessary labeled data. Specifically, the semi-supervised approach achieved Dice Similarity Coefficients of 0.84±0.04, Average Surface Distances of 13.98±0.93, and 95% Hausdorff Distances of 2.15±0.40 for the uterus and 0.92±0.05, 2.51±0.83, and 2.88±0.17 for the bladder, respectively (P value <.001 for all), outperforming both the baseline supervised learning and transfer learning models. Additionally, 3-dimensional reconstructions using the semi-supervised method exhibited superior details in the visualized organs.

CONCLUSION

This study's semi-supervised learning framework wherein the full use of unlabeled data markedly reduces the necessity for extensive manual annotations, achieving high segmentation accuracy with substantially fewer labeled images that can enhance clinical evaluation and advance medical image analysis by reducing the dependency on large-scale labeled pelvic magnetic resonance imaging datasets for training.