Automatic Segmentation of Metastatic Breast Cancer Lesions on 18F-FDG PET/CT Longitudinal Acquisitions for Treatment Response Assessment

Abstract

Metastatic breast cancer patients receive lifelong medication and are regularly monitored for disease progression. The aim of this work was to (1) propose networks to segment breast cancer metastatic lesions on longitudinal whole-body PET/CT and (2) extract imaging biomarkers from the segmentations and evaluate their potential to determine treatment response. Baseline and follow-up PET/CT images of 60 patients from the EPICUREseinmeta study were used to train two deep-learning models to segment breast cancer metastatic lesions: One for baseline images and one for follow-up images. From the automatic segmentations, four imaging biomarkers were computed and evaluated: SULpeak, Total Lesion Glycolysis (TLG), PET Bone Index (PBI) and PET Liver Index (PLI). The first network obtained a mean Dice score of 0.66 on baseline acquisitions. The second network obtained a mean Dice score of 0.58 on follow-up acquisitions. SULpeak, with a 32% decrease between baseline and follow-up, was the biomarker best able to assess patients’ response (sensitivity 87%, specificity 87%), followed by TLG (43% decrease, sensitivity 73%, specificity 81%) and PBI (8% decrease, sensitivity 69%, specificity 69%). Our networks constitute promising tools for the automatic segmentation of lesions in patients with metastatic breast cancer allowing treatment response assessment with several biomarkers.

PET-Derived Radiomics and Artificial Intelligence in Breast Cancer: A Systematic Review

“…Nevertheless, manual segmentation of all lesions is time consuming in clinical practice, especially in patients with multiple metastases. For this reason, Moreau and colleagues [ 46 ] trained two deep-learning models in order to automatically segment BC metastatic lesions on the baseline and follow-up FDG PET/CT of 60 patients. The authors assessed four imaging biomarkers, i.e., SULpeak, TLG, PET Bone Index, and PET Liver Index, with SULpeak identified as the best biomarker to assess patients’ response (sensitivity 87%, specificity 87%), representing a promising tool for automatic segmentation of metastatc BC lesions.…”

“…In particular, the prediction of metastasis at distance at baseline PET imaging was performed by only one paper [ 50 ], but with encouraging results. Similarly, radiomic-assisted therapy response assessment was only explored by one study [ 46 ], which introduced an interesting automatic segmentation of BC lesions using DL. We encourage researchers to investigate the potentialities of radiomic analysis and AI also on these clinical settings of BC.…”

Systematic Review of Tumor Segmentation Strategies for Bone Metastases

“…Even though LFPN achieves a slightly lower F1-score alone than self-training, models with self-training require twice the training time than the semi-supervised approach. Some studies have also suggested pretraining the model with unlabeled data from related datasets to overcome the lack of labeled data [ 45 , 114 ].…”

“…Apiparakoon et al [ 56 ] augmented a dataset by changing the light, contrast, and brightness to ensure consistency with the physician’s process. Several other augmentation techniques have been employed, including rescaling [ 13 , 33 , 39 , 43 , 44 , 45 , 54 , 58 , 59 , 60 , 98 , 107 ], rotation [ 13 , 33 , 39 , 43 , 44 , 45 , 54 , 58 , 59 , 60 , 77 , 98 , 107 ], zooming [ 13 , 44 , 107 ], shifting intensity [ 58 , 77 ], reflecting horizontally [ 77 ], translating the image [ 39 , 43 , 77 ], cropping [ 59 ], applying elastic deformations [ 45 ], gamma augmentation [ 45 ], and flipping [ 13 , 44 , 54 , 59 , 107 ]. Da Cruz et al [ 60 ] further applied a probabilistic Gaussian blur and linear contrast filters to augment the dataset.…”

“…A limitation of the model-based approaches is potential biases in small datasets, such as studying gender-biased data [ 47 ]. Insufficient data to train the models was an issue in most of the studies in this review [ 13 , 39 , 44 , 45 , 47 , 65 , 90 , 105 , 106 , 107 ]. Lu et al, 2020 [ 66 ] adopted a strategy of transforming the image segmentation issue into a pixel-wise classification issue.…”