Radiomics workflow definition & challenges - German priority program 2177 consensus statement on clinically applied radiomics.

OBJECTIVES: Achieving a consensus on a definition for different aspects of radiomics workflows to support their translation into clinical usage. Furthermore, to assess the perspective of experts on important challenges for a successful clinical workflow implementation. MATERIALS AND METHODS: The consensus was achieved by a multi-stage process. Stage 1 comprised a definition screening, a retrospective analysis with semantic mapping of terms found in 22 workflow definitions, and the compilation of an initial baseline definition. Stages 2 and 3 consisted of a Delphi process with over 45 experts hailing from sites participating in the German Research Foundation (DFG) Priority Program 2177. Stage 2 aimed to achieve a broad consensus for a definition proposal, while stage 3 identified the importance of translational challenges. RESULTS: Workflow definitions from 22 publications (published 2012-2020) were analyzed. Sixty-nine definition terms were extracted, mapped, and semantic ambiguities (e.g., homonymous and synonymous terms) were identified and resolved. The consensus definition was developed via a Delphi process. The final definition comprising seven phases and 37 aspects reached a high overall consensus (> 89% of experts "agree" or "strongly agree"). Two aspects reached no strong consensus. In addition, the Delphi process identified and characterized from the participating experts' perspective the ten most important challenges in radiomics workflows. CONCLUSION: To overcome semantic inconsistencies between existing definitions and offer a well-defined, broad, referenceable terminology, a consensus workflow definition for radiomics-based setups and a terms mapping to existing literature was compiled. Moreover, the most relevant challenges towards clinical application were characterized. CRITICAL RELEVANCE STATEMENT: Lack of standardization represents one major obstacle to successful clinical translation of radiomics. Here, we report a consensus workflow definition on different aspects of radiomics studies and highlight important challenges to advance the clinical adoption of radiomics. KEY POINTS: Published radiomics workflow terminologies are inconsistent, hindering standardization and translation. A consensus radiomics workflow definition proposal with high agreement was developed. Publicly available result resources for further exploitation by the scientific community.

Language model-based labeling of German thoracic radiology reports.

The aim of this study was to explore the potential of weak supervision in a deep learning-based label prediction model. The goal was to use this model to extract labels from German free-text thoracic radiology reports on chest X-ray images and for training chest X-ray classification models.The proposed label extraction model for German thoracic radiology reports uses a German BERT encoder as a backbone and classifies a report based on the CheXpert labels. For investigating the efficient use of manually annotated data, the model was trained using manual annotations, weak rule-based labels, and both. Rule-based labels were extracted from 66071 retrospectively collected radiology reports from 2017-2021 (DS 0), and 1091 reports from 2020-2021 (DS 1) were manually labeled according to the CheXpert classes. Label extraction performance was evaluated with respect to mention extraction, negation detection, and uncertainty detection by measuring F1 scores. The influence of the label extraction method on chest X-ray classification was evaluated on a pneumothorax data set (DS 2) containing 6434 chest radiographs with associated reports and expert diagnoses of pneumothorax. For this, DenseNet-121 models trained on manual annotations, rule-based and deep learning-based label predictions, and publicly available data were compared.The proposed deep learning-based labeler (DL) performed on average considerably stronger than the rule-based labeler (RB) for all three tasks on DS 1 with F1 scores of 0.938 vs. 0.844 for mention extraction, 0.891 vs. 0.821 for negation detection, and 0.624 vs. 0.518 for uncertainty detection. Pre-training on DS 0 and fine-tuning on DS 1 performed better than only training on either DS 0 or DS 1. Chest X-ray pneumothorax classification results (DS 2) were highest when trained with DL labels with an area under the receiver operating curve (AUC) of 0.939 compared to RB labels with an AUC of 0.858. Training with manual labels performed slightly worse than training with DL labels with an AUC of 0.934. In contrast, training with a public data set resulted in an AUC of 0.720.Our results show that leveraging a rule-based report labeler for weak supervision leads to improved labeling performance. The pneumothorax classification results demonstrate that our proposed deep learning-based labeler can serve as a substitute for manual labeling requiring only 1000 manually annotated reports for training. · The proposed deep learning-based label extraction model for German thoracic radiology reports performs better than the rule-based model.. · Training with limited supervision outperformed training with a small manually labeled data set.. · Using predicted labels for pneumothorax classification from chest radiographs performed equally to using manual annotations.. Wollek A, Haitzer P, Sedlmeyr T et al. Language modelbased labeling of German thoracic radiology reports. Fortschr Röntgenstr 2024; DOI 10.1055/a-2287-5054.

German CheXpert Chest X-ray Radiology Report Labeler.

PURPOSE: The aim of this study was to develop an algorithm to automatically extract annotations from German thoracic radiology reports to train deep learning-based chest X-ray classification models. MATERIALS AND METHODS: An automatic label extraction model for German thoracic radiology reports was designed based on the CheXpert architecture. The algorithm can extract labels for twelve common chest pathologies, the presence of support devices, and "no finding". For iterative improvements and to generate a ground truth, a web-based multi-reader annotation interface was created. With the proposed annotation interface, a radiologist annotated 1086 retrospectively collected radiology reports from 2020-2021 (data set 1). The effect of automatically extracted labels on chest radiograph classification performance was evaluated on an additional, in-house pneumothorax data set (data set 2), containing 6434 chest radiographs with corresponding reports, by comparing a DenseNet-121 model trained on extracted labels from the associated reports, image-based pneumothorax labels, and publicly available data, respectively. RESULTS: Comparing automated to manual labeling on data set 1: "mention extraction" class-wise F1 scores ranged from 0.8 to 0.995, the "negation detection" F1 scores from 0.624 to 0.981, and F1 scores for "uncertainty detection" from 0.353 to 0.725. Extracted pneumothorax labels on data set 2 had a sensitivity of 0.997 [95 % CI: 0.994, 0.999] and specificity of 0.991 [95 % CI: 0.988, 0.994]. The model trained on publicly available data achieved an area under the receiver operating curve (AUC) for pneumothorax classification of 0.728 [95 % CI: 0.694, 0.760], while the models trained on automatically extracted labels and on manual annotations achieved values of 0.858 [95 % CI: 0.832, 0.882] and 0.934 [95 % CI: 0.918, 0.949], respectively. CONCLUSION: Automatic label extraction from German thoracic radiology reports is a promising substitute for manual labeling. By reducing the time required for data annotation, larger training data sets can be created, resulting in improved overall modeling performance. Our results demonstrated that a pneumothorax classifier trained on automatically extracted labels strongly outperformed the model trained on publicly available data, without the need for additional annotation time and performed competitively compared to manually labeled data. KEY POINTS: · An algorithm for automatic German thoracic radiology report annotation was developed.. · Automatic label extraction is a promising substitute for manual labeling.. · The classifier trained on extracted labels outperformed the model trained on publicly available data..

Radiological age assessment based on clavicle ossification in CT: enhanced accuracy through deep learning.

BACKGROUND: Radiological age assessment using reference studies is inherently limited in accuracy due to a finite number of assignable skeletal maturation stages. To overcome this limitation, we present a deep learning approach for continuous age assessment based on clavicle ossification in computed tomography (CT). METHODS: Thoracic CT scans were retrospectively collected from the picture archiving and communication system. Individuals aged 15.0 to 30.0 years examined in routine clinical practice were included. All scans were automatically cropped around the medial clavicular epiphyseal cartilages. A deep learning model was trained to predict a person's chronological age based on these scans. Performance was evaluated using mean absolute error (MAE). Model performance was compared to an optimistic human reader performance estimate for an established reference study method. RESULTS: The deep learning model was trained on 4,400 scans of 1,935 patients (training set: mean age = 24.2 years ± 4.0, 1132 female) and evaluated on 300 scans of 300 patients with a balanced age and sex distribution (test set: mean age = 22.5 years ± 4.4, 150 female). Model MAE was 1.65 years, and the highest absolute error was 6.40 years for females and 7.32 years for males. However, performance could be attributed to norm-variants or pathologic disorders. Human reader estimate MAE was 1.84 years and the highest absolute error was 3.40 years for females and 3.78 years for males. CONCLUSIONS: We present a deep learning approach for continuous age predictions using CT volumes highlighting the medial clavicular epiphyseal cartilage with performance comparable to the human reader estimate.

Out-of-distribution detection with in-distribution voting using the medical example of chest x-ray classification.

BACKGROUND: Deep learning models are being applied to more and more use cases with astonishing success stories, but how do they perform in the real world? Models are typically tested on specific cleaned data sets, but when deployed in the real world, the model will encounter unexpected, out-of-distribution (OOD) data. PURPOSE: To investigate the impact of OOD radiographs on existing chest x-ray classification models and to increase their robustness against OOD data. METHODS: The study employed the commonly used chest x-ray classification model, CheXnet, trained on the chest x-ray 14 data set, and tested its robustness against OOD data using three public radiography data sets: IRMA, Bone Age, and MURA, and the ImageNet data set. To detect OOD data for multi-label classification, we proposed in-distribution voting (IDV). The OOD detection performance is measured across data sets using the area under the receiver operating characteristic curve (AUC) analysis and compared with Mahalanobis-based OOD detection, MaxLogit, MaxEnergy, self-supervised OOD detection (SS OOD), and CutMix. RESULTS: Without additional OOD detection, the chest x-ray classifier failed to discard any OOD images, with an AUC of 0.5. The proposed IDV approach trained on ID (chest x-ray 14) and OOD data (IRMA and ImageNet) achieved, on average, 0.999 OOD AUC across the three data sets, surpassing all other OOD detection methods. Mahalanobis-based OOD detection achieved an average OOD detection AUC of 0.982. IDV trained solely with a few thousand ImageNet images had an AUC 0.913, which was considerably higher than MaxLogit (0.726), MaxEnergy (0.724), SS OOD (0.476), and CutMix (0.376). CONCLUSIONS: The performance of all tested OOD detection methods did not translate well to radiography data sets, except Mahalanobis-based OOD detection and the proposed IDV method. Consequently, training solely on ID data led to incorrect classification of OOD images as ID, resulting in increased false positive rates. IDV substantially improved the model's ID classification performance, even when trained with data that will not occur in the intended use case or test set (ImageNet), without additional inference overhead or performance decrease in the target classification. The corresponding code is available at https://gitlab.lrz.de/IP/a-knee-cannot-have-lung-disease.

Imaging-Derived Biomarkers Integrated with Clinical and Laboratory Values Predict Recurrence of Hepatocellular Carcinoma After Liver Transplantation.

Purpose: To investigate the prognostic value of computed tomography (CT) derived imaging biomarkers in hepatocellular carcinoma (HCC) recurrence after liver transplantation (LT) and develop a predictive nomogram model. Patients and Methods: This retrospective study included 178 patients with histopathologically confirmed HCC who underwent liver transplantation between 2007 and 2021 at the two academic liver centers. We evaluated dedicated imaging features from baseline multiphase contrast-enhanced CT supplemented by several clinical findings and laboratory parameters. Time-to-recurrence was estimated by Kaplan-Meier analysis. Univariable Cox proportional hazard regression and multivariable Least Absolute Shrinkage and Selection Operator (LASSO) regression were used to assess independent prognostic factors for recurrence. A nomogram model was then built based on the independent factors selected through LASSO regression, to predict the probabilities of HCC recurrence at one, three, and five years. Results: The rate of HCC recurrence after LT was 17.4% (31 of 178). The LASSO analysis revealed six independent predictors associated with an elevated risk of tumor recurrence. These predictors included the presence of peritumoral enhancement, the presence of over three tumor lesions, the largest tumor diameter greater than 3 cm, serum alpha-fetoprotein (AFP) levels exceeding 400 ng/mL, and the presence of a tumor capsule. Conversely, a history of bridging therapies was found to be correlated with a reduced risk of HCC recurrence. In addition, Kaplan-Meier curves showed patients with irregular margin, satellite nodules, or small lesions displayed shorter time-to-recurrence. Our nomogram demonstrated good performance, yielding a C-index of 0.835 and AUC values of 0.86, 0.88, and 0.85 for the predictions of 1-year, 3-year, and 5-year TTR, respectively. Conclusion: Imaging parameters derived from baseline contrast-enhanced CT showing malignant behavior and aggressive growth patterns, along with serum AFP and history of bridging therapies, show potential as biomarkers for predicting HCC recurrence after transplantation.

WindowNet: Learnable Windows for Chest X-ray Classification.

Public chest X-ray (CXR) data sets are commonly compressed to a lower bit depth to reduce their size, potentially hiding subtle diagnostic features. In contrast, radiologists apply a windowing operation to the uncompressed image to enhance such subtle features. While it has been shown that windowing improves classification performance on computed tomography (CT) images, the impact of such an operation on CXR classification performance remains unclear. In this study, we show that windowing strongly improves the CXR classification performance of machine learning models and propose WindowNet, a model that learns multiple optimal window settings. Our model achieved an average AUC score of 0.812 compared with the 0.759 score of a commonly used architecture without windowing capabilities on the MIMIC data set.

ChatGPT makes medicine easy to swallow: an exploratory case study on simplified radiology reports.

OBJECTIVES: To assess the quality of simplified radiology reports generated with the large language model (LLM) ChatGPT and to discuss challenges and chances of ChatGPT-like LLMs for medical text simplification. METHODS: In this exploratory case study, a radiologist created three fictitious radiology reports which we simplified by prompting ChatGPT with "Explain this medical report to a child using simple language." In a questionnaire, we tasked 15 radiologists to rate the quality of the simplified radiology reports with respect to their factual correctness, completeness, and potential harm for patients. We used Likert scale analysis and inductive free-text categorization to assess the quality of the simplified reports. RESULTS: Most radiologists agreed that the simplified reports were factually correct, complete, and not potentially harmful to the patient. Nevertheless, instances of incorrect statements, missed relevant medical information, and potentially harmful passages were reported. CONCLUSION: While we see a need for further adaption to the medical field, the initial insights of this study indicate a tremendous potential in using LLMs like ChatGPT to improve patient-centered care in radiology and other medical domains. CLINICAL RELEVANCE STATEMENT: Patients have started to use ChatGPT to simplify and explain their medical reports, which is expected to affect patient-doctor interaction. This phenomenon raises several opportunities and challenges for clinical routine. KEY POINTS: • Patients have started to use ChatGPT to simplify their medical reports, but their quality was unknown. • In a questionnaire, most participating radiologists overall asserted good quality to radiology reports simplified with ChatGPT. However, they also highlighted a notable presence of errors, potentially leading patients to draw harmful conclusions. • Large language models such as ChatGPT have vast potential to enhance patient-centered care in radiology and other medical domains. To realize this potential while minimizing harm, they need supervision by medical experts and adaption to the medical field.

Artificial Intelligence in Oncological Hybrid Imaging.

BACKGROUND: Artificial intelligence (AI) applications have become increasingly relevant across a broad spectrum of settings in medical imaging. Due to the large amount of imaging data that is generated in oncological hybrid imaging, AI applications are desirable for lesion detection and characterization in primary staging, therapy monitoring, and recurrence detection. Given the rapid developments in machine learning (ML) and deep learning (DL) methods, the role of AI will have significant impact on the imaging workflow and will eventually improve clinical decision making and outcomes. METHODS AND RESULTS: The first part of this narrative review discusses current research with an introduction to artificial intelligence in oncological hybrid imaging and key concepts in data science. The second part reviews relevant examples with a focus on applications in oncology as well as discussion of challenges and current limitations. CONCLUSION: AI applications have the potential to leverage the diagnostic data stream with high efficiency and depth to facilitate automated lesion detection, characterization, and therapy monitoring to ultimately improve quality and efficiency throughout the medical imaging workflow. The goal is to generate reproducible, structured, quantitative diagnostic data for evidence-based therapy guidance in oncology. However, significant challenges remain regarding application development, benchmarking, and clinical implementation. KEY POINTS: · Hybrid imaging generates a large amount of multimodality medical imaging data with high complexity and depth.. · Advanced tools are required to enable fast and cost-efficient processing along the whole radiology value chain.. · AI applications promise to facilitate the assessment of oncological disease in hybrid imaging with high quality and efficiency for lesion detection, characterization, and response assessment. The goal is to generate reproducible, structured, quantitative diagnostic data for evidence-based oncological therapy guidance.. · Selected applications in three oncological entities (lung, prostate, and neuroendocrine tumors) demonstrate how AI algorithms may impact imaging-based tasks in hybrid imaging and potentially guide clinical decision making..

Integration of clinical parameters and CT-based radiomics improves machine learning assisted subtyping of primary hyperaldosteronism.

Objectives: The aim of this study was to investigate an integrated diagnostics approach for prediction of the source of aldosterone overproduction in primary hyperaldosteronism (PA). Methods: 269 patients from the prospective German Conn Registry with PA were included in this study. After segmentation of adrenal glands in native CT images, radiomic features were calculated. The study population consisted of a training (n = 215) and a validation (n = 54) cohort. The k = 25 best radiomic features, selected using maximum-relevance minimum-redundancy (MRMR) feature selection, were used to train a baseline random forest model to predict the result of AVS from imaging alone. In a second step, clinical parameters were integrated. Model performance was assessed via area under the receiver operating characteristic curve (ROC AUC). Permutation feature importance was used to assess the predictive value of selected features. Results: Radiomics features alone allowed only for moderate discrimination of the location of aldosterone overproduction with a ROC AUC of 0.57 for unilateral left (UL), 0.61 for unilateral right (UR), and 0.50 for bilateral (BI) aldosterone overproduction (total 0.56, 95% CI: 0.45-0.65). Integration of clinical parameters into the model substantially improved ROC AUC values (0.61 UL, 0.68 UR, and 0.73 for BI, total 0.67, 95% CI: 0.57-0.77). According to permutation feature importance, lowest potassium value at baseline and saline infusion test (SIT) were the two most important features. Conclusion: Integration of clinical parameters into a radiomics machine learning model improves prediction of the source of aldosterone overproduction and subtyping in patients with PA.

A Comprehensive Machine Learning Benchmark Study for Radiomics-Based Survival Analysis of CT Imaging Data in Patients With Hepatic Metastases of CRC.

OBJECTIVES: Optimizing a machine learning (ML) pipeline for radiomics analysis involves numerous choices in data set composition, preprocessing, and model selection. Objective identification of the optimal setup is complicated by correlated features, interdependency structures, and a multitude of available ML algorithms. Therefore, we present a radiomics-based benchmarking framework to optimize a comprehensive ML pipeline for the prediction of overall survival. This study is conducted on an image set of patients with hepatic metastases of colorectal cancer, for which radiomics features of the whole liver and of metastases from computed tomography images were calculated. A mixed model approach was used to find the optimal pipeline configuration and to identify the added prognostic value of radiomics features. MATERIALS AND METHODS: In this study, a large-scale ML benchmark pipeline consisting of preprocessing, feature selection, dimensionality reduction, hyperparameter optimization, and training of different models was developed for radiomics-based survival analysis. Portal-venous computed tomography imaging data from a previous prospective randomized trial evaluating radioembolization of liver metastases of colorectal cancer were quantitatively accessible through a radiomics approach. One thousand two hundred eighteen radiomics features of hepatic metastases and the whole liver were calculated, and 19 clinical parameters (age, sex, laboratory values, and treatment) were available for each patient. Three ML algorithms-a regression model with elastic net regularization (glmnet), a random survival forest (RSF), and a gradient tree-boosting technique (xgboost)-were evaluated for 5 combinations of clinical data, tumor radiomics, and whole-liver features. Hyperparameter optimization and model evaluation were optimized toward the performance metric integrated Brier score via nested cross-validation. To address dependency structures in the benchmark setup, a mixed-model approach was developed to compare ML and data configurations and to identify the best-performing model. RESULTS: Within our radiomics-based benchmark experiment, 60 ML pipeline variations were evaluated on clinical data and radiomics features from 491 patients. Descriptive analysis of the benchmark results showed a preference for RSF-based pipelines, especially for the combination of clinical data with radiomics features. This observation was supported by the quantitative analysis via a linear mixed model approach, computed to differentiate the effect of data sets and pipeline configurations on the resulting performance. This revealed the RSF pipelines to consistently perform similar or better than glmnet and xgboost. Further, for the RSF, there was no significantly better-performing pipeline composition regarding the sort of preprocessing or hyperparameter optimization. CONCLUSIONS: Our study introduces a benchmark framework for radiomics-based survival analysis, aimed at identifying the optimal settings with respect to different radiomics data sources and various ML pipeline variations, including preprocessing techniques and learning algorithms. A suitable analysis tool for the benchmark results is provided via a mixed model approach, which showed for our study on patients with intrahepatic liver metastases, that radiomics features captured the patients' clinical situation in a manner comparable to the provided information solely from clinical parameters. However, we did not observe a relevant additional prognostic value obtained by these radiomics features.

Patterns of pseudoprogression across different cancer entities treated with immune checkpoint inhibitors.

BACKGROUND: Pseudoprogression (PsPD) is a rare response pattern to immune checkpoint inhibitor (ICI) therapy in oncology. This study aims to reveal imaging features of PsPD, and their association to other relevant findings. METHODS: Patients with PsPD who had at least three consecutive cross-sectional imaging studies at our comprehensive cancer center were retrospectively analyzed. Treatment response was assessed according to immune Response Evaluation Criteria in Solid Tumors (iRECIST). PsPD was defined as the occurrence of immune unconfirmed progressive disease (iUPD) without follow-up confirmation. Target lesions (TL), non-target lesions (NTL), new lesions (NL) were analyzed over time. Tumor markers and immune-related adverse events (irAE) were correlated. RESULTS: Thirty-two patients were included (mean age: 66.7 ± 13.6 years, 21.9% female) with mean baseline STL of 69.7 mm ± 55.6 mm. PsPD was observed in twenty-six patients (81.3%) at FU1, and no cases occurred after FU4. Patients with iUPD exhibited the following: TL increase in twelve patients, (37.5%), NTL increase in seven patients (21.9%), NL appearance in six patients (18.8%), and combinations thereof in four patients (12.5%). The mean and maximum increase for first iUPD in sum of TL was 19.8 and 96.8 mm (+ 700.8%). The mean and maximum decrease in sum of TL between iUPD and consecutive follow-up was - 19.1 mm and - 114.8 mm (-60.9%) respectively. The mean and maximum sum of new TL at first iUPD timepoint were 7.6 and 82.0 mm respectively. In two patients (10.5%), tumor-specific serologic markers were elevated at first iUPD, while the rest were stable or decreased among the other PsPD cases (89.5%). In fourteen patients (43.8%), irAE were observed. CONCLUSIONS: PsPD occurred most frequently at FU1 after initiation of ICI treatment. The two most prevalent reasons for PsPD were TL und NTL progression, with an increase in TL diameter commonly below + 100%. In few cases, PsPD was observed even if tumor markers were rising compared to baseline. Our findings also suggest a correlation between PsPD and irAE. These findings may guide decision-making of ICI continuation in suspected PsPD.

Blood-brain barrier leakage hotspots collocating with brain lesions due to sporadic and monogenic small vessel disease.

Blood-brain barrier (BBB) is known to be impaired in cerebral small vessel disease (SVD), and is measurable by dynamic-contrast enhancement (DCE)-MRI. In a cohort of 69 patients (42 sporadic, 27 monogenic SVD), who underwent 3T MRI, including DCE and cerebrovascular reactivity (CVR) sequences, we assessed the relationship of BBB-leakage hotspots to SVD lesions (lacunes, white matter hyperintensities (WMH), and microbleeds). We defined as hotspots the regions with permeability surface area product highest decile on DCE-derived maps within the white matter. We assessed factors associated with the presence and number of hotspots corresponding to SVD lesions in multivariable regression models adjusted for age, WMH volume, number of lacunes, and SVD type. We identified hotspots at lacune edges in 29/46 (63%) patients with lacunes, within WMH in 26/60 (43%) and at the WMH edges in 34/60 (57%) patients with WMH, and microbleed edges in 4/11 (36%) patients with microbleeds. In adjusted analysis, lower WMH-CVR was associated with presence and number of hotspots at lacune edges, and higher WMH volume with hotspots within WMH and at WMH edges, independently of the SVD type. In conclusion, SVD lesions frequently collocate with high BBB-leakage in patients with sporadic and monogenic forms of SVD.

Automated localization of the medial clavicular epiphyseal cartilages using an object detection network: a step towards deep learning-based forensic age assessment.

BACKGROUND: Deep learning is a promising technique to improve radiological age assessment. However, expensive manual annotation by experts poses a bottleneck for creating large datasets to appropriately train deep neural networks. We propose an object detection approach to automatically annotate the medial clavicular epiphyseal cartilages in computed tomography (CT) scans. METHODS: The sternoclavicular joints were selected as structure-of-interest (SOI) in chest CT scans and served as an easy-to-identify proxy for the actual medial clavicular epiphyseal cartilages. CT slices containing the SOI were manually annotated with bounding boxes around the SOI. All slices in the training set were used to train the object detection network RetinaNet. Afterwards, the network was applied individually to all slices of the test scans for SOI detection. Bounding box and slice position of the detection with the highest classification score were used as the location estimate for the medial clavicular epiphyseal cartilages inside the CT scan. RESULTS: From 100 CT scans of 82 patients, 29,656 slices were used for training and 30,846 slices from 110 CT scans of 110 different patients for testing the object detection network. The location estimate from the deep learning approach for the SOI was in a correct slice in 97/110 (88%), misplaced by one slice in 5/110 (5%), and missing in 8/110 (7%) test scans. No estimate was misplaced by more than one slice. CONCLUSIONS: We demonstrated a robust automated approach for annotating the medial clavicular epiphyseal cartilages. This enables training and testing of deep neural networks for age assessment.

Integrated intravoxel incoherent motion tensor and diffusion tensor brain MRI in a single fast acquisition.

The acquisition of intravoxel incoherent motion (IVIM) data and diffusion tensor imaging (DTI) data from the brain can be integrated into a single measurement, which offers the possibility to determine orientation-dependent (tensorial) perfusion parameters in addition to established IVIM and DTI parameters. The purpose of this study was to evaluate the feasibility of such a protocol with a clinically feasible scan time below 6 min and to use a model-selection approach to find a set of DTI and IVIM tensor parameters that most adequately describes the acquired data. Diffusion-weighted images of the brain were acquired at 3 T in 20 elderly participants with cerebral small vessel disease using a multiband echoplanar imaging sequence with 15 b-values between 0 and 1000 s/mm2 and six non-collinear diffusion gradient directions for each b-value. Seven different IVIM-diffusion models with 4 to 14 parameters were implemented, which modeled diffusion and pseudo-diffusion as scalar or tensor quantities. The models were compared with respect to their fitting performance based on the goodness of fit (sum of squared fit residuals, chi2 ) and their Akaike weights (calculated from the corrected Akaike information criterion). Lowest chi2 values were found using the model with the largest number of model parameters. However, significantly highest Akaike weights indicating the most appropriate models for the acquired data were found with a nine-parameter IVIM-DTI model (with isotropic perfusion modeling) in normal-appearing white matter (NAWM), and with an 11-parameter model (IVIM-DTI with additional pseudo-diffusion anisotropy) in white matter with hyperintensities (WMH) and in gray matter (GM). The latter model allowed for the additional calculation of the fractional anisotropy of the pseudo-diffusion tensor (with a median value of 0.45 in NAWM, 0.23 in WMH, and 0.36 in GM), which is not accessible with the usually performed IVIM acquisitions based on three orthogonal diffusion-gradient directions.

Artificial Intelligence in Oncological Hybrid Imaging.

BACKGROUND: Artificial intelligence (AI) applications have become increasingly relevant across a broad spectrum of settings in medical imaging. Due to the large amount of imaging data that is generated in oncological hybrid imaging, AI applications are desirable for lesion detection and characterization in primary staging, therapy monitoring, and recurrence detection. Given the rapid developments in machine learning (ML) and deep learning (DL) methods, the role of AI will have significant impact on the imaging workflow and will eventually improve clinical decision making and outcomes. METHODS AND RESULTS: The first part of this narrative review discusses current research with an introduction to artificial intelligence in oncological hybrid imaging and key concepts in data science. The second part reviews relevant examples with a focus on applications in oncology as well as discussion of challenges and current limitations. CONCLUSION: AI applications have the potential to leverage the diagnostic data stream with high efficiency and depth to facilitate automated lesion detection, characterization, and therapy monitoring to ultimately improve quality and efficiency throughout the medical imaging workflow. The goal is to generate reproducible, structured, quantitative diagnostic data for evidence-based therapy guidance in oncology. However, significant challenges remain regarding application development, benchmarking, and clinical implementation. KEY POINTS: · Hybrid imaging generates a large amount of multimodality medical imaging data with high complexity and depth.. · Advanced tools are required to enable fast and cost-efficient processing along the whole radiology value chain.. · AI applications promise to facilitate the assessment of oncological disease in hybrid imaging with high quality and efficiency for lesion detection, characterization, and response assessment. The goal is to generate reproducible, structured, quantitative diagnostic data for evidence-based oncological therapy guidance.. · Selected applications in three oncological entities (lung, prostate, and neuroendocrine tumors) demonstrate how AI algorithms may impact imaging-based tasks in hybrid imaging and potentially guide clinical decision making.. CITATION FORMAT: · Feuerecker B, Heimer M, Geyer T et al. Artificial Intelligence in Oncological Hybrid Imaging. Fortschr Röntgenstr 2023; 195: 105 - 114.

Radiomics Signature Using Manual Versus Automated Segmentation for Lymph Node Staging of Bladder Cancer.

BACKGROUND: Bladder cancer (BC) treatment algorithms depend on accurate tumor staging. To date, computed tomography (CT) is recommended for assessment of lymph node (LN) metastatic spread in muscle-invasive and high-risk BC. However, the diagnostic efficacy of radiologist-evaluated CT imaging studies is limited. OBJECTIVE: To evaluate the performance of quantitative radiomics signatures for detection of LN metastases in BC. DESIGN, SETTING, AND PARTICIPANTS: Of 1354 patients with BC who underwent radical cystectomy (RC) with lymphadenectomy who were screened, 391 with pathological nodal staging (pN0: n = 297; pN+: n = 94) were included and randomized into training (n = 274) and test (n = 117) cohorts. Pelvic LNs were segmented manually and automatically. A total of 1004 radiomics features were extracted from each LN and a machine learning model was trained to assess pN status using histopathology labels as the ground truth. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: Radiologist assessment was compared to radiomics-based analysis using manual and automated LN segmentations for detection of LN metastases in BC. Statistical analysis was performed using the receiver operating characteristics curve method and evaluated in terms of sensitivity, specificity, and area under the curve (AUC). RESULTS AND LIMITATIONS: In total, 1845 LNs were manually segmented. Automated segmentation correctly located 361/557 LNs in the test cohort. Manual and automatic masks achieved an AUC of 0.80 (95% confidence interval [CI] 0.69-0.91; p = 0.64) and 0.70 (95% CI: 0.58-0.82; p = 0.17), respectively, in the test cohort compared to radiologist assessment, with an AUC of 0.78 (95% CI 0.67-0.89). A combined model of a manually segmented radiomics signature and radiologist assessment reached an AUC of 0.81 (95% CI 0.71-0.92; p = 0.63). CONCLUSIONS: A radiomics signature allowed discrimination of nodal status with high diagnostic accuracy. The model based on manual LN segmentation outperformed the fully automated approach. PATIENT SUMMARY: For patients with bladder cancer, evaluation of computed tomography (CT) scans before surgery using a computer-based method for image analysis, called radiomics, may help in standardizing and improving the accuracy of assessment of lymph nodes. This could be a valuable tool for optimizing treatment options.

18F-FET PET radiomics-based survival prediction in glioblastoma patients receiving radio(chemo)therapy.

BACKGROUND: Quantitative image analysis based on radiomic feature extraction is an emerging field for survival prediction in oncological patients. 18F-Fluorethyltyrosine positron emission tomography (18F-FET PET) provides important diagnostic and grading information for brain tumors, but data on its use in survival prediction is scarce. In this study, we aim at investigating survival prediction based on multiple radiomic features in glioblastoma patients undergoing radio(chemo)therapy. METHODS: A dataset of 37 patients with glioblastoma (WHO grade 4) receiving radio(chemo)therapy was analyzed. Radiomic features were extracted from pre-treatment 18F-FET PET images, following intensity rebinning with a fixed bin width. Principal component analysis (PCA) was applied for variable selection, aiming at the identification of the most relevant features in survival prediction. Random forest classification and prediction algorithms were optimized on an initial set of 25 patients. Testing of the implemented algorithms was carried out in different scenarios, which included additional 12 patients whose images were acquired with a different scanner to check the reproducibility in prediction results. RESULTS: First order intensity variations and shape features were predominant in the selection of most important radiomic signatures for survival prediction in the available dataset. The major axis length of the 18F-FET-PET volume at tumor to background ratio (TBR) 1.4 and 1.6 correlated significantly with reduced probability of survival. Additional radiomic features were identified as potential survival predictors in the PTV region, showing 76% accuracy in independent testing for both classification and regression. CONCLUSIONS: 18F-FET PET prior to radiation provides relevant information for survival prediction in glioblastoma patients. Based on our preliminary analysis, radiomic features in the PTV can be considered a robust dataset for survival prediction.

Performance variability of radiomics machine learning models for the detection of clinically significant prostate cancer in heterogeneous MRI datasets.

Background: Radiomics promises to enhance the discriminative performance for clinically significant prostate cancer (csPCa), but still lacks validation in real-life scenarios. This study investigates the classification performance and robustness of machine learning radiomics models in heterogeneous MRI datasets to characterize suspicious prostate lesions for non-invasive prediction of prostate cancer (PCa) aggressiveness compared to conventional imaging biomarkers. Methods: A total of 142 patients with clinical suspicion of PCa underwent 1.5T or 3T biparametric MRI (7 scanner types, 14 institutions) and exhibited suspicious lesions [prostate Imaging Reporting and Data System (PI-RADS) score ≥3] in peripheral or transitional zones. Whole-gland and index-lesion segmentations were performed semi-automatically. A total of 1,482 quantitative morphologic, shape, texture, and intensity-based radiomics features were extracted from T2-weighted and apparent diffusion coefficient (ADC)-images and assessed using random forest and logistic regression models. Five-fold cross-validation performance in terms of area under the ROC curve was compared to mean ADC (mADC), PI-RADS and prostate-specific antigen density (PSAD). Bias mitigation techniques targeting the high-dimensional feature space and inherent class imbalance were applied and robustness of results was systematically evaluated. Results: Trained models showed mean area under the curves (AUCs) ranging from 0.78 to 0.83 in csPCa classification. Despite using mitigation techniques, high performance variability of results could be demonstrated. Trained models achieved on average numerically higher classification performance compared to clinical parameters PI-RADS (AUC =0.78), mADC (AUC =0.71) and PSAD (AUC =0.63). Conclusions: Radiomics models' classification performance of csPCa was numerically but not significantly higher than PI-RADS scoring. Overall, clinical applicability in heterogeneous MRI datasets is limited because of high variability of results. Performance variability, robustness and reproducibility of radiomics-based measures should be addressed more transparently in future research to enable broad clinical application.

Radiomics for the Prediction of Overall Survival in Patients with Bladder Cancer Prior to Radical Cystectomy.

(1) Background: To evaluate radiomics features as well as a combined model with clinical parameters for predicting overall survival in patients with bladder cancer (BCa). (2) Methods: This retrospective study included 301 BCa patients who received radical cystectomy (RC) and pelvic lymphadenectomy. Radiomics features were extracted from the regions of the primary tumor and pelvic lymph nodes as well as the peritumoral regions in preoperative CT scans. Cross-validation was performed in the training cohort, and a Cox regression model with an elastic net penalty was trained using radiomics features and clinical parameters. The models were evaluated with the time-dependent area under the ROC curve (AUC), Brier score and calibration curves. (3) Results: The median follow-up time was 56 months (95% CI: 48−74 months). In the follow-up period from 1 to 7 years after RC, radiomics models achieved comparable predictive performance to validated clinical parameters with an integrated AUC of 0.771 (95% CI: 0.657−0.869) compared to an integrated AUC of 0.761 (95% CI: 0.617−0.874) for the prediction of overall survival (p = 0.98). A combined clinical and radiomics model stratified patients into high-risk and low-risk groups with significantly different overall survival (p < 0.001). (4) Conclusions: Radiomics features based on preoperative CT scans have prognostic value in predicting overall survival before RC. Therefore, radiomics may guide early clinical decision-making.