CT assessed morphological features can predict higher mitotic index in gastric gastrointestinal stromal tumors.

OBJECTIVES: To investigate the correlation of the mitotic index (MI) of 1-5 cm gastric gastrointestinal stromal tumors (gGISTs) with CT-identified morphological and first-order radiomics features, incorporating subgroup analysis based on tumor size. METHODS: We enrolled 344 patients across four institutions, each pathologically diagnosed with 1-5 cm gGISTs and undergoing preoperative contrast-enhanced CT scans. Univariate and multivariate analyses were performed to investigate the independent CT morphological high-risk features of MI. Lesions were categorized into four subgroups based on their pathological LD: 1-2 cm (n = 69), 2-3 cm (n = 96), 3-4 cm (n = 107), and 4-5 cm (n = 72). CT morphological high-risk features of MI were evaluated in each subgroup. In addition, first-order radiomics features were extracted on CT images of the venous phase, and the association between these features and MI was investigated. RESULTS: Tumor size (p = 0.04, odds ratio, 1.41; 95% confidence interval: 1.01-1.96) and invasive margin (p < 0.01, odds ratio, 4.55; 95% confidence interval: 1.77-11.73) emerged as independent high-risk features for MI > 5 of 1-5 cm gGISTs from multivariate analysis. In the subgroup analysis, the invasive margin was correlated with MI > 5 in 3-4 cm and 4-5 cm gGISTs (p = 0.02, p = 0.03), and potentially correlated with MI > 5 in 2-3 cm gGISTs (p = 0.07). The energy was the sole first-order radiomics feature significantly correlated with gGISTs of MI > 5, displaying a strong correlation with CT-detected tumor size (Pearson's ρ = 0.85, p < 0.01). CONCLUSIONS: The invasive margin stands out as the sole independent CT morphological high-risk feature for 1-5 cm gGISTs after tumor size-based subgroup analysis, overshadowing intratumoral morphological characteristics and first-order radiomics features. KEY POINTS: Question How can accurate preoperative risk stratification of gGISTs be achieved to support treatment decision-making? Findings Invasive margins may serve as a reliable marker for risk prediction in gGISTs up to 5 cm, rather than surface ulceration, irregular shape, necrosis, or heterogeneous enhancement. Clinical relevance For gGISTs measuring up to 5 cm, preoperative prediction of the metastatic risk could help select patients who could be treated by endoscopic resection, thereby avoiding overtreatment.

Synthetic CT generation for pelvic cases based on deep learning in multi-center datasets.

BACKGROUND AND PURPOSE: To investigate the feasibility of synthesizing computed tomography (CT) images from magnetic resonance (MR) images in multi-center datasets using generative adversarial networks (GANs) for rectal cancer MR-only radiotherapy. MATERIALS AND METHODS: Conventional T2-weighted MR and CT images were acquired from 90 rectal cancer patients at Peking University People's Hospital and 19 patients in public datasets. This study proposed a new model combining contrastive learning loss and consistency regularization loss to enhance the generalization of model for multi-center pelvic MRI-to-CT synthesis. The CT-to-sCT image similarity was evaluated by computing the mean absolute error (MAE), peak signal-to-noise ratio (SNRpeak), structural similarity index (SSIM) and Generalization Performance (GP). The dosimetric accuracy of synthetic CT was verified against CT-based dose distributions for the photon plan. Relative dose differences in the planning target volume and organs at risk were computed. RESULTS: Our model presented excellent generalization with a GP of 0.911 on unseen datasets and outperformed the plain CycleGAN, where MAE decreased from 47.129 to 42.344, SNRpeak improved from 25.167 to 26.979, SSIM increased from 0.978 to 0.992. The dosimetric analysis demonstrated that most of the relative differences in dose and volume histogram (DVH) indicators between synthetic CT and real CT were less than 1%. CONCLUSION: The proposed model can generate accurate synthetic CT in multi-center datasets from T2w-MR images. Most dosimetric differences were within clinically acceptable criteria for photon radiotherapy, demonstrating the feasibility of an MRI-only workflow for patients with rectal cancer.

PGR-KITLG signaling drives a tumor-mast cell regulatory feedback to modulate apoptosis of breast cancer cells.

The immune microenvironment constructed by tumor-infiltrating immune cells and the molecular phenotype defined by hormone receptors (HRs) have been implicated as decisive factors in the regulation of breast cancer (BC) progression. Here, we found that the infiltration of mast cells (MCs) informed impaired prognoses in HR(+) BC but predicted improved prognoses in HR(-) BC. However, molecular features of MCs in different BC remain unclear. We next discovered that HR(-) BC cells were prone to apoptosis under the stimulation of MCs, whereas HR(+) BC cells exerted anti-apoptotic effects. Mechanistically, in HR(+) BC, the KIT ligand (KITLG), a major mast cell growth factor in recruiting and activating MCs, could be transcriptionally upregulated by the progesterone receptor (PGR), and elevate the production of MC-derived granulin (GRN). GRN attenuates TNFα-induced apoptosis in BC cells by competitively binding to TNFR1. Furthermore, disruption of PGR-KITLG signaling by knocking down PGR or using the specific KITLG-cKIT inhibitor iSCK03 potently enhanced the sensitivity of HR(+) BC cells to MC-induced apoptosis and exerted anti-tumor activity. Collectively, these results demonstrate that PGR-KITLG signaling in BC cells preferentially induces GRN expression in MCs to exert anti-apoptotic effects, with potential value in developing precision medicine approaches for diagnosis and treatment.

Semi-supervised auto-segmentation method for pelvic organ-at-risk in magnetic resonance images based on deep-learning.

BACKGROUND AND PURPOSE: In radiotherapy, magnetic resonance (MR) imaging has higher contrast for soft tissues compared to computed tomography (CT) scanning and does not emit radiation. However, manual annotation of the deep learning-based automatic organ-at-risk (OAR) delineation algorithms is expensive, making the collection of large-high-quality annotated datasets a challenge. Therefore, we proposed the low-cost semi-supervised OAR segmentation method using small pelvic MR image annotations. METHODS: We trained a deep learning-based segmentation model using 116 sets of MR images from 116 patients. The bladder, femoral heads, rectum, and small intestine were selected as OAR regions. To generate the training set, we utilized a semi-supervised method and ensemble learning techniques. Additionally, we employed a post-processing algorithm to correct the self-annotation data. Both 2D and 3D auto-segmentation networks were evaluated for their performance. Furthermore, we evaluated the performance of semi-supervised method for 50 labeled data and only 10 labeled data. RESULTS: The Dice similarity coefficient (DSC) of the bladder, femoral heads, rectum and small intestine between segmentation results and reference masks is 0.954, 0.984, 0.908, 0.852 only using self-annotation and post-processing methods of 2D segmentation model. The DSC of corresponding OARs is 0.871, 0.975, 0.975, 0.783, 0.724 using 3D segmentation network, 0.896, 0.984, 0.890, 0.828 using 2D segmentation network and common supervised method. CONCLUSION: The outcomes of our study demonstrate that it is possible to train a multi-OAR segmentation model using small annotation samples and additional unlabeled data. To effectively annotate the dataset, ensemble learning and post-processing methods were employed. Additionally, when dealing with anisotropy and limited sample sizes, the 2D model outperformed the 3D model in terms of performance.

Prediction of macrotrabecular-massive hepatocellular carcinoma by using MR-based models and their prognostic implications.

PURPOSE: To evaluate the efficacy of MRI-based radiomics and clinical models in predicting MTM-HCC. Additionally, to investigate the ability of the radiomics model designed for MTM-HCC identification in predicting disease-free survival (DFS) in patients with HCC. METHODS: A total of 336 patients who underwent oncological resection for HCC between June 2007 and March 2021 were included. 127 patients in Cohort1 were used for MTM-HCC identification, and 209 patients in Cohort2 for prognostic analyses. Radiomics analysis was performed using volumes of interest of HCC delineated on pre-operative MRI images. Radiomics and clinical models were developed using Random Forest algorithm in Cohort1 and a radiomics probability (RP) of MTM-HCC was obtained from the radiomics model. Based on the RP, patients in Cohort2 were divided into a RAD-MTM-HCC (RAD-M) group and a RAD-non-MTM-HCC (RAD-nM) group. Univariate and multivariate Cox regression analyses were employed to identify the independent predictors for DFS of patients in Cohort2. Kaplan-Meier curves were used to compare the DFS between different groups pf patients based on the predictors. RESULTS: The radiomics model for identifying MTM-HCC showed AUCs of 0.916 (95% CI: 0.858-0.960) and 0.833 (95% CI: 0.675-0.935), and the clinical model showed AUCs of 0.760 (95% CI: 0.669-0.836) and 0.704 (95% CI: 0.532-0.843) in the respective training and validation sets. Furthermore, the radiomics biomarker RP, portal or hepatic vein tumor thrombus, irregular rim-like arterial phase hyperenhancement (IRE) and AFP were independent predictors of DFS in patients with HCC. The DFS of RAD-nM group was significantly higher than that of the RAD-M group (p < .001). CONCLUSION: MR-based clinical and radiomic models have the potential to accurately diagnose MTM-HCC. Moreover, the radiomics signature designed to identify MTM-HCC also can be used to predict prognosis in patients with HCC, realizing the diagnostic and prognostic aims at the same time.

Computed tomography-detected extramural venous invasion-related gene signature: a potential negative biomarker of immune checkpoint inhibitor treatment in patients with gastric cancer.

BACKGROUND: To investigate the association between computed tomography (CT)-detected extramural venous invasion (EMVI)-related genes and immunotherapy resistance and immune escape in patients with gastric cancer (GC). METHODS: Thirteen patients with pathologically proven locally advanced GC who had undergone preoperative abdominal contrast-enhanced CT and radical resection surgery were included in this study. Transcriptome sequencing was multidetector performed on the cancerous tissue obtained during surgery, and EMVI-related genes (P value for association < 0.001) were selected. A single-sample gene set enrichment analysis algorithm was also used to divide all GC samples (n = 377) in The Cancer Genome Atlas (TCGA) database into high and low EMVI-immune related groups based on immune-related differential genes. Cluster analysis was used to classify EMVI-immune-related genotypes, and survival among patients was validated in TCGA and Gene Expression Omnibus (GEO) cohorts. The EMVI scores were calculated using principal component analysis (PCA), and GC samples were divided into high and low EMVI score groups. Microsatellite instability (MSI) status, tumor mutation burden (TMB), response rate to immune checkpoint inhibitors (ICIs), immune escape were compared between the high and low EMVI score groups. Hub gene of the model in pan-cancer analysis was also performed. RESULTS: There were 17 EMVI-immune-related genes used for cluster analysis. PCA identified 8 genes (PCH17, SEMA6B, GJA4, CD34, ACVRL1, SOX17, CXCL12, DYSF) that were used to calculate EMVI scores. High EMVI score groups had lower MSI, TMB and response rate of ICIs, status but higher immune escape status. Among the 8 genes used for EMVI scores, CXCL12 and SOX17 were at the core of the protein-protein interaction (PPI) network and had a higher priority in pan-cancer analysis. Immunohistochemical analysis showed that the expression of CXCL12 and SOX17 was significantly higher in CT-detected EMVI-positive samples than in EMVI-negative samples (P < 0.0001). CONCLUSION: A CT-detected EMVI gene signature could be a potential negative biomarker for ICIs treatment, as the signature is negatively correlated with TMB, and MSI, resulting in poorer prognosis.

Risk stratification for 1- to 2-cm gastric gastrointestinal stromal tumors: visual assessment of CT and EUS high-risk features versus CT radiomics analysis.

OBJECTIVES: To investigate the ability of CT and endoscopic sonography (EUS) in predicting the malignant risk of 1-2-cm gastric gastrointestinal stromal tumors (gGISTs) and to clarify whether radiomics could be applied for risk stratification. METHODS: A total of 151 pathologically confirmed 1-2-cm gGISTs from seven institutions were identified by contrast-enhanced CT scans between January 2010 and March 2021. A detailed description of EUS morphological features was available for 73 gGISTs. The association between EUS or CT high-risk features and pathological malignant potential was evaluated. gGISTs were randomly divided into three groups to build the radiomics model, including 74 in the training cohort, 37 in validation cohort, and 40 in testing cohort. The ROIs covering the whole tumor volume were delineated on the CT images of the portal venous phase. The Pearson test and least absolute shrinkage and selection operator (LASSO) algorithm were used for feature selection, and the ROC curves were used to evaluate the model performance. RESULTS: The presence of EUS- and CT-based morphological high-risk features, including calcification, necrosis, intratumoral heterogeneity, irregular border, or surface ulceration, did not differ between very-low and intermediate risk 1-2-cm gGISTs (p > 0.05). The radiomics model consisting of five radiomics features showed favorable performance in discrimination of malignant 1-2-cm gGISTs, with the AUC of the training, validation, and testing cohort as 0.866, 0.812, and 0.766, respectively. CONCLUSIONS: Instead of CT- and EUS-based morphological high-risk features, the CT radiomics model could potentially be applied for preoperative risk stratification of 1-2-cm gGISTs. KEY POINTS: • The presence of EUS- and CT-based morphological high-risk factors, including calcification, necrosis, intratumoral heterogeneity, irregular border, or surface ulceration, did not correlate with the pathological malignant potential of 1-2-cm gGISTs. • The CT radiomics model could potentially be applied for preoperative risk stratification of 1-2-cm gGISTs.

Factors in the occurrence and restoration of hypoparathyroidism after total thyroidectomy for thyroid cancer patients with intraoperative parathyroid autotransplantation.

Introduction: Postoperative hypoparathyroidism (POH) is the most common and important complication for thyroid cancer patients who undergo total thyroidectomy. Intraoperative parathyroid autotransplantation has been demonstrated to be essential in maintaining functional parathyroid tissue, and it has clinical significance in identifying essential factors of serum parathyroid hormone (PTH) levels for patients with parathyroid autotransplantation. This retrospective cohort study aimed to comprehensively investigate influential factors in the occurrence and restoration of POH for patients who underwent total thyroidectomy with intraoperative parathyroid autotransplantation (TTIPA). Method: This study was conducted in a tertiary referral hospital, with a total of 525 patients who underwent TTIPA. The postoperative serum PTH levels were collected after six months, and demographic characteristics, clinical features and associated operative information were analyzed. Results: A total of 66.48% (349/525) of patients who underwent TTIPA were diagnosed with POH. Multivariate logistic regression indicated that Hashimoto's thyroiditis (OR=1.93, 95% CI: 1.09-3.42), P=0.024), the number of transplanted parathyroid glands (OR=2.70, 95% CI: 1.91-3.83, P<0.001) and postoperative blood glucose levels (OR=1.36, 95% CI: 1.06-1.74, P=0.016) were risk factors for POH, and endoscopic surgery (OR=0.39, 95% CI: 0.22-0.68, P=0.001) was a protective factor for POH. Multivariate Cox regression indicated that PTG autotransplantation patients with same-side central lymph node dissection (CLND) (HR=0.50; 95% CI: 0.34-0.73, P<0.001) demonstrated a longer time for increases PTH, and female patients (HR=1.35, 95% CI: 1.00-1.81, P=0.047) were more prone to PTH increases. Additionally, PTG autotransplantation with same-side CLND (HR=0.56, 95% CI: 0.38-0.82, P=0.003) patients had a longer time to PTH restoration, and patients with endoscopic surgery (HR=1.54, 95% CI: 1.04-2.28, P=0.029) were more likely to recover within six months. Conclusion: High postoperative fasting blood glucose levels, a large number of transplanted PTGs, open surgery and Hashimoto's thyroiditis are risk factors for postoperative POH in TTIPA patients. Elevated PTH levels occur earlier in female patients and patients without CLND on the transplant side. PTH returns to normal earlier in patients without CLND and endoscopic surgery on the transplant side.

Detecting and monitoring tumors in orthotopic colorectal liver metastatic animal models with high-resolution ultrasound.

The ability to noninvasively detect and monitor the growth of orthotopic liver transplantation tumors is critical for replicating advanced colorectal cancer liver metastases (CRLMs) in animal models. We assessed the use of high-resolution ultrasound (HRU) to monitor CRLMs transplanted using various cell concentrations. Sixty BALB/c female mice were randomly divided into 3 groups, and murine colonic CT26 cells were injected into the left liver lobe at concentrations of 1 × 102 (group 1), 1 × 103 (group 2), or 1 × 104 (group 3). Tumor presentation, location, number, size, shape, and echogenicity were assessed daily with 24-MHz center frequency HRU starting 6 days after injection. Animals were sacrificed when the largest tumor was ≥ 1 cm in diameter. Sensitivity, specificity, and area under curve (AUC) of CRLMs diagnosed with HRU were calculated using receiver operating characteristic curve analysis. In group 1, 94% of mice formed < 5 tumors, and 41% formed a single tumor. Tumors were first detected with HRU on day 12 in group 1, day 10 in group 2, and day 7 in group 3; tumor volume doubling times were 14-15 days, 11-12 days, and 7-8 days, respectively. With a long diameter threshold of 2.4 mm, diagnostic sensitivity and specificity of HRU were 94.1% and 88.7%, respectively, and the AUC was 0.962. These findings suggest that HRU can be used to accurately detect and monitor the growth of CRLMs in an orthotopic transplantation mouse model, especially when a lower concentration of cells is used.

CT-detected extramural venous invasion-related gene signature for the overall survival prediction in patients with gastric cancer.

BACKGROUND: Computed tomography (CT)-detected extramural venous invasion (EMVI) has been identified as an independent factor that can be used for risk stratification and prediction of prognosis in patients with gastric cancer (GC). Overall survival (OS) is identified as the most important prognostic indicator for GC patients. However, the molecular mechanism of EMVI development and its potential relationship with OS in GC are not fully understood. In this radiogenomics-based study, we sought to investigate the molecular mechanism underlying CT-detected EMVI in patients with GC, and aimed to construct a genomic signature based on EMVI-related genes with the goal of using this signature to predict the OS. MATERIALS AND METHODS: Whole mRNA genome sequencing of frozen tumor samples from 13 locally advanced GC patients was performed to identify EMVI-related genes. EMVI-prognostic hub genes were selected based on overlapping EMVI-related differentially expressed genes and OS-related genes, using a training cohort of 176 GC patients who were included in The Cancer Genome Atlas database. Another 174 GC patients from this database comprised the external validation cohort. A risk stratification model using a seven-gene signature was constructed through the use of a least absolute shrinkage and selection operator Cox regression model. RESULTS: Patients with high risk score showed significantly reduced OS (training cohort, p = 1.143e-04; validation cohort, p = 2.429e-02). Risk score was an independent predictor of OS in multivariate Cox regression analyses (training cohort, HR = 2.758; 95% CI: 1.825-4.169; validation cohort, HR = 2.173; 95% CI: 1.347-3.505; p < 0.001 for both). Gene functions/pathways of the seven-gene signature mainly included cell proliferation, cell adhesion, regulation of metal ion transport, and epithelial to mesenchymal transition. CONCLUSIONS: A CT-detected EMVI-related gene model could be used to predict the prognosis in GC patients, potentially providing clinicians with additional information regarding appropriate therapeutic strategy and medical decision-making.