Multimodal PET/CT Tumour Segmentation and Prediction of Progression-Free Survival using a Full-Scale UNet with Attention

Abstract

Segmentation of head and neck (H&N) tumours and prediction of patient outcome are crucial for patient's disease diagnosis and treatment monitoring. Current developments of robust deep learning models are hindered by the lack of large multi-centre, multi-modal data with quality annotations. The MICCAI 2021 HEad and neCK TumOR (HECKTOR) segmentation and outcome prediction challenge creates a platform for comparing segmentation methods of the primary gross target volume on fluorodeoxyglucose (FDG)-PET and Computed Tomography images and prediction of progression-free survival in H&N oropharyngeal cancer. For the segmentation task, we proposed a new network based on an encoder-decoder architecture with full inter-and intra-skip connections to take advantage of low-level and high-level semantics at full scales. Additionally, we used Conditional Random Fields as a post-processing step to refine the predicted segmentation maps. We trained multiple neural networks for tumor volume segmentation, and these segmentations were ensembled achieving an average Dice Similarity Coefficient of 0.75 in cross-validation, and 0.76 on the challenge testing data set. For prediction of patient progression free survival task, we propose a Cox proportional hazard regression combining clinical, radiomic, and deep learning features. Our survival prediction model achieved a concordance index of 0.82 in cross-validation, and 0.62 on the challenge testing data set.

Effective Approaches to Fetal Brain Segmentation in MRI and Gestational Age Estimation by Utilizing a Multiview Deep Inception Residual Network and Radiomics

“…The Clinical + DeepFeat + Radiomics combination led to a C-index of 0.786, which was lower than the Deep-Features + Radiomics combination, meaning that we could achieve accurate predictions of patients’ survival days without employing clinical data. Table 11 also demonstrated that the features based on radiomics and DeepFeat achieved the highest C-index scores (0.821) compared to those of methods that were specially designed for the prediction of patients’ survival days, such as [ 38 , 39 , 40 , 41 ].…”

Deep Learning Based Cancer Survival Analysis Algorithm

“…Finally, 1505 regions of interest were left as inputs of model. In experiments, two SOTA methods were selected as benchmark methods, which are derived from reference [6] and [7] respectively. In reference [6], SqueezeNet was used to process WSI data in survival analysis task.…”

“…In reference [6], SqueezeNet was used to process WSI data in survival analysis task. In reference [7], a survival analysis method based on U-Net was proposed. Both approaches achieve SOTA performance on the corresponding data set.…”

“…Anika Cheerla et al [6] used SqueezeNet to process WSI data. Bourigault et al [7] proposed a survival analysis method based on U-Net network. Limited by network depth, these methods cannot extract the deeper features of histological images.…”