e12511 Background: Left-sided breast cancer radiotherapy poses a significant challenge in balancing target coverage with the sparing of critical cardiac substructures. RapidArc Dynamic (RAD) is a novel hybrid technique that integrates the continuous arc delivery of volumetric modulated arc therapy (VMAT) with the fixed-beam modulation of intensity modulated radiotherapy (IMRT) through strategic gantry pauses and dynamic collimator rotation. This study aims to evaluate the dosimetric performance of RAD, benchmarking it against IMRT and VMAT, with a specific focus on cardiac substructure sparing and normal tissue complication probability (NTCP). Methods: Twenty patients with left-sided breast cancer treated with post-mastectomy adjuvant radiotherapy were retrospectively enrolled. For each patient, RAD (2 partial arcs with 6 gantry pauses), IMRT (9 fields), and VMAT (5 partial arcs) plans were generated with a prescription of 50 Gy in 25 fractions. Dosimetric metrics, plan complexity, and conformity index were analyzed, and NTCP values were calculated using the Lyman-Kutcher-Burman (LKB) model for multiple endpoints. Results: RAD demonstrated significantly lower (P < 0.05) doses to the heart, left lung, and spinal cord compared to both IMRT and VMAT. Specifically, RAD achieved superior sparing of the left anterior descending coronary artery (LAD) and left ventricle (LV), reducing mean and maximum doses significantly compared to competing modalities. While VMAT resulted in a higher low-dose bath to the contralateral breast, RAD successfully mitigated this issue, delivering significantly lower contralateral doses. Furthermore, RAD yielded significantly lower predicted NTCPs for cardiac perfusion defects, LA major adverse cardiac events, and pulmonary pneumonitis compared to IMRT and VMAT. Conclusions: RAD offers a feasible and effective alternative for left-sided chest radiotherapy, providing an optimal balance between high dose conformity and superior ipsilateral organ sparing. By significantly reducing doses to critical cardiac substructures and minimizing predicted complications, RAD may reduce the risk of late toxicity in breast cancer survivors.

Contrast-enhanced (CE) T2-fluid-attenuated inversion recovery (T2-FLAIR) and CE T1-weighted imaging(T1WI) are used in the routine clinic to assist radiation oncologists for accurate gross tumor target volume (GTV) delineation of brain metastases (BMs). In our previous study, CE T1WI should be performed ≥ 10 min after contrast agent injection when determining the GTV of large-volume BMs. Furthermore, we demonstrated that a combination of enhancement scans with different delay times is necessary to accurately assess the boundaries of the GTV. However, this approach requires a significant amount of time. Therefore, this prospective study aimed to analyze the imaging effect of delayed CE T2-FLAIR on large-volume BMs and investigate its feasibility for GTV delineation in radiotherapy. A total of 76 patients with BMs (184 lesions) were prospectively enrolled. All patients underwent magnetic resonance (MR) simulation scans. CE T1WI scans were performed 1, 3, 5, 10, 18, and 20 min after gadolinium-based contrast agent injection. CE T2-FLAIR was performed after the 10-min delayed CE T1WI. The BMs GTVs were determined on various sequence images and defined as regions of interest (ROIs), such as ROI-1 min, ROI-3 min… ROI-20 min, and ROI-T2-FLAIR. ROI-1 min, ROI-3 min, ROI-5 min and ROI-T2-FLAIR were merged as the fusion ROI. The fusion of all CE T1WI ROIs (1, 3, 5, 10, 18, and 20 min) was defined as ROI-total, and used as the reference. The signal intensity, volume, and shape of the ROIs were compared. ROI-T2-FLAIR had the highest contrast ratio (0.77 ± 0.39), which was 148.4%, 126.5%, and 126.5% higher than that of ROI-1 min, ROI-3min, and ROI-5 min, respectively. The ROI-T2-FLAIR volume increased by 11.5%, 8.6%, and 6.6% compared to ROI-1 min, ROI-3min, and ROI-5 min volumes, while decreasing by 17.44% compared to ROI-total(P < 0.05). Compared with ROI-total, all fusion ROIs all had volume differences < 0.08 cm3 and increase rates < 5%. Compared with ROI-total, ROI-T2-FLAIRhad the smallest Dice similarity coefficient (DSC); whereas the DSCs of the fusion ROIs were 0.903 to 0.917. The Hausdorff Distance of each ROI was < 3 mm. The combined use of delayed CE T2-FLAIR and CE T1WI may improve visualization of brain metastases and provide additional information to support tumor target delineation.

To further explore relative biological effectiveness (RBE) variability, the RBE of different intracerebral cells at various irradiation (IR) dosages and time were determined in this study. A total of 120 rabbits were randomly divided into proton groups (0, 10, 20, 30, 40 Gy, RBE) (n = 3) and photon groups (0, 10, 20, 30, 40 Gy) (n = 3). The rabbits were sacrificed at 2, 4, 6, 8 weeks after brain IR. Neuronal survival, identified via Hematoxylin and Eosin (H&E) staining, and immunohistochemical detection of neurofilament (NF), Olig2, and CD68 in the hippocampus and thalamus, were analyzed. Dose- and time-dependent RBE curves were fitted using the LQ model. Proton IR showed higher neuronal survival at 4-, 6-, and 8-weeks post 10 Gy, 20 Gy, 30 Gy IR (p < 0.05) compared to photon IR. Oligodendrocyte populations in photon group at 4-, 6-, and 8-weeks post 10 Gy IR and 6-, 8-weeks post 20 Gy were consistently higher than proton subgroups (p < 0.05). While proton IR showed higher microglial activation which was observed only at 4-weeks post 30y IR. Proton RBE for neurons and oligodendrocytes remained below 1.1 but exceeded 1.1 for microglial activation. These findings demonstrate the dose- and time- dependent nature of proton RBE and suggest brain tissue tolerates higher proton IR doses compared to photon IR, which fully confirmed the biological advantages of proton IR. These will help clinicians more precisely set the organ limit at risk and tailor radiotherapy plans.

Background Left-sided breast cancer radiotherapy poses a challenge in balancing target coverage with the sparing of critical cardiac substructures. A novel hybrid technique - RapidArc Dynamic (RAD) - integrated continuous arc delivery of volumetric modulated arc therapy (VMAT) with fixed-beam modulation of intensity modulated radiotherapy (IMRT) through strategic gantry pauses and dynamic collimator rotation. Purpose This study aims to evaluate the dosimetric performance of RAD, benchmarking it against IMRT and VMAT, with a specific focus on cardiac substructure sparing and normal tissue complication probability (NTCP). Methods Twenty patients with left-sided breast cancer treated with post-mastectomy adjuvant radiotherapy were retrospectively enrolled. For each patient, three plans (RAD, IMRT, and VMAT) were generated with a prescription of 50 Gy in 25 fractions. RAD plans utilized a “balanced” optimization strategy combining static ports and dynamic arcs with strategic gantry pauses to enhance modulation. Dosimetric metrics were analyzed for the planning target volume (PTV) and organs at risk (OARs). NTCP values were calculated using the Lyman-Kutcher-Burman (LKB) model for multiple endpoints. Results RAD demonstrated significantly lower dose metrics for the heart, left lung, and spinal cord compared to both IMRT and VMAT (P < 0.05). Specifically, RAD achieved superior sparing of the left anterior descending coronary artery (LAD) and left ventricle (LV), reducing mean and maximum doses significantly compared to competing modalities. While VMAT provided better conformity than IMRT, it resulted in a higher low-dose bath to the contralateral breast. Radiobiological modelling demonstrated that RAD successfully mitigated this issue, delivering significantly lower contralateral doses than VMAT. RAD yielded significantly lower predicted NTCPs for cardiac perfusion defects, LA major adverse cardiac events, and pulmonary pneumonitis compared to IMRT and VMAT. Conclusion RAD offers a feasible and effective alternative for left-sided chest radiotherapy, providing an optimal balance between high dose conformity and superior organ sparing.

Automatic brain lesion segmentation enhances diagnostic efficiency by enabling detailed texture analysis and precise delineation of tumor subregions. Multimodal MRI has improved segmentation accuracy by combining complementary information from different modalities. Conventional methods either fuse all modalities uniformly, obscuring how individual modalities contribute to specific segmentation subtasks, or predefine modality-to-subregion mappings based on prior medical knowledge. The former limits interpretability on modality contribution during training, while the latter relies on parameter-heavy architectures like cascaded subnetworks, making models struggle to adapt to varying modalities. To address these challenges, this paper proposes SwitchNet, a novel model that integrates interpretability into the training process and optimizes parameter efficiency without relying on predefined modality selection. First, we propose Adaptive Encoder and Decoder Blocks employing dynamic switching mechanisms to efficiently allocate feature space and prioritize critical subtasks. These blocks enable the model to automatically identify and utilize the most informative modalities. By strategically allocating parameters to modalities, our model optimizes overall parameter complexity while maintaining strong performance. Second, we propose a Guide-Contribution Mechanism to provide interpretability during training by quantitatively revealing the contributions of individual modalities to the segmentation process. This mechanism offers valuable insights into how the model delineates tumor subregions. SwitchNet was validated on three benchmark datasets, including BraTS 2023, ISLES 2022, and UCSF-PDGM, achieving competitive segmentation performance while significantly enhancing interpretability and maintaining parameter efficiency without extra cost. These results highlight its potential for efficient tumor segmentation and clinical explainability.

Background:Pancreatic cancer is characterized by an insidious onset and rapid progression, and the accurate determination of the gross tumor volume (GTV) constitutes a critical prerequisite for ensuring the efficacy of radiotherapy. Multiphase contrast-enhanced magnetic resonance imaging (CE-MRI) enables the dynamic visualization of tumor hemodynamic perfusion characteristics; however, the tissue discrimination capability varies substantially across different enhancement phases. To date, there is no universal consensus on the optimal imaging phase for GTV determination in pancreatic cancer. This study aimed to quantitatively analyze the differences in imaging and GTV determination of pancreatic cancer using multiphase CE-MRI, thereby providing a basis for selecting the optimal phase for GTV determination. Methods:Thirty patients with advanced pancreatic cancer [American Joint Committee on Cancer (AJCC) stage III-IV] who underwent magnetic resonance (MR) simulation were retrospectively enrolled in this study. MR T1-weighted images (T1WI) and contrast-enhanced T1-weighted images (CE-T1WI) were obtained at 15 s, 45 s, 75 s, 150 s, and >20 min after contrast injection. The GTV was determined from these different image sequences and named GTV-T1WI, GTV-15 s, GTV-45 s, GTV-75 s, GTV-150 s, and GTV-20 min. Differences in mean signal intensity (SI), SI contrast, volume, and shape among the different GTVs were compared. Normal pancreatic tissue was defined as a 1 cm3 region of interest of the parenchyma, strictly excluding blood vessels and pancreatic ducts. Results:The mean SI of the GTV was lower than that of the normal pancreatic tissue at each time phase (P<0.05), ranging from 9.93% to 45.01%. At CE-T1WI-15 s, the SI contrast between GTV and normal pancreatic tissue was the highest at 0.45±0.10, significantly superior to the T1WI (0.34±0.13, P<0.001). The GTV-15 s volume was 21.02±12.43 cm3. Compared with CE-T1WI-15 s, the SI contrast between the GTV and normal pancreatic tissue on T1WI and CE-T1WI-45 s to CE-T1WI-20 min decreased by 22.42-77.43% (P<0.05). Compared with GTV-15 s, the volume of GTV-T1WI and GTV-45 s-GTV-20 min decreased by -14.10-22.75%. Except for GTV-15 s and GTV-45 s, GTV-15 s and GTV-75 s, and GTV-45 s and GTV-75 s, the differences in GTV volumes in the other phases were statistically significant (P<0.05). The shape change trend of GTV at different phases was consistent with the volume compared with that of GTV-15 s. The Dice similarity coefficients (DSCs) of GTV-T1WI, GTV-45 s, GTV-75 s, GTV-150 s, and GTV-20 min were 0.74±0.10, 0.79±0.11, 0.76±0.13, 0.72±0.15, and 0.64±0.13, respectively. Conclusions:The CE-T1WI-15 s sequence demonstrated significant improvements in SI contrast and boundary definition. Consequently, it holds significant potential as an optimal sequence for GTV determination in pancreatic cancer radiotherapy, warranting further validation in larger cohorts.

PurposeThe aim of this study was to evaluate the feasibility of gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid (Gd-EOB-DTPA) contrast-enhanced magnetic resonance imaging (CE-MRI) for determining the gross tumor volume (GTV) of hepatocellular carcinoma (HCC).MethodsA retrospective analysis was conducted on 12 patients diagnosed with HCC (18 lesions) who received radiotherapy and underwent magnetic resonance (MR) simulation. Six series images, including MR T1-weighted image (T1WI) and contrast-enhanced T1WI (CE-T1WI) at 15 s, 45 s, 75 s, 150 s, and >20 min after Gd-EOB-DTPA injection, were obtained, and the GTV was determined in the different temporal images. The differences in mean signal intensity (SI), SI contrast between the HCC and liver tissue, volume and shape of HCC GTV among different phases were compared.Results(1) The mean SI of liver tissue reached its peak enhancement at >20 min, showing a 140.90 ± 64.69% increase, compared with T1WI (p < 0.05). (2) Compared with CE-T1WI-20min, the mean SI of the HCC increased by -41.19~18.09% from T1WI, CE-T1WI-15s to CE-T1WI-150s. Conversely, the mean SI of liver tissue decreased by 5.27~55.87% over the same period. Consequently, the SI contrast between HCC and liver tissue decreased by 53.30~89.37%. (3) The maximum GTV volume determined by CE-T1WI-20min was (22.80 ± 18.57) cm3, coinciding with the highest value of SI contrast (0.29 ± 0.16). (4) Compared with GTV-20min, GTV-T1WI and GTV-15s~GTV-150s had volume reductions of 6.73~19.35%. (5) Compared with GTV-20min, the Dice similarity coefficients (DSC) of GTV-T1WI and GTV-15s~GTV-150s ranged from 0.745 to 0.819. Additionally, the shape change trend of GTV in the CE-T1WI images was generally consistent with the volume change trend.ConclusionCE-T1WI MR images acquired more than 20 min post-injection of Gd-EOB-DTPA exhibited significant advantages in determining the GTV boundaries and enhancing the contrast of SI between HCC and liver tissue. The CE-T1WI-20min sequence is recommended for determining HCC GTV.

Accurately segmenting the pancreas from computed tomography (CT) images is critical for the early diagnosis and timely treatment of pancreatic cancer. However, segmentation tasks encounter substantial challenges due to the pancreas's complex shape, noise interference from surrounding tissues, and the variability in its appearance across different cases. To address these issues, we propose a diverse kernel mutual adaptive learning (DKMAL) framework to effectively capture detailed and overall structural features of the pancreas. Specifically, DKMAL employs denoising-aware multi-kernel feature aggregation (DMFA) module to effectively enhance informative representations while suppressing noise redundancy. Adaptive attention contrastive learning (AACL) block is developed to facilitate adaptive similarity learning between features extracted by diverse branches through the integration of beneficial noise. Furthermore, boundary sensitive region optimization (BSRO) strategy is designed to improve segmentation accuracy by balancing the focus between the pancreas's core regions and detailed boundaries during training. Evaluations on the NIH Pancreas dataset demonstrate that the DKMAL framework outperforms ten state-of-the-art algorithms across multiple evaluation metrics. To further validate the applicability of the DKMAL framework, we extend our experiments on the MSD Pancreas dataset for pancreas segmentation, thereby providing additional evidence of its robustness.

Current neural-like P systems use "point neurons" as the computing entities, and the computations in these neurons are simplified, ignoring the fact that, in organisms, subcellular compartments (such as neuronal dendrites) can also perform operations as independent computing units in addition to computing at the individual neuron level. The nervous system has a strong ability for optimization learning. Therefore, we propose learnable dendrite neural P (LDNP) systems with new plasticity rules, in which the dendrite structure and learning function can be adaptively changed when solving different application problems. Specifically, the dendrites of neurons are designed as dendritic trees composed of multiple dendritic branches, each of which serves as an independent computing unit. The multilevel complex topological structure of dendrites provides powerful computing capabilities for neurons. A model for predicting the overall survival of glioblastoma (GBM) patients was developed based on LDNP systems and validated on the GBM cohort from the Cancer Genome Atlas. Compared with thirteen state-of-the-art methods, the LDNP system achieves the best performance.

Accurate segmentation of multiple brain metastases (MBMs) is crucial for stereotactic radiosurgery (SRS) treatment. Semi-supervised learning performs well under the scarcity of labeled medical images in clinic. However, distribution discrepancy (various positions, numbers and sizes) of MBMs inner/inter patients, and ambiguous boundaries of MBMs hinder its effective feature distillation learning, leading to poor results. We propose a novel neuropeptide-regulated P (NRP) system with three dynamic neuropeptide rules to address the above issues. Specifically, neuropeptide-driven dynamic smoothing mechanism is designed to segment MBMs with rare labels interactively under a parallel ensemble mode. To mitigate those mis-segmentation caused by distribution discrepancy, neuropeptide anatomical masks are generated by rules to do phenotype transfer fusion among samples. To further suppress error accumulation during training, neuropeptide bias masks are also obtained to correct features in biased regions continuously. Experiments on a private MBMs dataset and the public PROMISE12 dataset show the NRP system outperforms ten state-of-the-art approaches, demonstrating the effectiveness and robustness of the NRP system across different medical MRI tasks.

Purpose: To compare the imaging manifestations of BM at 3-min and > 60-min delayed-enhanced MRI and explore their imaging characteristics and changing patterns in ultra-long delayed-enhanced MRI > 60 min. Methods: Twenty-six participants with BM were prospectively enrolled from May to October 2019. Contrast-enhanced (CE) T1-weighted imaging (T1WI) was performed for 3 min and > 60 min after contrast injection. BM were defined as regions of interest (ROI-3min and ROI-60min). The two ROIs were fused (ROI-total); the additional display area of the lesion at 3-min (ROI-A) and > 60-min (ROI-B) was determined. Signal intensity (SI), volume, and shape differences between ROIs and brain white matter were compared; the imaging characteristics of BM on ultra-long delayed-enhanced MRI were analyzed. Results: 12 males (age: 37–72 [median: 62] years) and 14 females (age: 41–79 [median: 61] years) were included. BM were divided into disappeared (n = 79, 38.0 %) and non-disappeared (n = 129, 62.0 %) groups. The average volume of the two groups at 3-min was 0.18 ± 0.25 and 3.53 ± 10.47 cm3 (p < .05). At > 60-min, BM in the non-disappeared group had four imaging manifestations: adduction (95.3 %), abduction (72.7 %), signal reversal (10.1 %), and filling (38 %) effects. In the non-disappeared group, ROI-3min, ROI-60min, ROI-total, ROI-A, and ROI-B volumes averaged 3.53 ± 10.47, 3.72 ± 11.51, 4.06 ± 11.76, 0.36 ± 0.54, and 0.54 ± 1.48 cm3 (p < .05). The DSC obtained from ROI-3min compared to ROI-60min was largest in the > 5 cm3 group. Conclusion: BM changed significantly in > 60-min delayed CE T1WI; no consistent regularity was observed.

Automated segmentation for variable pancreas and tumors is challenging due to 1) large variations occurring within/crossing the narrow and irregular pancreas, 2) a small proportion of voxels on pancreatic tumors in the abdomen and 3) ambiguous lesion boundaries caused by low contrast between target surroundings. Most existing methods mainly focus on extracting local semantic features corresponding to ground truth, while ignoring the mislead by voxels similar to boundaries and irrelevant information from input images, and the lack of shallow global information. In this paper, we present a novel deep network, MF2N network, equipped with three feature fusion modules for automated segmentation of pancreas and tumors. We first propose adaptive morphological feature fusion (AMF2) module to dynamically learn morphological features of targets from skeleton to boundaries for mitigating undersegmentation. Then, bidirectional semantic feature fusion (BSF2) module is proposed to optimize mutual information between prediction and manual delineation as well as discard redundant information between features to capture more consistent feature expressions and alleviate noise interference. Furthermore, we develop a local-global dependency feature fusion (LGDF2) module to fuse local features from CNNs and global information provided by shallow features through lightweight transformers to enhance MF2N's capability to grasp more boundary and content features. The experimental results show that MF2N network outperforms twelve state-of-the-art methods on the MSD pancreas dataset and can be applied to other potential clinical scenarios.

PURPOSE:Multisequence magnetic resonance imaging (MRI) radiomic features were used to analyze dynamic changes in the hippocampus after whole-brain radiotherapy (WBRT), thus providing an objective basis for the early prediction of hippocampal radiation injury. METHODS:Seventy-five patients with brain metastases (BMs) who received WBRT underwent MRI scanning (including T1-weighted imaging [T1WI], contrast-enhanced [CE]-T1WI, T2-weighted imaging [T2WI], T2-weighted Fluid-Attenuated Inversion Recovery imaging [T2 FLAIR] and diffusion weighted imaging [DWI]) before WBRT (MRIpre), after WBRT (MRIpost, 26.22 ± 13.05 days after the MRIpre scan), and at follow-up WBRT (MRIfollow, 393.45 ± 210.33 days after the MRIpost scan). Radiomics features were subsequently extracted from delineations of the hippocampus on the different sequences. Changes in the hippocampal volume and radiomics features of the sequences were analyzed in the MRIpost and MRIfollow sequences relative to the MRIpre sequences. The features were then organized as follows: (1) Group1 features included those features that were significantly different among MRIpre, MRIpost, and MRIfollow scans; and (2) Group2 features included those features that were significantly different between MRIpre and MRIfollow scans and between MRIpost and MRIfollow scans. RESULTS:(1) The average MRIpost and MRIfollow hippocampal volumes were 3.32 ± 0.49 cm3 and 2.95±0.45 cm3, respectively, which were 1.68 % and 12.51% lower than the MRIpre volume (3.41 ± 0.49 cm3), respectively (p < 0.05). (2) Radiomics analysis revealed that 88 features were significantly different (p < 0.05) across the MRIpre, MRIpost, and MRIfollow scans. The T2WI sequence contained the greatest number of significant features (n = 42). Among Group1 features (n = 57), enrichment was observed in T2WI (n = 34) and T1WI (n = 22). The feature exhibiting the highest rate of change was GLCM-ClusterShade (range: 83.87-281.62 %). All 12 significant change features in CE-T1WI were observed in Group2. Although the overall timing difference for T2 FLAIR was not significant (p = 0.064), DWI contained a single Group2 feature (p = 0.032). Within Group2, GLCM-ClusterTendency exhibited the largest rate of change (range: 37.16-51.27 %). CONCLUSIONS:Compared with volume, multisequence MRI radiomics features more directly reflect dynamic microscopic hippocampal changes across MRIpre, MRIpost, and MRIfollow time points. T2WI and T1WI captured early sustained radiomics alterations, whereas CE-T1WI reflected delayed changes, thus serving as potential biomarkers for the monitoring of hippocampal dynamics following WBRT.

Detection of various lesions in brain MRI is clinically critical, but challenging due to the diversity of lesions and variability in imaging conditions. Current unsupervised learning methods detect anomalies mainly through reconstructing abnormal images into pseudo-healthy images (PHIs) by normal samples learning and then analyzing differences between images. However, these unsupervised models face two significant limitations: restricted generalizability to multi-modality and multi-center MRIs due to their reliance on the specific imaging information in normal training data, and constrained performance due to abnormal residuals propagated from input images to reconstructed PHIs. To address these limitations, two novel modules are proposed, forming a new PHI reconstruction framework. Firstly, the disentangled representation module is proposed to improve generalizability by decoupling brain MRI into imaging information and essential imaging-invariant anatomical images, ensuring that the reconstruction focuses on the anatomy. Specifically, brain anatomical priors and a differentiable one-hot encoding operator are introduced to constrain the disentanglement results and enhance the disentanglement stability. Secondly, the edge-to-image restoration module is designed to reconstruct high-quality PHIs by restoring the anatomical representation from the high-frequency edge information of anatomical images, and then recoupling the disentangled imaging information. This module not only suppresses abnormal residuals in PHI by reducing abnormal pixels input through edge-only input, but also effectively reconstructs normal regions using the preserved structural details in the edges. Evaluated on nine public datasets (4,443 patients’ MRIs from multiple centers), our method outperforms 17 state-of-the-art methods, achieving absolute improvements of +18.32% in average precision and +13.64% in Dice similarity coefficient.

PURPOSE:The distribution analysis of the morphologic characteristics and spatial relations among brain metastases (BMs) to guide screening and early diagnosis. MATERIAL AND METHODS:This retrospective study analysed 4314 BMs across 30 brain regions from MRIs of 304 patients. This paper proposed a unified analysis model based on persistent homology (PH) and graph modelling to provide a comprehensive portrait of BMs distribution. Spatial relationships are quantified through dynamic multiple-scale graphs constructed with Rips filtration. The multi-scale centrality importance and clustering coefficients are extracted to decode BMs spatial relations. Morphologic BMs characteristics are further analysed by varying radius and volume values that are considered as clinically influential factors. Finally, two-tailed proportional hypothesis testing is used for BM statistical distribution analysis. RESULTS:For spatial analysis, results have shown a statistical increase in the proportions of high-level centrality BMs at the left cerebellum (p<0.01). BMs rapidly form graphs with high clustering rather than those with high centrality. For demographic analysis, the cerebellum and frontal are the top high-frequency areas of BMs with 0-4 and 5-10 radii. Statistical increases in the proportions of BMs at cerebellum (p<0.01). CONCLUSION:Results indicate that distributions of both BMs spatial relations and demographics are statistically non-random. This research offers novel insights into the BMs distribution analysis, providing physicians with the BMs demographic to guide screening and early diagnosis.

Purpose: This study developed a hippocampal segmentation model that can be used by clinicians by applying the Dual Path Networks U-Net (DPNU-Net), Mask-Region Convolution Neural Networks (Mask-RCNN), and No New U-Net (nnU-Net) algorithms for segmenting the hippocampus to provide a reference for the accurate implementation of hippocampal-avoidance whole-brain radiotherapy (HA-WBRT). Methods: We retrospectively collected T1-weighted imaging (T1WI) and contrast-enhanced (CE)- T1WI sequence magnetic resonance images of 312 patients with brain metastases, of which 62 served as the test set and 250 as the training set. Manual segmentation was used as the gold standard to compare the differences in the dice similarity coefficient (DSC), relative volume error (RVE), 95% hausdorff distance (95%HD) and average surface distance (ASD) of the DPNU-Net, Mask-RCNN, and nnU-Net models for segmenting the hippocampus in different images. Results: (1) Compared with manual segmentation, the DPNU-Net, Mask-RCNN, and nnU-Net models segmented the hippocampus based on T1WI with DSCs of 0.819-0.897, RVEs of-3.40% to 1.40%, 95%HDs of 0.813-37.425 mm, ASDs of 0.155-3.907 mm. The best segmentation effect was achieved by using the DPNU-Net model. (2) Based on the CE-T1WI, the DSCs of the DPNU-Net, Mask-RCNN, and nnU-Net models compared with manual segmentation were 0.791-0.879, whereas the RVEs were-1.90% to 2.20%, 95%HDs were 0.915-47.812 mm, and the ASDs were 0.210-5.384 mm. The segmentation effect of the DPNU-Net model was the best. (3) When comparing the DPNU-Net, Mask-RCNN, and nnU-Net models for T1WI and CE-T1WI segmentation, the differences in the DSC, 95%HD, ASD and volumes of the hippocampi segmented by the DPNU-Net model between the two sequences of images were the smallest. Conclusions: Considering automatic hippocampal segmentation, the DPNU-Net model had a higher accuracy and more stable performance than the Mask-RCNN and nnU-Net models. Thus the DPNU-Net model can be used as a practical method for automatic hippocampal segmentation in the HA-WBRT technique.

Whole-brain radiotherapy (WBRT) plays an irreplaceable role in the treatment of brain metastases (BMs), but cognitive decline after WBRT seriously affects patients’ quality of life. The development of cognitive dysfunction is closely related to hippocampal injury, but standardized criteria for predicting hippocampal injury and dose limits for hippocampal protection have not yet been developed. This review systematically reviews the clinical efficacy of hippocampal avoidance - WBRT (HA-WBRT), the controversy over dose limits, common methods and characteristics of hippocampal imaging and segmentation, differences in hippocampal protection by common radiotherapy (RT) techniques, and the application of artificial intelligence (AI) and radiomic techniques for hippocampal protection. In the future, the application of new techniques and methods can improve the consistency of hippocampal dose limit determination and the prediction of the occurrence of cognitive dysfunction in WBRT patients, avoiding the occurrence of cognitive dysfunction in patients and thus benefiting more patients with BMs.

AbstractThe study aimed to determine the specific relative biological effectiveness (RBE) of various cells in the hippocampus following proton irradiation. Sixty Sprague–Dawley rats were randomly allocated to 5 groups receiving 20 or 30 Gy of proton or photon irradiation. Pathomorphological neuronal damage in the hippocampus was assessed using Hematoxylin–eosin (HE) staining. The expression level of NeuN, Nestin, Caspase-3, Olig2, CD68 and CD45 were determined by immunohistochemistry (IHC). The RBE range established by comparing the effects of proton and photon irradiation at equivalent biological outcomes. Proton20Gy induced more severe damage to neurons than photon20Gy, but showed no difference compared to photon30Gy. The RBE of neuron was determined to be 1.65. Similarly, both proton20Gy and proton30Gy resulted in more inhibition of oligodendrocytes and activation of microglia in the hippocampal regions than photon20Gy and photon30Gy. However, the expression of Olig2 was higher and CD68 was lower in the proton20Gy group than in the photon30Gy group. The RBE of oligodendrocyte and microglia was estimated to be between 1.1 to 1.65. For neural stem cells (NSCs) and immune cells, there were no significant difference in the expression of Nestin and CD45 between proton and photon irradiation (both 20 and 30 Gy). Therefore, the RBE for NSCs and immune cell was determined to be 1.1. These findings highlight the varying RBE values of different cells in the hippocampus in vivo. Moreover, the actual RBE of the hippocampus may be higher than 1.1, suggesting that using as RBE value of 1.1 in clinical practice may underestimate the toxicities induced by proton radiation.

Background and Purpose To assess the variation of large-volume brain metastases (BMs) boundaries and shapes using enhanced magnetic resonance (MR) scanning with different delay times and to provide a basis for determining the gross tumor target volume (GTV) for radiotherapy of BMs. Materials and Methods We prospectively enrolled 155 patients initially diagnosed with BMs (561 lesions > 1 cm). Contrast-enhanced (CE) T1-weighted imaging scans were performed 1, 3, 5, 10, 18, and 20 min after gadolinium-based contrast agent injection and GTVs were determined as GTV(-1min), GTV(-3min), GTV(-5min), GTV(-10min), GTV(-18min), and GTV(-20min,) respectively, which were subsequently fused in different phases. Fusion of the six GTVs was defined as GTV(-total), which was set as the reference GTV. The volume, shape, and signal intensity of the GTVs and brain white matter (BWM) were compared at different delay times. Results GTV(-3min), GTV(-5min), GTV(-10min), GTV(-18min), and GTV(-20min) volumes increased by 2.2 %, 3.8 %, 6.5 %, 9.5 %, and 10.6 %, respectively (P < 0.05) compared with GTV(-1min). Compared with GTV(-total), GTV(-1min), GTV(-3min), GTV(-5min), GTV(-10min), GTV(-18min), and GTV(-20min) volumes reduced by 25.4 %, 22.1 %, 18.7 %, 15.0 %, 11.2 %, and 10.3 %, respectively (P < 0.05). Compared with GTV(-total), 29 (51.8 %) fused GTVs had a volume reduction rate < 5 %, 45 (80.4 %) had a Dice similarity coefficient > 0.95, and all contained GTV(-10min), GTV(-18min) or GTV(-20min). The signal intensity ratio between the GTV and BWM peaked at 5 min (0.351 +/- 0.24). Conclusion Enhanced MR scans with different delay times show significant differences in the boundaries and shapes of large-volume BMs, and time-delayed multi-phase CE scanning should be used in GTV determination, with time phases >= 10 min being mandatory.

Purpose This study seeks to examine the influence of the heartbeat on the position, volume, and shape of the heart and its substructures during various breathing states. The findings of this study will serve as a valuable reference for dose-volume evaluation of the heart and its substructures in radiotherapy for treating thoracic tumors. Methods Twenty-three healthy volunteers were enrolled in this study, and cine four-dimensional magnetic resonance images were acquired during periods of end-inspiration breath holding (EIBH), end-expiration breath holding (EEBH), and deep end-inspiration breath holding (DIBH). The MR images were used to delineate the heart and its substructures, including the heart, pericardium, left ventricle (LV), left ventricular myocardium, right ventricle (RV), right ventricular myocardium (RVM), ventricular septum (VS), atrial septum (AS), proximal and middle portions of the left anterior descending branch (pmLAD), and proximal portion of the left circumflex coronary branch (pLCX). The changes in each structure with heartbeat were compared among different respiratory states. Results Compared with EIBH, EEBH increased the volume of the heart and its substructures by 0.25–3.66%, while the average Dice similarity coefficient (DSC) increased by − 0.25 to 8.7%; however, the differences were not statistically significant. Conversely, the VS decreased by 0.89 mm in the left–right (LR) direction, and the displacement of the RV in the anterior–posterior (AP) direction significantly decreased by 0.76 mm ( p < 0.05). Compared with EIBH and EEBH, the average volume of the heart and its substructures decreased by 3.08–17.57% and 4.09–20.43%, respectively, during DIBH. Accordingly, statistically significant differences ( p < 0.05) were observed in the volume of the heart, pericardium, LV, RV, RVM, and AS. The average DSC increased by 0–37.04% and − 2.6 to 32.14%, respectively, with statistically significant differences ( p < 0.05) found in the right ventricular myocardium and interatrial septum. Furthermore, the displacements under DIBH decreased in the three directions (i.e.,− 1.73 to 3.47 mm and − 0.36 to 2.51 mm). In this regard, the AP displacement of the heart, LV, RV, RVM, LR direction, LV, RV, and AS showed statistically significant differences ( p < 0.05). The Hausdorff distance (HD) of the heart and its substructures under the three breathing states are all greater than 11 mm. Conclusion The variations in the displacement and shape alterations of the heart and its substructures during cardiac motion under various respiratory states are significant. When assessing the dose-volume index of the heart and its substructures during radiotherapy for thoracic tumors, it is essential to account for the combined impacts of cardiac motion and respiration.