Development and benchmarking of a Deep Learning-based MRI-guided gross tumor segmentation algorithm for Radiomics analyses in extremity soft tissue sarcomas.

BACKGROUND: Volume of interest (VOI) segmentation is a crucial step for Radiomics analyses and radiotherapy (RT) treatment planning. Because it can be time-consuming and subject to inter-observer variability, we developed and tested a Deep Learning-based automatic segmentation (DLBAS) algorithm to reproducibly predict the primary gross tumor as VOI for Radiomics analyses in extremity soft tissue sarcomas (STS). METHODS: A DLBAS algorithm was trained on a cohort of 157 patients and externally tested on an independent cohort of 87 patients using contrast-enhanced MRI. Manual tumor delineations by a radiation oncologist served as ground truths (GTs). A benchmark study with 20 cases from the test cohort compared the DLBAS predictions against manual VOI segmentations of two residents (ERs) and clinical delineations of two radiation oncologists (ROs). The ROs rated DLBAS predictions regarding their direct applicability. RESULTS: The DLBAS achieved a median dice similarity coefficient (DSC) of 0.88 against the GTs in the entire test cohort (interquartile range (IQR): 0.11) and a median DSC of 0.89 (IQR 0.07) and 0.82 (IQR 0.10) in comparison to ERs and ROs, respectively. Radiomics feature stability was high with a median intraclass correlation coefficient of 0.97, 0.95 and 0.94 for GTs, ERs, and ROs, respectively. DLBAS predictions were deemed clinically suitable by the two ROs in 35% and 20% of cases, respectively. CONCLUSION: The results demonstrate that the DLBAS algorithm provides reproducible VOI predictions for radiomics feature extraction. Variability remains regarding direct clinical applicability of predictions for RT treatment planning.

Toward image-based personalization of glioblastoma therapy: A clinical and biological validation study of a novel, deep learning-driven tumor growth model.

Background: The diffuse growth pattern of glioblastoma is one of the main challenges for accurate treatment. Computational tumor growth modeling has emerged as a promising tool to guide personalized therapy. Here, we performed clinical and biological validation of a novel growth model, aiming to close the gap between the experimental state and clinical implementation. Methods: One hundred and twenty-four patients from The Cancer Genome Archive (TCGA) and 397 patients from the UCSF Glioma Dataset were assessed for significant correlations between clinical data, genetic pathway activation maps (generated with PARADIGM; TCGA only), and infiltration (Dw) as well as proliferation (ρ) parameters stemming from a Fisher-Kolmogorov growth model. To further evaluate clinical potential, we performed the same growth modeling on preoperative magnetic resonance imaging data from 30 patients of our institution and compared model-derived tumor volume and recurrence coverage with standard radiotherapy plans. Results: The parameter ratio Dw/ρ (P < .05 in TCGA) as well as the simulated tumor volume (P < .05 in TCGA/UCSF) were significantly inversely correlated with overall survival. Interestingly, we found a significant correlation between 11 proliferation pathways and the estimated proliferation parameter. Depending on the cutoff value for tumor cell density, we observed a significant improvement in recurrence coverage without significantly increased radiation volume utilizing model-derived target volumes instead of standard radiation plans. Conclusions: Identifying a significant correlation between computed growth parameters and clinical and biological data, we highlight the potential of tumor growth modeling for individualized therapy of glioblastoma. This might improve the accuracy of radiation planning in the near future.

Tumor Contact With Internal Mammary Perforator Vessels as Risk Factor for Gross Internal Mammary Lymph Node Involvement in Patients With Breast Cancer.

PURPOSE: The identification of internal mammary lymph node metastases and the assessment of associated risk factors are crucial for adjuvant regional lymph node irradiation in patients with breast cancer. The current study aims to investigate whether tumor contact with internal mammary perforator vessels is associated with gross internal mammary lymph node involvement. METHODS AND MATERIALS: We included 297 patients with primary breast cancer and gross internal mammary (IMN+) and/or axillary metastases as well as 230 patients without lymph node metastases. Based on pretreatment dynamic contrast-enhanced magnetic resonance imaging, we assessed contact of the tumor with the internal mammary perforating vessels (IMPV). RESULTS: A total of 59 patients had ipsilateral IMN+ (iIMN+), 10 patients had contralateral IMN+ (cIMN+), and 228 patients had ipsilateral axillary metastases without IMN; 230 patients had node-negative breast cancer. In patients with iIMN+, 100% of tumors had contact with ipsilateral IMPV, with 94.9% (n = 56) classified as major contact. In iIMN- patients, major IMPV contact was observed in only 25.3% (n = 116), and 36.2% (n = 166) had no IMPV contact at all. Receiver operating characteristic analysis revealed that "major IMPV contact" was more accurate in predicting iIMN+ (area under the curve, 0.85) compared with a multivariate model combining grade of differentiation, tumor site, size, and molecular subtype (area under the curve, 0.65). Strikingly, among patients with cIMN+, 100% of tumors had contact with a crossing contralateral IMPV, whereas in cIMN- patients, IMPVs to the contralateral side were observed in only 53.4% (iIMN+) and 24.8% (iIMN-), respectively. CONCLUSIONS: Tumor contact with the IMPV is highly associated with risk of gross IMN involvement. Further studies are warranted to investigate whether this identified risk factor is also associated with microscopic IMN involvement and whether it can assist in the selection of patients with breast cancer for irradiation of the internal mammary lymph nodes.

European association of urology risk stratification predicts outcome in patients receiving PSMA-PET-planned salvage radiotherapy for biochemical recurrence following radical prostatectomy.

PURPOSE: The European Association of Urology (EAU) proposed a risk stratification (high vs. low risk) for patients with biochemical recurrence (BR) following radical prostatectomy (RP). Here we investigated whether this stratification accurately predicts outcome, particularly in patients staged with PSMA-PET. METHODS: For this study, we used a retrospective database including 1222 PSMA-PET-staged prostate cancer patients who were treated with salvage radiotherapy (SRT) for BR, at 11 centers in 5 countries. Patients with lymph node metastases (pN1 or cN1) or unclear EAU risk group were excluded. The remaining cohort comprised 526 patients, including 132 low-risk and 394 high-risk patients. RESULTS: The median follow-up time after SRT was 31.0 months. The 3-year biochemical progression-free survival (BPFS) was 85.7 % in EAU low-risk versus 69.4 % in high-risk patients (p = 0.002). The 3-year metastasis-free survival (MFS) was 94.4 % in low-risk versus 87.6 % in high-risk patients (p = 0.005). The 3-year overall survival (OS) was 99.0 % in low-risk versus 99.6 % in high-risk patients (p = 0.925). In multivariate analysis, EAU risk group remained a statistically significant predictor of BPFS (p = 0.003, HR 2.022, 95 % CI 1.262-3.239) and MFS (p = 0.013, HR 2.986, 95 % CI 1.262-7.058). CONCLUSION: Our data support the EAU risk group definition. EAU risk grouping for BCR reliably predicted outcome in patients staged lymph node-negative after RP and with PSMA-PET before SRT. To our knowledge, this is the first study validating the EAU risk grouping in patients treated with PSMA-PET-planned SRT.

The prognostic significance of a negative PSMA-PET scan prior to salvage radiotherapy following radical prostatectomy.

AIM: The optimal management for early recurrent prostate cancer following radical prostatectomy (RP) in patients with negative prostate-specific membrane antigen positron-emission tomography (PSMA-PET) scan is an ongoing subject of debate. The aim of this study was to evaluate the outcome of salvage radiotherapy (SRT) in patients with biochemical recurrence with negative PSMA PET finding. METHODS: This retrospective, multicenter (11 centers, 5 countries) analysis included patients who underwent SRT following biochemical recurrence (BR) of PC after RP without evidence of disease on PSMA-PET staging. Biochemical recurrence-free survival (bRFS), metastatic-free survival (MFS) and overall survival (OS) were assessed using Kaplan-Meier method. Multivariable Cox proportional hazards regression assessed predefined predictors of survival outcomes. RESULTS: Three hundred patients were included, 253 (84.3%) received SRT to the prostate bed only, 46 (15.3%) additional elective pelvic nodal irradiation, respectively. Only 41 patients (13.7%) received concomitant androgen deprivation therapy (ADT). Median follow-up after SRT was 33 months (IQR: 20-46 months). Three-year bRFS, MFS, and OS following SRT were 73.9%, 87.8%, and 99.1%, respectively. Three-year bRFS was 77.5% and 48.3% for patients with PSA levels before PSMA-PET ≤ 0.5 ng/ml and > 0.5 ng/ml, respectively. Using univariate analysis, the International Society of Urological Pathology (ISUP) grade > 2 (p = 0.006), metastatic pelvic lymph nodes at surgery (p = 0.032), seminal vesicle involvement (p < 0.001), pre-SRT PSA level of > 0.5 ng/ml (p = 0.004), and lack of concomitant ADT (p = 0.023) were significantly associated with worse bRFS. On multivariate Cox proportional hazards, seminal vesicle infiltration (p = 0.007), ISUP score >2 (p = 0.048), and pre SRT PSA level > 0.5 ng/ml (p = 0.013) remained significantly associated with worse bRFS. CONCLUSION: Favorable bRFS after SRT in patients with BR and negative PSMA-PET following RP was achieved. These data support the usage of early SRT for patients with negative PSMA-PET findings.

The importance of planning CT-based imaging features for machine learning-based prediction of pain response.

Patients suffering from painful spinal bone metastases (PSBMs) often undergo palliative radiation therapy (RT), with an efficacy of approximately two thirds of patients. In this exploratory investigation, we assessed the effectiveness of machine learning (ML) models trained on radiomics, semantic and clinical features to estimate complete pain response. Gross tumour volumes (GTV) and clinical target volumes (CTV) of 261 PSBMs were segmented on planning computed tomography (CT) scans. Radiomics, semantic and clinical features were collected for all patients. Random forest (RFC) and support vector machine (SVM) classifiers were compared using repeated nested cross-validation. The best radiomics classifier was trained on CTV with an area under the receiver-operator curve (AUROC) of 0.62 ± 0.01 (RFC; 95% confidence interval). The semantic model achieved a comparable AUROC of 0.63 ± 0.01 (RFC), significantly below the clinical model (SVM, AUROC: 0.80 ± 0.01); and slightly lower than the spinal instability neoplastic score (SINS; LR, AUROC: 0.65 ± 0.01). A combined model did not improve performance (AUROC: 0,74 ± 0,01). We could demonstrate that radiomics and semantic analyses of planning CTs allowed for limited prediction of therapy response to palliative RT. ML predictions based on established clinical parameters achieved the best results.

Multitask Learning with Convolutional Neural Networks and Vision Transformers Can Improve Outcome Prediction for Head and Neck Cancer Patients.

Neural-network-based outcome predictions may enable further treatment personalization of patients with head and neck cancer. The development of neural networks can prove challenging when a limited number of cases is available. Therefore, we investigated whether multitask learning strategies, implemented through the simultaneous optimization of two distinct outcome objectives (multi-outcome) and combined with a tumor segmentation task, can lead to improved performance of convolutional neural networks (CNNs) and vision transformers (ViTs). Model training was conducted on two distinct multicenter datasets for the endpoints loco-regional control (LRC) and progression-free survival (PFS), respectively. The first dataset consisted of pre-treatment computed tomography (CT) imaging for 290 patients and the second dataset contained combined positron emission tomography (PET)/CT data of 224 patients. Discriminative performance was assessed by the concordance index (C-index). Risk stratification was evaluated using log-rank tests. Across both datasets, CNN and ViT model ensembles achieved similar results. Multitask approaches showed favorable performance in most investigations. Multi-outcome CNN models trained with segmentation loss were identified as the optimal strategy across cohorts. On the PET/CT dataset, an ensemble of multi-outcome CNNs trained with segmentation loss achieved the best discrimination (C-index: 0.29, 95% confidence interval (CI): 0.22-0.36) and successfully stratified patients into groups with low and high risk of disease progression (p=0.003). On the CT dataset, ensembles of multi-outcome CNNs and of single-outcome ViTs trained with segmentation loss performed best (C-index: 0.26 and 0.26, CI: 0.18-0.34 and 0.18-0.35, respectively), both with significant risk stratification for LRC in independent validation (p=0.002 and p=0.011). Further validation of the developed multitask-learning models is planned based on a prospective validation study, which has recently completed recruitment.

Identifying core MRI sequences for reliable automatic brain metastasis segmentation.

BACKGROUND: Many automatic approaches to brain tumor segmentation employ multiple magnetic resonance imaging (MRI) sequences. The goal of this project was to compare different combinations of input sequences to determine which MRI sequences are needed for effective automated brain metastasis (BM) segmentation. METHODS: We analyzed preoperative imaging (T1-weighted sequence ± contrast-enhancement (T1/T1-CE), T2-weighted sequence (T2), and T2 fluid-attenuated inversion recovery (T2-FLAIR) sequence) from 339 patients with BMs from seven centers. A baseline 3D U-Net with all four sequences and six U-Nets with plausible sequence combinations (T1-CE, T1, T2-FLAIR, T1-CE + T2-FLAIR, T1-CE + T1 + T2-FLAIR, T1-CE + T1) were trained on 239 patients from two centers and subsequently tested on an external cohort of 100 patients from five centers. RESULTS: The model based on T1-CE alone achieved the best segmentation performance for BM segmentation with a median Dice similarity coefficient (DSC) of 0.96. Models trained without T1-CE performed worse (T1-only: DSC = 0.70 and T2-FLAIR-only: DSC = 0.73). For edema segmentation, models that included both T1-CE and T2-FLAIR performed best (DSC = 0.93), while the remaining four models without simultaneous inclusion of these both sequences reached a median DSC of 0.81-0.89. CONCLUSIONS: A T1-CE-only protocol suffices for the segmentation of BMs. The combination of T1-CE and T2-FLAIR is important for edema segmentation. Missing either T1-CE or T2-FLAIR decreases performance. These findings may improve imaging routines by omitting unnecessary sequences, thus allowing for faster procedures in daily clinical practice while enabling optimal neural network-based target definitions.

Salvage radiotherapy is effective in patients with PSMA-PET-negative biochemical recurrence- results of a retrospective study.

BACKGROUND/PURPOSE: The present study aimed to assess whether SRT to the prostatic fossa should be initiated in a timely manner after detecting biochemical recurrence (BR) in patients with prostate cancer, when no correlate was identified with prostate-specific membrane antigen positron emission tomography (PSMA-PET). MATERIALS AND METHODS: This retrospective, multicenter analysis included 1222 patients referred for PSMA-PET after a radical prostatectomy due to BR. Exclusion criteria were: pathological lymph node metastases, prostate-specific antigen (PSA) persistence, distant or lymph node metastases, nodal irradiation, and androgen deprivation therapy (ADT). This led to a cohort of 341 patients. Biochemical progression-free survival (BPFS) was the primary study endpoint. RESULTS: The median follow-up was 28.0 months. The 3-year BPFS was 71.6% in PET-negative cases and 80.8% in locally PET-positive cases. This difference was significant in univariate (p = 0.019), but not multivariate analyses (p = 0.366, HR: 1.46, 95%CI: 0.64-3.32). The 3-year BPFS in PET-negative cases was significantly influenced by age (p = 0.005), initial pT3/4 (p < 0.001), pathology scores (ISUP) ≥ 3 (p = 0.026), and doses to fossa > 70 Gy (p = 0.027) in univariate analyses. In multivariate analyses, only age (HR: 1.096, 95%CI: 1.023-1.175, p = 0.009) and PSA-doubling time (HR: 0.339, 95%CI: 0.139-0.826, p = 0.017) remained significant. CONCLUSION: To our best knowledge, this study provided the largest SRT analysis in patients without ADT that were lymph node-negative on PSMA-PET. A multivariate analysis showed no significant difference in BPFS between locally PET-positive and PET-negative cases. These results supported the current EAU recommendation to initiate SRT in a timely manner after detecting BR in PET negative patients.

Development and Validation of a Multi-institutional Nomogram of Outcomes for PSMA-PET-Based Salvage Radiotherapy for Recurrent Prostate Cancer.

Importance: Prostate-specific antigen membrane positron-emission tomography (PSMA-PET) is increasingly used to guide salvage radiotherapy (sRT) after radical prostatectomy for patients with recurrent or persistent prostate cancer. Objective: To develop and validate a nomogram for prediction of freedom from biochemical failure (FFBF) after PSMA-PET-based sRT. Design, Setting, and Participants: This retrospective cohort study included 1029 patients with prostate cancer treated between July 1, 2013, and June 30, 2020, at 11 centers from 5 countries. The initial database consisted of 1221 patients. All patients had a PSMA-PET scan prior to sRT. Data were analyzed in November 2022. Exposures: Patients with a detectable post-radical prostatectomy prostate-specific antigen (PSA) level treated with sRT to the prostatic fossa with or without additional sRT to pelvic lymphatics or concurrent androgen deprivation therapy (ADT) were eligible. Main Outcomes and Measures: The FFBF rate was estimated, and a predictive nomogram was generated and validated. Biochemical relapse was defined as a PSA nadir of 0.2 ng/mL after sRT. Results: In the nomogram creation and validation process, 1029 patients (median age at sRT, 70 years [IQR, 64-74 years]) were included and further divided into a training set (n = 708), internal validation set (n = 271), and external outlier validation set (n = 50). The median follow-up was 32 months (IQR, 21-45 months). Based on the PSMA-PET scan prior to sRT, 437 patients (42.5%) had local recurrences and 313 patients (30.4%) had nodal recurrences. Pelvic lymphatics were electively irradiated for 395 patients (38.4%). All patients received sRT to the prostatic fossa: 103 (10.0%) received a dose of less than 66 Gy, 551 (53.5%) received a dose of 66 to 70 Gy, and 375 (36.5%) received a dose of more than 70 Gy. Androgen deprivation therapy was given to 325 (31.6%) patients. On multivariable Cox proportional hazards regression analysis, pre-sRT PSA level (hazard ratio [HR], 1.80 [95% CI, 1.41-2.31]), International Society of Urological Pathology grade in surgery specimen (grade 5 vs 1+2: HR, 2.39 [95% CI, 1.63-3.50], pT stage (pT3b+pT4 vs pT2: HR, 1.91 [95% CI, 1.39-2.67]), surgical margins (R0 vs R1+R2+Rx: HR, 0.60 [95% CI, 0.48-0.78]), ADT use (HR, 0.49 [95% CI, 0.37-0.65]), sRT dose (>70 vs ≤66 Gy: HR, 0.44 [95% CI, 0.29-0.67]), and nodal recurrence detected on PSMA-PET scans (HR, 1.42 [95% CI, 1.09-1.85]) were associated with FFBF. The mean (SD) nomogram concordance index for FFBF was 0.72 (0.06) for the internal validation cohort and 0.67 (0.11) in the external outlier validation cohort. Conclusions and Relevance: This cohort study of patients with prostate cancer presents an internally and externally validated nomogram that estimated individual patient outcomes after PSMA-PET-guided sRT.

Development and Evaluation of MR-Based Radiogenomic Models to Differentiate Atypical Lipomatous Tumors from Lipomas.

BACKGROUND: The aim of this study was to develop and validate radiogenomic models to predict the MDM2 gene amplification status and differentiate between ALTs and lipomas on preoperative MR images. METHODS: MR images were obtained in 257 patients diagnosed with ALTs (n = 65) or lipomas (n = 192) using histology and the MDM2 gene analysis as a reference standard. The protocols included T2-, T1-, and fat-suppressed contrast-enhanced T1-weighted sequences. Additionally, 50 patients were obtained from a different hospital for external testing. Radiomic features were selected using mRMR. Using repeated nested cross-validation, the machine-learning models were trained on radiomic features and demographic information. For comparison, the external test set was evaluated by three radiology residents and one attending radiologist. RESULTS: A LASSO classifier trained on radiomic features from all sequences performed best, with an AUC of 0.88, 70% sensitivity, 81% specificity, and 76% accuracy. In comparison, the radiology residents achieved 60-70% accuracy, 55-80% sensitivity, and 63-77% specificity, while the attending radiologist achieved 90% accuracy, 96% sensitivity, and 87% specificity. CONCLUSION: A radiogenomic model combining features from multiple MR sequences showed the best performance in predicting the MDM2 gene amplification status. The model showed a higher accuracy compared to the radiology residents, though lower compared to the attending radiologist.

Outcome of patients with soft tissue sarcomas of the extremities and trunk treated by (neo)adjuvant intensity modulated radiation therapy with curative intent.

BACKGROUND: Soft tissue sarcomas (STS) are a relatively rare group of malignant tumors. Currently, there is very little published clinical data, especially in the context of curative multimodal therapy with image-guided, conformal, intensity-modulated radiotherapy. METHODS: Patients who received preoperative or postoperative intensity-modulated radiotherapy for STS of the extremities or trunk with curative intent were included in this single centre retrospective analysis. A Kaplan-Meier analysis was performed to evaluate survival endpoints. Multivariable proportional hazard models were used to investigate the association between survival endpoints and tumour-, patient-, and treatment-specific characteristics. RESULTS: 86 patients were included in the analysis. The most common histological subtypes were undifferentiated pleomorphic high-grade sarcoma (UPS) (27) and liposarcoma (22). More than two third of the patients received preoperative radiation therapy (72%). During the follow-up period, 39 patients (45%) suffered from some type of relapse, mainly remote (31%). The two-years overall survival rate was 88%. The median DFS was 48 months and the median DMFS was 51 months. Female gender (HR 0.460 (0.217; 0.973)) and histology of liposarcomas compared to UPS proved to be significantly more favorable in terms of DFS (HR 0.327 (0.126; 0.852)). CONCLUSION: Conformal, intensity-modulated radiotherapy is an effective treatment modality in the preoperative or postoperative management of STS. Especially for the prevention of distant metastases, the establishment of modern systemic therapies or multimodal therapy approaches is necessary.

PSMA-PET/CT-guided salvage radiotherapy in recurrent or persistent prostate cancer and PSA < 0.2 ng/ml.

PURPOSE: The purpose of this retrospective, multicenter study was to assess efficacy of PSMA-PET/CT-guided salvage radiotherapy (sRT) in patients with recurrent or persistent PSA after primary surgery and PSA levels < 0.2 ng/ml. METHODS: The study included patients from a pooled cohort (n = 1223) of 11 centers from 6 countries. Patients with PSA levels > 0.2 ng/ml prior to sRT or without sRT to the prostatic fossa were excluded. The primary study endpoint was biochemical recurrence-free survival (BRFS) and BR was defined as PSA nadir after sRT + 0.2 ng/ml. Cox regression analysis was performed to assess the impact of clinical parameters on BRFS. Recurrence patterns after sRT were analyzed. RESULTS: The final cohort consisted of 273 patients; 78/273 (28.6%) and 48/273 (17.6%) patients had local or nodal recurrence on PET/CT. The most frequently applied sRT dose to the prostatic fossa was 66-70 Gy (n = 143/273, 52.4%). SRT to pelvic lymphatics was delivered in 87/273 (31.9%) patients and androgen deprivation therapy was given to 36/273 (13.2%) patients. After a median follow-up time of 31.1 months (IQR: 20-44), 60/273 (22%) patients had biochemical recurrence. The 2- and 3-year BRFS was 90.1% and 79.2%, respectively. The presence of seminal vesicle invasion in surgery (p = 0.019) and local recurrences in PET/CT (p = 0.039) had a significant impact on BR in multivariate analysis. In 16 patients, information on recurrence patterns on PSMA-PET/CT after sRT was available and one had recurrent disease inside the RT field. CONCLUSION: This multicenter analysis suggests that implementation of PSMA-PET/CT imaging for sRT guidance might be of benefit for patients with very low PSA levels after surgery due to promising BRFS rates and a low number of relapses within the sRT field.

Development of PSMA-PET-guided CT-based radiomic signature to predict biochemical recurrence after salvage radiotherapy.

PURPOSE: To develop a CT-based radiomic signature to predict biochemical recurrence (BCR) in prostate cancer patients after sRT guided by positron-emission tomography targeting prostate-specific membrane antigen (PSMA-PET). MATERIAL AND METHODS: Consecutive patients, who underwent 68Ga-PSMA11-PET/CT-guided sRT from three high-volume centers in Germany, were included in this retrospective multicenter study. Patients had PET-positive local recurrences and were treated with intensity-modulated sRT. Radiomic features were extracted from volumes of interests on CT guided by focal PSMA-PET uptakes. After preprocessing, clinical, radiomics, and combined clinical-radiomic models were developed combining different feature reduction techniques and Cox proportional hazard models within a nested cross validation approach. RESULTS: Among 99 patients, median interval until BCR was the radiomic models outperformed clinical models and combined clinical-radiomic models for prediction of BCR with a C-index of 0.71 compared to 0.53 and 0.63 in the test sets, respectively. In contrast to the other models, the radiomic model achieved significantly improved patient stratification in Kaplan-Meier analysis. The radiomic and clinical-radiomic model achieved a significantly better time-dependent net reclassification improvement index (0.392 and 0.762, respectively) compared to the clinical model. Decision curve analysis demonstrated a clinical net benefit for both models. Mean intensity was the most predictive radiomic feature. CONCLUSION: This is the first study to develop a PSMA-PET-guided CT-based radiomic model to predict BCR after sRT. The radiomic models outperformed clinical models and might contribute to guide personalized treatment decisions.

Development and external validation of an MRI-based neural network for brain metastasis segmentation in the AURORA multicenter study.

BACKGROUND: Stereotactic radiotherapy is a standard treatment option for patients with brain metastases. The planning target volume is based on gross tumor volume (GTV) segmentation. The aim of this work is to develop and validate a neural network for automatic GTV segmentation to accelerate clinical daily routine practice and minimize interobserver variability. METHODS: We analyzed MRIs (T1-weighted sequence ± contrast-enhancement, T2-weighted sequence, and FLAIR sequence) from 348 patients with at least one brain metastasis from different cancer primaries treated in six centers. To generate reference segmentations, all GTVs and the FLAIR hyperintense edematous regions were segmented manually. A 3D-U-Net was trained on a cohort of 260 patients from two centers to segment the GTV and the surrounding FLAIR hyperintense region. During training varying degrees of data augmentation were applied. Model validation was performed using an independent international multicenter test cohort (n = 88) including four centers. RESULTS: Our proposed U-Net reached a mean overall Dice similarity coefficient (DSC) of 0.92 ± 0.08 and a mean individual metastasis-wise DSC of 0.89 ± 0.11 in the external test cohort for GTV segmentation. Data augmentation improved the segmentation performance significantly. Detection of brain metastases was effective with a mean F1-Score of 0.93 ± 0.16. The model performance was stable independent of the center (p = 0.3). There was no correlation between metastasis volume and DSC (Pearson correlation coefficient 0.07). CONCLUSION: Reliable automated segmentation of brain metastases with neural networks is possible and may support radiotherapy planning by providing more objective GTV definitions.

Nearest Neighbor-Based Strategy to Optimize Multi-View Triplet Network for Classification of Small-Sample Medical Imaging Data.

Multi-view classification with limited sample size and data augmentation is a very common machine learning (ML) problem in medicine. With limited data, a triplet network approach for two-stage representation learning has been proposed. However, effective training and verifying the features from the representation network for their suitability in subsequent classifiers are still unsolved problems. Although typical distance-based metrics for the training capture the overall class separability of the features, the performance according to these metrics does not always lead to an optimal classification. Consequently, an exhaustive tuning with all feature-classifier combinations is required to search for the best end result. To overcome this challenge, we developed a novel nearest-neighbor (NN) validation strategy based on the triplet metric. This strategy is supported by a theoretical foundation to provide the best selection of the features with a lower bound of the highest end performance. The proposed strategy is a transparent approach to identify whether to improve the features or the classifier. This avoids the need for repeated tuning. Our evaluations on real-world medical imaging tasks (i.e., radiation therapy delivery error prediction and sarcoma survival prediction) show that our strategy is superior to other common deep representation learning baselines [i.e., autoencoder (AE) and softmax]. The strategy addresses the issue of feature's interpretability which enables more holistic feature creation such that the medical experts can focus on specifying relevant data as opposed to tedious feature engineering.

Incidental nodal irradiation in patients with esophageal cancer undergoing (chemo)radiation with 3D-CRT or VMAT.

The extent of elective nodal irradiation (ENI) in patients undergoing definitive chemoradiotherapy (dCRT) for esophageal squamous cell carcinoma (ESCC) remains unclear. The aim of this dosimetric study was to evaluate the extent of incidental nodal irradiation using modern radiation techniques. A planning target volume (PTV) was generated for 30 patients with node-negative esophageal carcinoma (13 cervical/upper third, 7 middle third, 10 lower third/abdomen). Thereby, no elective nodal irradiation (ENI) was intended. Both three-dimensional conformal radiotherapy (3D-CRT) and volumetric-modulated arc therapy (VMAT) treatment plans (50 Gy in 25 fractions) were calculated for all patients. Fifteen nodal stations were contoured according to the definitions of the AJCC and investigated in regard to dosimetric parameters. Compared to 3D-CRT, VMAT was associated with lower dose distribution to the organs at risk (lower Dmean, V20 and V30 for the lungs and lower Dmean and V30 for the heart). For both techniques, the median Dmean surpassed 40 Gy in 12 of 15 (80%) nodal stations. However, VMAT resulted in significantly lower Dmeans and equivalent uniform doses (EUD) compared to 3D-CRT for eight nodal stations (1L, 2L, 2R, 4L, 7, 8L, 10L, 15), while differences did not reach significance for seven nodal station (1R, 4R, 8U, 8M, 10R, 16). For dCRT of ESCC, the use of VMAT was associated with significantly lower median (incidental) doses to eight of 15 regional lymph node areas compared to 3D-CRT. However, given the small absolute differences, these differences probably do not impair (regional) tumor control rates.

Influence of radiation treatment technique (IMRT vs. 3D-RT) on acute toxicity and prognostic factors for survival for anal cancer.

We compared our institutional experience with intensity-modulated radiotherapy (IMRT) and 3D-conformal radiotherapy (3D-RT) for definitive treatment of primary anal cancer. We performed a single-institution retrospective review of all patients with anal squamous cell carcinoma treated with definitive (chemo) radiotherapy with curative intent from 2004 through 2018. We assessed several prognostic factors in respect to relevant survival endpoints. In addition, acute toxicities were determined and compared between IMRT and 3D-RT patients. This study included 94 patients (58 IMRT, 36 3D-RT). Mean follow up for all patients, for IMRT and 3D-RT patients was 61 months (range 6-176), 46 months (range 6-118), and 85 months (range 6-176), respectively. 5-year overall survival (OS) was 86%, disease-free survival (DFS) was 72%, and colostomy-free survival (CFS) was 75% in the IMRT cohort. In the 3D-RT cohort, OS was 87%, DFS was 71%, and CFS was 81% (all p > 0.05). Male gender and Karnofsky Index (KI) were revealed as independent prognostic factors for 5-year OS (p = 0.017; p = 0.023). UICC stage was an independent prognostic factor for DFS and CFS (p = 0.023; p = 0.042). In addition, the pre-treatment leukocyte count was an independent prognostic factor for CFS (p = 0.042). Acute grade ≥ 3 toxicity was not significantly different between IMRT and 3D-RT patients, but the IMRT cohort had favorable outcomes. This study confirmed IMRT as the primary definitive treatment of anal cancer. With similar survival rates, IMRT had the potential to reduce acute toxicity by sparing organs at risk. Promising prognostic factors such as BMI, KI, and leucocyte and hemoglobin levels should be further investigated.

Integrated radiogenomics analyses allow for subtype classification and improved outcome prognosis of patients with locally advanced HNSCC.

Patients with locally advanced head and neck squamous cell carcinoma (HNSCC) may benefit from personalised treatment, requiring biomarkers that characterize the tumour and predict treatment response. We integrate pre-treatment CT radiomics and whole-transcriptome data from a multicentre retrospective cohort of 206 patients with locally advanced HNSCC treated with primary radiochemotherapy to classify tumour molecular subtypes based on radiomics, develop surrogate radiomics signatures for gene-based signatures related to different biological tumour characteristics and evaluate the potential of combining radiomics features with full-transcriptome data for the prediction of loco-regional control (LRC). Using end-to-end machine-learning, we developed and validated a model to classify tumours of the atypical subtype (AUC [95% confidence interval] 0.69 [0.53-0.83]) based on CT imaging, observed that CT-based radiomics models have limited value as surrogates for six selected gene signatures (AUC < 0.60), and showed that combining a radiomics signature with a transcriptomics signature consisting of two metagenes representing the hedgehog pathway and E2F transcriptional targets improves the prognostic value for LRC compared to both individual sources (validation C-index [95% confidence interval], combined: 0.63 [0.55-0.73] vs radiomics: 0.60 [0.50-0.71] and transcriptomics: 0.59 [0.49-0.69]). These results underline the potential of multi-omics analyses to generate reliable biomarkers for future application in personalized oncology.

Self-Supervised Pretext Tasks in Model Robustness & Generalizability: A Revisit from Medical Imaging Perspective.

Self-supervised pretext tasks have been introduced as an effective strategy when learning target tasks on small annotated data sets. However, while current research focuses on exploring novel pretext tasks for meaningful and reusable representation learning for the target task, the study of its robustness and generalizability has remained relatively under-explored. Specifically, it is crucial in medical imaging to proactively investigate performance under different perturbations for reliable deployment of clinical applications. In this work, we revisit medical imaging networks pre-trained with self-supervised learnings and categorically evaluate robustness and generalizability compared to vanilla supervised learning. Our experiments on pneumonia detection in X-rays and multi-organ segmentation in CT yield conclusive results exposing the hidden benefits of self-supervision pre-training for learning robust feature representations.