The Image Biomarker Standardization Initiative: Standardized Convolutional Filters for Reproducible Radiomics and Enhanced Clinical Insights.

Filters are commonly used to enhance specific structures and patterns in images, such as vessels or peritumoral regions, to enable clinical insights beyond the visible image using radiomics. However, their lack of standardization restricts reproducibility and clinical translation of radiomics decision support tools. In this special report, teams of researchers who developed radiomics software participated in a three-phase study (September 2020 to December 2022) to establish a standardized set of filters. The first two phases focused on finding reference filtered images and reference feature values for commonly used convolutional filters: mean, Laplacian of Gaussian, Laws and Gabor kernels, separable and nonseparable wavelets (including decomposed forms), and Riesz transformations. In the first phase, 15 teams used digital phantoms to establish 33 reference filtered images of 36 filter configurations. In phase 2, 11 teams used a chest CT image to derive reference values for 323 of 396 features computed from filtered images using 22 filter and image processing configurations. Reference filtered images and feature values for Riesz transformations were not established. Reproducibility of standardized convolutional filters was validated on a public data set of multimodal imaging (CT, fluorodeoxyglucose PET, and T1-weighted MRI) in 51 patients with soft-tissue sarcoma. At validation, reproducibility of 486 features computed from filtered images using nine configurations × three imaging modalities was assessed using the lower bounds of 95% CIs of intraclass correlation coefficients. Out of 486 features, 458 were found to be reproducible across nine teams with lower bounds of 95% CIs of intraclass correlation coefficients greater than 0.75. In conclusion, eight filter types were standardized with reference filtered images and reference feature values for verifying and calibrating radiomics software packages. A web-based tool is available for compliance checking.

Correlation of microscopic tumor extension with tumor microenvironment in esophageal cancer patients.

OBJECTIVE: In the era of image-guided adaptive radiotherapy, definition of the clinical target volume (CTV) is a challenge in various solid tumors, including esophageal cancer (EC). Many tumor microenvironmental factors, e.g., tumor cell proliferation or cancer stem cells, are hypothesized to be involved in microscopic tumor extension (MTE). Therefore, this study assessed the expression of FAK, ILK, CD44, HIF-1α, and Ki67 in EC patients after neoadjuvant radiochemotherapy followed by tumor resection (NRCHT+R) and correlated these markers with the MTE. METHODS: Formalin-fixed paraffin-embedded tumor resection specimens of ten EC patients were analyzed using multiplex immunofluorescence staining. Since gold fiducial markers had been endoscopically implanted at the proximal and distal tumor borders prior to NRCHT+R, correlation of the markers with the MTE was feasible. RESULTS: In tumor resection specimens of EC patients, the overall percentages of FAK+, CD44+, HIF-1α+, and Ki67+ cells were higher in tumor nests than in the tumor stroma, with the outcome for Ki67+ cells reaching statistical significance (p < 0.001). Conversely, expression of ILK+ cells was higher in tumor stroma, albeit not statistically significantly. In three patients, MTE beyond the fiducial markers was found, reaching up to 31 mm. CONCLUSION: Our findings indicate that the overall expression of FAK, HIF-1α, Ki67, and CD44 was higher in tumor nests, whereas that of ILK was higher in tumor stroma. Differences in the TME between patients with residual tumor cells in the original CTV compared to those without were not found. Thus, there is insufficient evidence that the TME influences the required CTV margin on an individual patient basis. TRIAL REGISTRATION NUMBER AND DATE: BO-EK-148042017 and BO-EK-177042022 on 20.06.2022, DRKS00011886, https://drks.de/search/de/trial/DRKS00011886 .

Transfusions of packed red blood cells in surgery for liver cancer: predictor of impaired overall survival but not recurrence-free survival - impact of blood transfusions in liver surgery.

BACKGROUND: Liver surgery remains a cornerstone of potentially curative multimodal treatments for primary malignancies of the liver and hepatic metastases. Improving perioperative safety is a prerequisite in this context. Perioperative blood transfusions negatively influence postoperative recovery. This study aimed to identify risk factors for perioperative packed red blood cell (PRBC) transfusion and to elucidate its effect on postoperative outcomes. METHODS: This was an observational study of a prospective data collection. A monocentric, retrospective analysis of 1118 hepatectomies at the University Hospital Carl Gustav Carus between 2013 and 2020 was conducted to compare postoperative short- and long-term outcomes in patients undergoing curative intended liver resection of hepatic primary or secondary malignancies. The outcomes were compared between 356 patients (31.8%) who received PRBC transfusions during surgery or within 7 days after surgery and 762 patients (68.2%) who did not receive PRBC transfusions. RESULTS: Preoperative anemia could be observed in 45.0% of the whole cohort: 65.7% in the PRBC transfusion group and 35.3% in the nontransfused group. Postoperative complications were significantly more common in the PRBC transfusion group in association with prolonged lengths of hospital stay and increased 30-day mortality than in the nontransfused group. After adjustment for possible confounders, preexisting kidney failure, preoperative hemoglobin and albumin levels outside of the reference range, intraoperative plasma transfusions, and overall surgery time were recognized as negative predictors for perioperative PRBC transfusions. PRBC transfusion increased the risk of death by approximately 38.8% (hazard ratio, 1.388; 95% CI, 1.027-1.876; P = .033), whereas no influence on recurrence-free survival (RFS) was observed. CONCLUSION: PRBC transfusions were associated with postoperative morbidity and mortality after curative-intended surgery for liver cancers and represented an independent poor prognostic indicator for overall survival but not for RFS.

Diffusion decrease in normal-appearing white matter structures following photon or proton irradiation indicates differences in regional radiosensitivity.

PURPOSE: Radio(chemo)therapy (RCT) as part of the standard treatment of glioma patients, inevitably leads to radiation exposure of the tumor-surrounding normal-appearing (NA) tissues. The effect of radiotherapy on the brain microstructure can be assessed by magnetic resonance imaging (MRI) using diffusion tensor imaging (DTI). The aim of this study was to analyze regional DTI changes of white matter (WM) structures and to determine their dose- and time-dependency. METHODS: As part of a longitudinal prospective clinical study (NCT02824731), MRI data of 23 glioma patients treated with proton or photon beam therapy were acquired at three-monthly intervals until 36 months following irradiation. Mean, radial and axial diffusivity (MD, RD, AD) as well as fractional anisotropy (FA) were investigated in the NA tissue of 15 WM structures and their dependence on radiation dose, follow-up time and distance to the clinical target volume (CTV) was analyzed in a multivariate linear regression model. Due to the small and non-comparable patient numbers for proton and photon beam irradiation, a separate assessment of the findings per treatment modality was not performed. RESULTS: Four WM structures (i.e., internal capsule, corona radiata, posterior thalamic radiation, and superior longitudinal fasciculus) showed statistically significantly decreased RD and MD after RT, whereas AD decrease and FA increase occurred less frequently. The posterior thalamic radiation showed the most pronounced changes after RCT [i.e., ΔRD = -8.51 % (p = 0.012), ΔMD = -6.14 % (p = 0.012)]. The DTI changes depended significantly on mean dose and time. CONCLUSION: Significant changes in DTI for WM substructures were found even at low radiation doses. These findings may prompt new radiation dose constraints sparing the vulnerable structures from damage and subsequent side-effects.

Radiomics for residual tumour detection and prognosis in newly diagnosed glioblastoma based on postoperative [11C] methionine PET and T1c-w MRI.

Personalized treatment strategies based on non-invasive biomarkers have potential to improve patient management in patients with newly diagnosed glioblastoma (GBM). The residual tumour burden after surgery in GBM patients is a prognostic imaging biomarker. However, in clinical patient management, its assessment is a manual and time-consuming process that is at risk of inter-rater variability. Furthermore, the prediction of patient outcome prior to radiotherapy may identify patient subgroups that could benefit from escalated radiotherapy doses. Therefore, in this study, we investigate the capabilities of traditional radiomics and 3D convolutional neural networks for automatic detection of the residual tumour status and to prognosticate time-to-recurrence (TTR) and overall survival (OS) in GBM using postoperative [11C] methionine positron emission tomography (MET-PET) and gadolinium-enhanced T1-w magnetic resonance imaging (MRI). On the independent test data, the 3D-DenseNet model based on MET-PET achieved the best performance for residual tumour detection, while the logistic regression model with conventional radiomics features performed best for T1c-w MRI (AUC: MET-PET 0.95, T1c-w MRI 0.78). For the prognosis of TTR and OS, the 3D-DenseNet model based on MET-PET integrated with age and MGMT status achieved the best performance (Concordance-Index: TTR 0.68, OS 0.65). In conclusion, we showed that both deep-learning and conventional radiomics have potential value for supporting image-based assessment and prognosis in GBM. After prospective validation, these models may be considered for treatment personalization.

Dose and dose rate dependence of the tissue sparing effect at ultra-high dose rate studied for proton and electron beams using the zebrafish embryo model.

PURPOSE: A better characterization of the dependence of the tissue sparing effect at ultra-high dose rate (UHDR) on physical beam parameters (dose, dose rate, radiation quality) would be helpful towards a mechanistic understanding of the FLASH effect and for its broader clinical translation. To address this, a comprehensive study on the normal tissue sparing at UHDR using the zebrafish embryo (ZFE) model was conducted. METHODS: One-day-old ZFE were irradiated over a wide dose range (15-95 Gy) in three different beams (proton entrance channel, proton spread out Bragg peak and 30 MeV electrons) at UHDR and reference dose rate. After irradiation the ZFE were incubated for 4 days and then analyzed for four different biological endpoints (pericardial edema, curved spine, embryo length and eye diameter). RESULTS: Dose-effect curves were obtained and a sparing effect at UHDR was observed for all three beams. It was demonstrated that proton relative biological effectiveness and UHDR sparing are both relevant to predict the resulting dose response. Dose dependent FLASH modifying factors (FMF) for ZFE were found to be compatible with rodent data from the literature. It was found that the UHDR sparing effect saturates at doses above âˆ¼ 50 Gy with an FMF of âˆ¼ 0.7-0.8. A strong dose rate dependence of the tissue sparing effect in ZFE was observed. The magnitude of the maximum sparing effect was comparable for all studied biological endpoints. CONCLUSION: The ZFE model was shown to be a suitable pre-clinical high-throughput model for radiobiological studies on FLASH radiotherapy, providing results comparable to rodent models. This underlines the relevance of ZFE studies for FLASH radiotherapy research.

Type I conventional dendritic cells and CD8+ T cells predict favorable clinical outcome of head and neck squamous cell carcinoma patients.

Head and neck squamous cell carcinoma (HNSCC) is one of the most common tumor entities worldwide, with human papillomavirus (HPV) infection contributing to cancer development. Conventional therapies achieve only limited efficiency, especially in recurrent or metastatic HNSCC. As the immune landscape decisively impacts the survival of patients and treatment efficacy, this study comprehensively investigated the immunological tumor microenvironment (TME) and its association with patient outcome, with special focus on several dendritic cell (DC) and T lymphocyte subpopulations. Therefore, formalin-fixed paraffin-embedded tumor samples of 56 HNSCC patients, who have undergone resection and adjuvant radiotherapy, were analyzed by multiplex immunohistochemistry focusing on the detailed phenotypic characterization and spatial distribution of DCs, CD8+ T cells, and T-helper cell subsets in different tumor compartments. Immune cell densities and proportions were correlated with clinical characteristics of the whole HNSCC cohort and different HPV- or hypoxia-associated subcohorts. Tumor stroma was highly infiltrated by plasmacytoid DCs and T lymphocytes. Among the T-helper cells and CD8+ T cells, stromal regulatory T cells and intraepithelial exhausted CD8+ T cells expressing programmed cell death protein-1 (PD-1+) and/or lymphocyte-activation gene-3 (LAG-3+) were the predominant phenotypes, indicating an immunosuppressive TME. HPV-associated tumors showed significantly higher infiltration of type I and type II conventional DCs (cDC1, cDC2) as well as several CD8+ T cell phenotypes including exhausted, activated, and proliferating T cells. On the contrary, tumors with hypoxia-associated gene signatures exhibited reduced infiltration for these immune cells. By multivariate Cox regression, immune-related prognostic factors were identified. Patient clusters defined by high infiltration of DCs and T lymphocytes combined with HPV positivity or low hypoxia showed significantly prolonged survival. Thereby, cDC1 and CD8+ T cells emerged as independent prognostic factors for local and distant recurrence. These results might contribute to the implementation of an immune cell infiltration score predicting HNSCC patients' survival and such patient stratification might improve the design of future individualized radiochemo-(immuno)therapies.

Comparative analysis of radiomics and deep-learning algorithms for survival prediction in hepatocellular carcinoma.

To examine the comparative robustness of computed tomography (CT)-based conventional radiomics and deep-learning convolutional neural networks (CNN) to predict overall survival (OS) in HCC patients. Retrospectively, 114 HCC patients with pretherapeutic CT of the liver were randomized into a development (n = 85) and a validation (n = 29) cohort, including patients of all tumor stages and several applied therapies. In addition to clinical parameters, image annotations of the liver parenchyma and of tumor findings on CT were available. Cox-regression based on radiomics features and CNN models were established and combined with clinical parameters to predict OS. Model performance was assessed using the concordance index (C-index). Log-rank tests were used to test model-based patient stratification into high/low-risk groups. The clinical Cox-regression model achieved the best validation performance for OS (C-index [95% confidence interval (CI)] 0.74 [0.57-0.86]) with a significant difference between the risk groups (p = 0.03). In image analysis, the CNN models (lowest C-index [CI] 0.63 [0.39-0.83]; highest C-index [CI] 0.71 [0.49-0.88]) were superior to the corresponding radiomics models (lowest C-index [CI] 0.51 [0.30-0.73]; highest C-index [CI] 0.66 [0.48-0.79]). A significant risk stratification was not possible (p > 0.05). Under clinical conditions, CNN-algorithms demonstrate superior prognostic potential to predict OS in HCC patients compared to conventional radiomics approaches and could therefore provide important information in the clinical setting, especially when clinical data is limited.

Epigenetic Targeting to Overcome Radioresistance in Head and Neck Cancer.

(1) Background: The sensitivity of head and neck squamous cell carcinoma (HNSCC) to ionizing radiation, among others, is determined by the number of cells with high clonogenic potential and stem-like features. These cellular characteristics are dynamically regulated in response to treatment and may lead to an enrichment of radioresistant cells with a cancer stem cell (CSC) phenotype. Epigenetic mechanisms, particularly DNA and histone methylation, are key regulators of gene-specific transcription and cellular plasticity. Therefore, we hypothesized that specific epigenetic targeting may prevent irradiation-induced plasticity and may sensitize HNSCC cells to radiotherapy. (2) Methods: We compared the DNA methylome and intracellular concentrations of tricarboxylic acid cycle metabolites in radioresistant FaDu and Cal33 cell lines with their parental controls, as well as aldehyde dehydrogenase (ALDH)-positive CSCs with negative controls. Moreover, we conducted a screen of a chemical library targeting enzymes involved in epigenetic regulation in combination with irradiation and analyzed the clonogenic potential, sphere formation, and DNA repair capacity to identify compounds with both radiosensitizing and CSC-targeting potential. (3) Results: We identified the histone demethylase inhibitor GSK-J1, which targets UTX (KDM6A) and JMJD3 (KDM6B), leading to increased H3K27 trimethylation, heterochromatin formation, and gene silencing. The clonogenic survival assay after siRNA-mediated knock-down of both genes radiosensitized Cal33 and SAS cell lines. Moreover, high KDM6A expression in tissue sections of patients with HNSCC was associated with improved locoregional control after primary (n = 137) and post-operative (n = 187) radio/chemotherapy. Conversely, high KDM6B expression was a prognostic factor for reduced overall survival. (4) Conclusions: Within this study, we investigated cellular and molecular mechanisms underlying irradiation-induced cellular plasticity, a key inducer of radioresistance, with a focus on epigenetic alterations. We identified UTX (KDM6A) as a putative prognostic and therapeutic target for HNSCC patients treated with radiotherapy.

A deep-learning-based surrogate model for Monte-Carlo simulations of the linear energy transfer in primary brain tumor patients treated with proton-beam radiotherapy.

Objective.This study explores the use of neural networks (NNs) as surrogate models for Monte-Carlo (MC) simulations in predicting the dose-averaged linear energy transfer (LETd) of protons in proton-beam therapy based on the planned dose distribution and patient anatomy in the form of computed tomography (CT) images. As LETdis associated with variability in the relative biological effectiveness (RBE) of protons, we also evaluate the implications of using NN predictions for normal tissue complication probability (NTCP) models within a variable-RBE context.Approach.The predictive performance of three-dimensional NN architectures was evaluated using five-fold cross-validation on a cohort of brain tumor patients (n= 151). The best-performing model was identified and externally validated on patients from a different center (n= 107). LETdpredictions were compared to MC-simulated results in clinically relevant regions of interest. We assessed the impact on NTCP models by leveraging LETdpredictions to derive RBE-weighted doses, using the Wedenberg RBE model.Main results.We found NNs based solely on the planned dose distribution, i.e. without additional usage of CT images, can approximate MC-based LETddistributions. Root mean squared errors (RMSE) for the median LETdwithin the brain, brainstem, CTV, chiasm, lacrimal glands (ipsilateral/contralateral) and optic nerves (ipsilateral/contralateral) were 0.36, 0.87, 0.31, 0.73, 0.68, 1.04, 0.69 and 1.24 keV µm-1, respectively. Although model predictions showed statistically significant differences from MC outputs, these did not result in substantial changes in NTCP predictions, with RMSEs of at most 3.2 percentage points.Significance.The ability of NNs to predict LETdbased solely on planned dose distributions suggests a viable alternative to compute-intensive MC simulations in a variable-RBE setting. This is particularly useful in scenarios where MC simulation data are unavailable, facilitating resource-constrained proton therapy treatment planning, retrospective patient data analysis and further investigations on the variability of proton RBE.

ALDH1A1 drives prostate cancer metastases and radioresistance by interplay with AR- and RAR-dependent transcription.

Rationale: Current therapies for metastatic osseous disease frequently fail to provide a durable treatment response. To date, there are only limited therapeutic options for metastatic prostate cancer, the mechanisms that drive the survival of metastasis-initiating cells are poorly characterized, and reliable prognostic markers are missing. A high aldehyde dehydrogenase (ALDH) activity has been long considered a marker of cancer stem cells (CSC). Our study characterized a differential role of ALDH1A1 and ALDH1A3 genes as regulators of prostate cancer progression and metastatic growth. Methods: By genetic silencing of ALDH1A1 and ALDH1A3 in vitro, in xenografted zebrafish and murine models, and by comparative immunohistochemical analyses of benign, primary tumor, and metastatic specimens from patients with prostate cancer, we demonstrated that ALDH1A1 and ALDH1A3 maintain the CSC phenotype and radioresistance and regulate bone metastasis-initiating cells. We have validated ALDH1A1 and ALDH1A3 as potential biomarkers of clinical outcomes in the independent cohorts of patients with PCa. Furthermore, by RNAseq, chromatin immunoprecipitation (ChIP), and biostatistics analyses, we suggested the molecular mechanisms explaining the role of ALDH1A1 in PCa progression. Results: We found that aldehyde dehydrogenase protein ALDH1A1 positively regulates tumor cell survival in circulation, extravasation, and metastatic dissemination, whereas ALDH1A3 plays the opposite role. ALDH1A1 and ALDH1A3 are differentially expressed in metastatic tumors of patients with prostate cancer, and their expression levels oppositely correlate with clinical outcomes. Prostate cancer progression is associated with the increasing interplay of ALDH1A1 with androgen receptor (AR) and retinoid receptor (RAR) transcriptional programs. Polo-like kinase 3 (PLK3) was identified as a transcriptional target oppositely regulated by ALDH1A1 and ALDH1A3 genes in RAR and AR-dependent manner. PLK3 contributes to the control of prostate cancer cell proliferation, migration, DNA repair, and radioresistance. ALDH1A1 gain in prostate cancer bone metastases is associated with high PLK3 expression. Conclusion: This report provides the first evidence that ALDH1A1 and PLK3 could serve as biomarkers to predict metastatic dissemination and radiotherapy resistance in patients with prostate cancer and could be potential therapeutic targets to eliminate metastasis-initiating and radioresistant tumor cell populations.

The value of subcutaneous xenografts for individualised radiotherapy in HNSCC: Robust gene signature correlates with radiotherapy outcome in patients and xenografts.

PURPOSE: To assess the robustness of prognostic biomarkers and molecular tumour subtypes developed for patients with head and neck squamous cell carcinoma (HNSCC) on cell-line derived HNSCC xenograft models, and to develop a novel biomarker signature by combining xenograft and patient datasets. MATERIALS AND METHODS: Mice bearing xenografts (n = 59) of ten HNSCC cell lines and a retrospective, multicentre patient cohort (n = 242) of the German Cancer Consortium-Radiation Oncology Group (DKTK-ROG) were included. All patients received postoperative radiochemotherapy (PORT-C). Gene expression analysis was conducted using GeneChip Human Transcriptome Arrays. Xenografts were stratified based on their molecular subtypes and previously established gene classifiers. The dose to control 50 % of tumours (TCD50) was compared between these groups. Using differential gene expression analyses combining xenograft and patient data, a gene signature was developed to define risk groups for the primary endpoint loco-regional control (LRC). RESULTS: Tumours of mesenchymal subtype were characterized by a higher TCD50 (xenografts, p < 0.001) and lower LRC (patients, p < 0.001) compared to the other subtypes. Similar to previously published patient data, hypoxia- and radioresistance-related gene signatures were associated with high TCD50 values. A 2-gene signature (FN1, SERPINE1) was developed that was prognostic for TCD50 (xenografts, p < 0.001) and for patient outcome in independent validation (LRC: p = 0.007). CONCLUSION: Genetic prognosticators of outcome for patients after PORT-C and subcutaneous xenografts after primary clinically relevant irradiation show similarity. The identified robust 2-gene signature may help to guide patient stratification, after prospective validation. Thus, xenografts remain a valuable resource for translational research towards the development of individualized radiotherapy.