BACKGROUND: Dose to heart substructures is a better predictor for major adverse cardiac events (MACE) than mean heart dose (MHD). We propose an avoidance planning strategy for important cardiac substructures.
METHODS: Two plans, clinical and cardiac substructure-avoidance plan, were generated for twenty patients. Five dose-sensitive substructures, including left ventricle, pulmonary artery, left anterior descending branch, left circumflex branch and the coronary artery were chosen. The avoidance plan aims to meet the target criteria and organ-at-risk (OARs) constraints while minimizing the dose parameters of the above five substructures. The dosimetric assessments included the mean dose and the maximum dose of cardiac substructures and several volume parameters. In addition, we also evaluated the relative risk of coronary artery disease (CAD), chronic heart failure (CHF), and radiation pneumonia (RP).
RESULTS: Pearson correlation coefficient and R2 value of linear regression fitting demonstrated that MHD had poor prediction ability for the mean dose of the cardiac substructures. Compared to clinical plans, an avoidance plan is able to statistically significantly decrease the dose to key substructures. Meanwhile, the dose to OARs and the coverage of the target are comparable in the two plans. In addition, it can be observed that the avoidance plan statistically decreases the relative risks of CAD, CHF, and RP.
CONCLUSIONS: The substructure-avoidance planning strategy that incorporates the cardiac substructures into optimization process, can protect the important heart substructures, such as left ventricle, left anterior descending branch and pulmonary artery, achieving the substantive sparing of dose-sensitive cardiac structures, and have the potential to decrease the relative risks of CAD, CHF, and RP.

BACKGROUND: Multi-modal learning is widely adopted to learn the latent complementary information between different modalities in multi-modal medical image segmentation tasks. Nevertheless, the traditional multi-modal learning methods require spatially well-aligned and paired multi-modal images for supervised training, which cannot leverage unpaired multi-modal images with spatial misalignment and modality discrepancy. For training accurate multi-modal segmentation networks using easily accessible and low-cost unpaired multi-modal images in clinical practice, unpaired multi-modal learning has received comprehensive attention recently.
OBJECTIVE: Existing unpaired multi-modal learning methods usually focus on the intensity distribution gap but ignore the scale variation problem between different modalities. Besides, within existing methods, shared convolutional kernels are frequently employed to capture common patterns in all modalities, but they are typically inefficient at learning global contextual information. On the other hand, existing methods highly rely on a large number of labeled unpaired multi-modal scans for training, which ignores the practical scenario when labeled data is limited. To solve the above problems, we propose a modality-collaborative convolution and transformer hybrid network (MCTHNet) using semi-supervised learning for unpaired multi-modal segmentation with limited annotations, which not only collaboratively learns modality-specific and modality-invariant representations, but also could automatically leverage extensive unlabeled scans for improving performance.
METHODS: We make three main contributions to the proposed method. First, to alleviate the intensity distribution gap and scale variation problems across modalities, we develop a modality-specific scale-aware convolution (MSSC) module that can adaptively adjust the receptive field sizes and feature normalization parameters according to the input. Secondly, we propose a modality-invariant vision transformer (MIViT) module as the shared bottleneck layer for all modalities, which implicitly incorporates convolution-like local operations with the global processing of transformers for learning generalizable modality-invariant representations. Third, we design a multi-modal cross pseudo supervision (MCPS) method for semi-supervised learning, which enforces the consistency between the pseudo segmentation maps generated by two perturbed networks to acquire abundant annotation information from unlabeled unpaired multi-modal scans.
RESULTS: Extensive experiments are performed on two unpaired CT and MR segmentation datasets, including a cardiac substructure dataset derived from the MMWHS-2017 dataset and an abdominal multi-organ dataset consisting of the BTCV and CHAOS datasets. Experiment results show that our proposed method significantly outperforms other existing state-of-the-art methods under various labeling ratios, and achieves a comparable segmentation performance close to single-modal methods with fully labeled data by only leveraging a small portion of labeled data. Specifically, when the labeling ratio is 25%, our proposed method achieves overall mean DSC values of 78.56% and 76.18% in cardiac and abdominal segmentation, respectively, which significantly improves the average DSC value of two tasks by 12.84% compared to single-modal U-Net models.
CONCLUSIONS: Our proposed method is beneficial for reducing the annotation burden of unpaired multi-modal medical images in clinical applications.

Accurate segmentation of cardiac substructures in multi-modality heart images is an important prerequisite for the diagnosis and treatment of cardiovascular diseases. However, the segmentation of cardiac images remains a challenging task due to (1) the interference of multiple targets, (2) the imbalance of sample size. Therefore, in this paper, we propose a novel two-stage segmentation network with feature aggregation and multi-level attention mechanism (TSFM-Net) to comprehensively solve these challenges. Firstly, in order to improve the effectiveness of multi-target features, we adopt the encoder-decoder structure as the backbone segmentation framework and design a feature aggregation module (FAM) to realize the multi-level feature representation (Stage1). Secondly, because the segmentation results obtained from Stage1 are limited to the decoding of single scale feature maps, we design a multi-level attention mechanism (MLAM) to assign more attention to the multiple targets, so as to get multi-level attention maps. We fuse these attention maps and concatenate the output of Stage1 to carry out the second segmentation to get the final segmentation result (Stage2). The proposed method has better segmentation performance and balance on 2017 MM-WHS multi-modality whole heart images than the state-of-the-art methods, which demonstrates the feasibility of TSFM-Net for accurate segmentation of heart images.

We investigated to what extent can the dose-volumes of the coronary artery and the cardiac substructures be reduced by using IMRT technique in left-sided breast cancer patients. We chose 40 pN2M0 patients treated with postmastectomy IMRT. The original treatment plans were retrieved and the (internal mammary nodes) IMNs and cardiac substructure delineations were added. Three sets of dose-volume parameters including the original plans without internal mammary irradiation (IMNI), the plans with IMNI, and the plans with dose constraints to the heart, were derived. In left-sided patients, when IMNI was included, the V30 for right ventricle (RV), left ventricle (LV), pulmonic valve (PV), and left anterior descending artery (LADA) were 56.37% ± 7.9%, 25.3% ± 7.3%, 48.3% ± 6.3%, and 69.7% ± 6.4%, respectively. Of the 4 main coronary arteries, LADA had the highest dose followed by the left main coronary artery (LMCA). LADA had a V40 of 62% ± 9.7% vs 13.5% ± 3.5%, and a V50 of 27.5% ± 4.7% vs 0, with and without IMNI. For the right-sided patients, the V30s for all the heart substructures were 0 with or without IMNI. When we set a dose constraint of V40 < 10% for the LADA in the left-sided patients, the PTV volumes covered by 50 Gy decreased by only 1%. IMNI increased the V30 of the right and left ventricle and significantly increased the V40 and V50 to the LADA of left-sided breast cancer patients. IMRT markedly reduces the dose to the main coronary arteries and the right and left ventricle.