Ana Jimenez-Pastor

Background: Automated machine learning (AutoML) frameworks can lower technical barriers for predictive and prognostic model development in radiomics by enabling researchers without programming expertise to build models. However, their effectiveness in addressing radiomics-specific challenges remains unclear. Objective: This study aimed to evaluate the performance, efficiency, and accessibility of general-purpose and radiomics-specific AutoML frameworks on diverse radiomics classification tasks, thereby guiding researchers and highlighting development needs for radiomics. Methods: A total of 10 public and private radiomics datasets with varied imaging modalities (computed tomography and magnetic resonance imaging), sizes, anatomies, and end points were used. Six general-purpose and 5 radiomics-specific frameworks were tested with predefined parameters using standardized cross-validation. Evaluation metrics included area under the receiver operating characteristic curve, runtime, and qualitative aspects related to software status, accessibility, and interpretability. Results: Simplatab, a radiomics-specific tool with a no-code interface, achieved the best overall balance between performance and computational efficiency, recording the highest average test area under the receiver operating characteristic curve (mean 78.46%, SD 12.22%) with a moderate runtime (1.1 h). However, its performance was not statistically superior to the most intensive general-purpose solutions. Most radiomics-specific frameworks were excluded from the performance analysis due to obsolescence, extensive programming requirements, or computational inefficiency. Conversely, general-purpose frameworks demonstrated higher accessibility and ease of implementation. Conclusions: While no single framework demonstrated absolute predictive superiority, Simplatab provides an effective balance of performance, efficiency, and accessibility for radiomics classification problems. However, continued efforts are needed to further mature AutoML solutions in the radiomics domain.

Objectives: The aim of this study was to develop and validate an MRI radiomics-based predictive model to discriminate significant prostate cancer (sPCa), compare it with PI-RADS, and determine whether incorporating PI-RADS and other clinical variables improves clinical performance. Materials and methods: A retrospective observational study was conducted using a cohort of 1497 MRI cases from 1395 men to develop the models. For each case, the index-lesion PI-RADS score, systematic ± targeted biopsy results, and six additional clinical variables were collected. Prostate biopsy samples served as the reference standard, defining sPCa as Gleason Grade ≥ 7. Handcrafted radiomic features were extracted from automatically segmented prostate glands. Four machine learning models were developed: (1) Radiomics, (2) PI-RADS, (3) PI-RADS + Radiomics, and (4) PI-RADS + Radiomics + Clinical Variables. Model performance and comparisons were evaluated using the area under the curve (AUC), while clinical utility was assessed through the decision curve analysis plot, Clinical Utility plot, and the number of avoided biopsies. Results: The radiomics model did not perform significantly better than PI-RADS in the validation cohort (AUC 0.838 vs. 0.833, p = 0.874). The combination of radiomics, PI-RADS, and clinical variables achieved the highest performance, with an AUC of 0.891 (95% CI: 0.853–0.930), significantly outperforming the other models ( p < 0.05). It also showed the highest specificity (29.41%) and biopsy avoidance rate (18.15%), although the differences were not statistically significant ( p = 0.313). Conclusions: Incorporating radiomics and clinical variables into PI-RADS enhances its ability to discriminate sPCa, potentially decreasing false positives and unnecessary biopsies. Critical relevance statement: The incorporation of clinical variables and radiomics into PI-RADS enhances its ability to predict significant prostate cancer, helping mitigate some of PI-RADS’s current limitations, such as a significant false-positive rate, and might help reduce unnecessary biopsies. Key Points: PI-RADS limitations result in overdiagnosis of indolent prostatic lesions and unnecessary biopsies. Radiomics and clinical variables enhance PI-RADS ability to detect significant prostate cancer. Combined clinical-radiological models reduce false positives and help avoid unnecessary biopsies.

Background: Assessment of the prostatic neurovascular bundles on MRI is clinically relevant for staging and treatment planning but remains technically challenging and underexplored in automated imaging pipelines. Purpose: To develop and evaluate an automated framework for neurovascular bundle segmentation, proximity-based invasion risk stratification, and radiomics-based prediction of biochemical recurrence, perineural invasion, and extraprostatic extension. Methods: This retrospective study included 808 prostate MRI examinations from three datasets acquired between 2015 and 2020. Among them, 470 T2-weighted image sequences were manually annotated to train a 3D full-resolution nnU-Net segmentation model. Tumor-to-neurovascular bundle distance was used to define invasion risk categories (low, intermediate, high). Machine learning models were developed using radiomics features extracted from neurovascular bundles, lesions, and combined regions, with optional inclusion of prostate-specific antigen and age at MRI. Model performance was evaluated using area under the receiver operating characteristic curve and accuracy. Model interpretability was assessed using Shapley additive explanations. Results: The median patient age was 69 years (interquartile range, 63–73). Automatic neurovascular bundles segmentation achieved anatomically plausible contours, with average surface distance below 1 mm and volume difference under 0.4 cc. The resulting tumor-to-neurovascular bundle invasion risk classification reached 90% accuracy, supporting usability. Radiomics models showed predictive value across endpoints, with moderate testing performance for biochemical recurrence (AUC = 0.73), and higher discrimination for perineural invasion (AUC = 0.80) and extraprostatic extension (AUC = 0.80). Interpretability analysis revealed that tumor-to-neurovascular bundle proximity and neurovascular bundles imaging features among the most relevant contributors to outcome prediction. Conclusions: Automated neurovascular bundle segmentation enabled quantitative tumor proximity assessment and radiomics-based prediction of biochemical recurrence, perineural invasion, and extraprostatic extension.

BACKGROUND Automated machine learning (AutoML) frameworks can lower technical barriers for predictive and prognostic model development in radiomics by enabling researchers without programming expertise to build models. However, their effectiveness in addressing radiomics-specific challenges remains unclear. OBJECTIVE This study aimed to evaluate the performance, efficiency, and accessibility of general-purpose and radiomics-specific AutoML frameworks on diverse radiomics classification tasks, thereby guiding researchers and highlighting development needs for radiomics. METHODS A total of 10 public and private radiomics datasets with varied imaging modalities (computed tomography and magnetic resonance imaging), sizes, anatomies, and end points were used. Six general-purpose and 5 radiomics-specific frameworks were tested with predefined parameters using standardized cross-validation. Evaluation metrics included area under the receiver operating characteristic curve, runtime, and qualitative aspects related to software status, accessibility, and interpretability. RESULTS Simplatab, a radiomics-specific tool with a no-code interface, achieved the best overall balance between performance and computational efficiency, recording the highest average test area under the receiver operating characteristic curve (mean 78.46%, SD 12.22%) with a moderate runtime (1.1 h). However, its performance was not statistically superior to the most intensive general-purpose solutions. Most radiomics-specific frameworks were excluded from the performance analysis due to obsolescence, extensive programming requirements, or computational inefficiency. Conversely, general-purpose frameworks demonstrated higher accessibility and ease of implementation. CONCLUSIONS While no single framework demonstrated absolute predictive superiority, Simplatab provides an effective balance of performance, efficiency, and accessibility for radiomics classification problems. However, continued efforts are needed to further mature AutoML solutions in the radiomics domain.

Objectives: This study investigated the influence of hepatic vessels on the quantification of magnetic resonance imaging (MRI) proton density fat fraction (PDFF) and R2* using automated whole-liver segmentation. Materials and methods: This prospective multicenter study included patients with chronic liver disease having paired liver biopsy and MR exams with a standardized multiecho chemical-shift gradient echo sequence. Automated whole-liver segmentation was performed, generating two masks per patient, one including and the other excluding the major hepatic vessels. PDFF and R2* were quantified and graded for both masks. Histological grading of hepatic steatosis and iron overload severity was used as a reference standard. Results: A total of 377 patients were evaluated, of whom 54% had hepatic steatosis and 20% had iron overload on biopsy readings. Stratified by histological grades, there were no statistically significant differences in the distribution of PDFF or R2* between the two segmentation masks. Overall, PDFF and R2* values were minimally lower when vessels were included, with a bias of -0.06% for PDFF and -0.25 s ^-1 for R2*. A lower coefficient of variation was obtained for both imaging parameters after excluding vessels. Patients were classified in the same PDFF grades despite the segmentation approach, and only 7 cases (1.9% of the study population) were reclassified for R2* grading, all being upgraded after vessel exclusion. Conclusion: Excluding hepatic vessels entails nonsignificant differences in PDFF and R2* quantification. Although with limited impact, vessel exclusion improves biomarker precision in research settings demanding high accuracy and increases clinicians’ confidence when using automatic tools in clinical practice. Relevance statement: Fat and iron quantification on MRI are key imaging biomarkers for the accurate non-invasive assessment of patients with chronic liver disease. Proton density, fat fraction, and R2* quantification show minimal differences if hepatic vessels are included or excluded from the liver segmentation mask. Key Points: The effect of hepatic vessels on proton density, fat fraction, and R2* quantification was evaluated. No significant differences were found, excluding hepatic vessels, although their inclusion showed a small negative bias. Vessel exclusion may improve clinicians’ confidence and precision in high-sensitivity applications. Graphical

Introduction: Neuroblastoma, the most prevalent solid cancer in children, presents significant biological and clinical heterogeneity. This inherent heterogeneity underscores the need for more precise prognostic markers at the time of diagnosis to enhance patient stratification, allowing for more personalized treatment strategies. In response, this investigation developed a machine learning model using clinical, molecular, and magnetic resonance (MR) radiomics features at diagnosis to predict patient’s overall survival (OS) and improve their risk stratification.Methods: PRIMAGE database, including 513 patients (discovery cohort), was used for model training, validation, and testing. Additional 22 patients from different hospitals served as an external independent cohort. Primary tumor segmentation on T2-weighted MR images was semi-automatically edited by an experienced radiologist. From this area, 107 radiomics features were extracted. For the development of the prediction model, radiomics features were harmonized following the nested ComBat methodology and nested cross-validation approach was employed to determine the optimal preprocessing and model configuration.Results: The discovery cohort yielded a 78.8 ± 4.9 and 77.7 ± 6.1 of C index and time-dependent area under de curve (AUC), respectively, over the test set, with a random survival forest exhibiting the best performance. In the independent cohort, a C-index of 93.4 and a time-dependent AUC of 95.4 were achieved. Interpretability analysis identified lesion heterogeneity, size, and molecular variables as crucial factors in OS prediction. The model stratified neuroblastoma patients into low-, intermediate-, and high-risk categories, demonstrating a superior stratification compared to standard-of-care classification system in both cohorts.Discussion: Our results suggested that radiomics features improve current risk stratification systems in patients with neuroblastoma.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To evaluate the reproducibility of radiomics features extracted from T2-weighted MRI in patients with neuroblastoma. Materials and Methods A retrospective study included 419 patients (mean (SD) age, 29 (34) years; 220 male, 199 female) with neuroblastic tumors, diagnosed between 2002-2023, within the scope of the PRIMAGE-project, involving 746 MRI T2/T2*-weighted sequences at diagnosis and/or after initial chemotherapy. Images underwent processing steps (denoising, inhomogeneity bias field correction, normalization, and resampling). Tumors were automatically segmented and 107 shape, first-order and second-order radiomic features were extracted, considered as the reference standard. Subsequently, the previous image processing settings were modified, and volumetric masks were applied. New radiomics features were extracted and compared with the reference standard. Reproducibility was assessed using the concordance correlation coefficient (CCC), intrasubject repeatability was measured using the coefficient of variation (CoV). Results When normalization was omitted, only 5% of the radiomics features demonstrated high reproducibility. Statistical analysis revealed significant changes in the normalization and resampling processes (P < .001). Inhomogeneities removal had the least impact on radiomics (83% of parameters remained stable). Shape features remained stable after mask modifications, with a CCC > 0.90. Mask modifications were the most favorable changes for achieving high CCC values, with stability of 70% of the radiomics features. Only 7% of second-order radiomics features showed an excellent CoV of < 0.10. Conclusion Modifications in the T2-weighted MRI preparation process in patients with neuroblastoma resulted in changes in radiomics features, with normalization identified as the most influential factor for reproducibility. Inhomogeneities removal had the least impact on radiomics features. ©RSNA, 2024.