Detection of recurrent high-grade glioma using microstructure characteristics of distinct metabolic compartments in a multimodal and integrative 18F-FET PET/fast-DKI approach.

OBJECTIVES: Differentiation between high-grade glioma (HGG) and post-treatment-related effects (PTRE) is challenging, but advanced imaging techniques were shown to provide benefit. We aim to investigate microstructure characteristics of metabolic compartments identified from amino acid PET and to evaluate the diagnostic potential of this multimodal and integrative O-(2-18F-fluoroethyl)-L-tyrosine-(FET)-PET and fast diffusion kurtosis imaging (DKI) approach for the detection of recurrence and IDH genotyping. METHODS: Fifty-nine participants with neuropathologically confirmed recurrent HGG (n = 39) or PTRE (n = 20) were investigated using static 18F-FET PET and a fast-DKI variant. PET and advanced diffusion metrics of metabolically defined (80-100% and 60-75% areas of 18F-FET uptake) compartments were assessed. Comparative analysis was performed using Mann-Whitney U tests with Holm-Sídák multiple-comparison test and Wilcoxon signed-rank test. Receiver operating characteristic (ROC) curves, regression, and Spearman's correlation analysis were used for statistical evaluations. RESULTS: Compared to PTRE, recurrent HGG presented increased 18F-FET uptake and diffusivity (MD60), but lower (relative) mean kurtosis tensor (rMKT60) and fractional anisotropy (FA60) (respectively p < .05). Diffusion metrics determined from the metabolic periphery showed improved diagnostic performance - most pronounced for FA60 (AUC = 0.86, p < .001), which presented similar benefit to 18F-FET PET (AUC = 0.86, p < .001) and was negatively correlated with amino acid uptake (rs = - 0.46, p < .001). When PET and DKI metrics were evaluated in a multimodal biparametric approach, TBRmax + FA60 showed highest diagnostic accuracy (AUC = 0.93, p < .001), which improved the detection of relapse compared to PET alone (difference in AUC = 0.069, p = .04). FA60 and MD60 distinguished the IDH genotype in the post-treatment setting. CONCLUSION: Detection of glioma recurrence benefits from a multimodal and integrative PET/DKI approach, which presented significant diagnostic advantage to the assessment based on PET alone. CLINICAL RELEVANCE STATEMENT: A multimodal and integrative 18F-FET PET/fast-DKI approach for the non-invasive microstructural characterization of metabolic compartments provided improved diagnostic capability for differentiation between recurrent glioma and post-treatment-related changes, suggesting a role for the diagnostic workup of patients in post-treatment settings. KEY POINTS: • Multimodal PET/MRI with integrative analysis of 18F-FET PET and fast-DKI presents clinical benefit for the assessment of CNS cancer, particularly for the detection of recurrent high-grade glioma. • Microstructure markers of the metabolic periphery yielded biologically pertinent estimates characterising the tumour microenvironment, and, thereby, presented improved diagnostic accuracy with similar accuracy to amino acid PET. • Combined 18F-FET PET/fast-DKI achieved the best diagnostic performance for detection of high-grade glioma relapse with significant benefit to the assessment based on PET alone.

Reduction of false positives using zone-specific prostate-specific antigen density for prostate MRI-based biopsy decision strategies.

OBJECTIVES: To develop and test zone-specific prostate-specific antigen density (sPSAD) combined with PI-RADS to guide prostate biopsy decision strategies (BDS). METHODS: This retrospective study included consecutive patients, who underwent prostate MRI and biopsy (01/2012-10/2018). The whole gland and transition zone (TZ) were segmented at MRI using a retrained deep learning system (DLS; nnU-Net) to calculate PSAD and sPSAD, respectively. Additionally, sPSAD and PI-RADS were combined in a BDS, and diagnostic performances to detect Grade Group ≥ 2 (GG ≥ 2) prostate cancer were compared. Patient-based cancer detection using sPSAD was assessed by bootstrapping with 1000 repetitions and reported as area under the curve (AUC). Clinical utility of the BDS was tested in the hold-out test set using decision curve analysis. Statistics included nonparametric DeLong test for AUCs and Fisher-Yates test for remaining performance metrics. RESULTS: A total of 1604 patients aged 67 (interquartile range, 61-73) with 48% GG ≥ 2 prevalence (774/1604) were evaluated. By employing DLS-based prostate and TZ volumes (DICE coefficients of 0.89 (95% confidence interval, 0.80-0.97) and 0.84 (0.70-0.99)), GG ≥ 2 detection using PSAD was inferior to sPSAD (AUC, 0.71 (0.68-0.74)/0.73 (0.70-0.76); p < 0.001). Combining PI-RADS with sPSAD, GG ≥ 2 detection specificity doubled from 18% (10-20%) to 43% (30-44%; p < 0.001) with similar sensitivity (93% (89-96%)/97% (94-99%); p = 0.052), when biopsies were taken in PI-RADS 4-5 and 3 only if sPSAD was ≥ 0.42 ng/mL/cc as compared to all PI-RADS 3-5 cases. Additionally, using the sPSAD-based BDS, false positives were reduced by 25% (123 (104-142)/165 (146-185); p < 0.001). CONCLUSION: Using sPSAD to guide biopsy decisions in PI-RADS 3 lesions can reduce false positives at MRI while maintaining high sensitivity for GG ≥ 2 cancers. CLINICAL RELEVANCE STATEMENT: Transition zone-specific prostate-specific antigen density can improve the accuracy of prostate cancer detection compared to MRI assessments alone, by lowering false-positive cases without significantly missing men with ISUP GG ≥ 2 cancers. KEY POINTS: • Prostate biopsy decision strategies using PI-RADS at MRI are limited by a substantial proportion of false positives, not yielding grade group ≥ 2 prostate cancer. • PI-RADS combined with transition zone (TZ)-specific prostate-specific antigen density (PSAD) decreased the number of unproductive biopsies by 25% compared to PI-RADS only. • TZ-specific PSAD also improved the specificity of MRI-directed biopsies by 9% compared to the whole gland PSAD, while showing identical sensitivity.

Intra- and Interobserver Study Investigating the Adapted EFSUMB Bosniak Cyst Categorization Proposed for Contrast-Enhanced Ultrasound (CEUS) in 2020.

BACKGROUND: To investigate the inter- and intraobserver variability in comparison to an expert gold standard of the new and modified renal cyst Bosniak classification proposed for contrast-enhanced ultrasound findings (CEUS) by the European Federation of Societies for Ultrasound in Medicine and Biology (EFSUMB) in 2020. MATERIALS AND METHODS: 84 CEUS examinations for the evaluation of renal cysts were evaluated retrospectively by six readers with different levels of ultrasound expertise using the modified Bosniak classification proposed for CEUS. All cases were anonymized, and each case was rated twice in randomized order. The consensus reading of two experts served as the gold standard, to which all other readers were compared. Statistical analysis was performed using Cohen's weighted kappa tests, where appropriate. RESULTS: Intraobserver variability showed substantial to almost perfect agreement (lowest kappa κ=0.74; highest kappa κ=0.94), with expert level observers achieving the best results. Comparison to the gold standard was almost perfect for experts (highest kappa κ=0.95) and lower for beginner and intermediate level readers still achieving mostly substantial agreement (lowest kappa κ=0.59). Confidence of rating was highest for Bosniak classes I and IV and lowest for classes IIF and III. CONCLUSION: Categorization of cystic renal lesions based on the Bosniak classification proposed by the EFSUMB in 2020 showed very good reproducibility. While even less experienced observers achieved mostly substantial agreement, training remains a major factor for better diagnostic performance.

Epicardial fat volume and its association with cardiac arrhythmias in CT coronary angiography.

Purpose: This retrospective study aimed to investigate the epicardial fat volume in cardiac computed tomography (CT), its relationship with cardiac arrhythmias, and its correlation with the coronary artery disease reporting and data system (CAD-RADS) score. Material and methods: Ninety-six patients who underwent CT coronary angiography (CTCA) were included in this study. Patient data, including demographic information, clinical history, and imaging data were collected retrospectively. Epicardial fat volume was quantified using a standardised algorithm, the CAD-RADS scoring system was applied to assess the extent of coronary artery disease (CAD). Descriptive statistics, correlation analyses, and receiver operating characteristics methods were used. Results: The study found a significant correlation between epicardial fat volume and CAD-RADS score (r2 = 0.31; p < 0.001), indicating the known influence of epicardial fat on CAD risk. Moreover, patients with higher epicardial fat volumes were more likely to experience cardiac tachyarrhythmia (p < 0.001). Receiver operating characteristic analysis established a threshold value of 123 cm3 for epicardial fat volume to predict tachyarrhythmia with 80% sensitivity (AUC = 0.69). Conclusions: In this study a volume of at least 123 cm3 epicardial fat in native coronary calcium scans is associated with cardiac tachyarrhythmia. In these patients, careful selection of suitable imaging protocols is advised.

Interactive Explainable Deep Learning Model Informs Prostate Cancer Diagnosis at MRI.

Background Clinically significant prostate cancer (PCa) diagnosis at MRI requires accurate and efficient radiologic interpretation. Although artificial intelligence may assist in this task, lack of transparency has limited clinical translation. Purpose To develop an explainable artificial intelligence (XAI) model for clinically significant PCa diagnosis at biparametric MRI using Prostate Imaging Reporting and Data System (PI-RADS) features for classification justification. Materials and Methods This retrospective study included consecutive patients with histopathologic analysis-proven prostatic lesions who underwent biparametric MRI and biopsy between January 2012 and December 2017. After image annotation by two radiologists, a deep learning model was trained to detect the index lesion; classify PCa, clinically significant PCa (Gleason score ≥ 7), and benign lesions (eg, prostatitis); and justify classifications using PI-RADS features. Lesion- and patient-based performance were assessed using fivefold cross validation and areas under the receiver operating characteristic curve. Clinical feasibility was tested in a multireader study and by using the external PROSTATEx data set. Statistical evaluation of the multireader study included Mann-Whitney U and exact Fisher-Yates test. Results Overall, 1224 men (median age, 67 years; IQR, 62-73 years) had 3260 prostatic lesions (372 lesions with Gleason score of 6; 743 lesions with Gleason score of ≥ 7; 2145 benign lesions). XAI reliably detected clinically significant PCa in internal (area under the receiver operating characteristic curve, 0.89) and external test sets (area under the receiver operating characteristic curve, 0.87) with a sensitivity of 93% (95% CI: 87, 98) and an average of one false-positive finding per patient. Accuracy of the visual and textual explanations of XAI classifications was 80% (1080 of 1352), confirmed by experts. XAI-assisted readings improved the confidence (4.1 vs 3.4 on a five-point Likert scale; P = .007) of nonexperts in assessing PI-RADS 3 lesions, reducing reading time by 58 seconds (P = .009). Conclusion The explainable AI model reliably detected and classified clinically significant prostate cancer and improved the confidence and reading time of nonexperts while providing visual and textual explanations using well-established imaging features. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Chapiro in this issue.

Gd-EOB MRI for HCC subtype differentiation in a western population according to the 5th edition of the World Health Organization classification.

OBJECTIVES: To investigate the value of gadoxetic acid (Gd-EOB)-enhanced magnetic resonance imaging (MRI) for noninvasive subtype differentiation of HCCs according to the 5th edition of the WHO Classification of Digestive System Tumors in a western population. METHODS: This retrospective study included 262 resected lesions in 240 patients with preoperative Gd-EOB-enhanced MRI. Subtypes were assigned by two pathologists. Gd-EOB-enhanced MRI datasets were assessed by two radiologists for qualitative and quantitative imaging features, including imaging features defined in LI-RADS v2018 and area of hepatobiliary phase (HBP) iso- to hyperintensity. RESULTS: The combination of non-rim arterial phase hyperenhancement with non-peripheral portal venous washout was more common in "not otherwise specified" (nos-ST) (88/168, 52%) than other subtypes, in particular macrotrabecular massive (mt-ST) (3/15, 20%), chromophobe (ch-ST) (1/8, 13%), and scirrhous subtypes (sc-ST) (2/9, 22%) (p = 0.035). Macrovascular invasion was associated with mt-ST (5/16, p = 0.033) and intralesional steatosis with steatohepatitic subtype (sh-ST) (28/32, p < 0.001). Predominant iso- to hyperintensity in the HBP was only present in nos-ST (16/174), sh-ST (3/33), and clear cell subtypes (cc-ST) (3/13) (p = 0.031). Associations were found for the following non-imaging parameters: age and sex, as patients with fibrolamellar subtype (fib-ST) were younger (median 44 years (19-66), p < 0.001) and female (4/5, p = 0.023); logarithm of alpha-fetoprotein (AFP) was elevated in the mt-ST (median 397 µg/l (74-5370), p < 0.001); type II diabetes mellitus was more frequent in the sh-ST (20/33, p = 0.027). CONCLUSIONS: Gd-EOB-MRI reproduces findings reported in the literature for extracellular contrast-enhanced MRI and CT and may be a valuable tool for noninvasive HCC subtype differentiation. CLINICAL RELEVANCE STATEMENT: Better characterization of the heterogeneous phenotypes of HCC according to the revised WHO classification potentially improves both diagnostic accuracy and the precision of therapeutic stratification for HCC. KEY POINTS: • Previously reported imaging features of common subtypes in CT and MRI enhanced with extracellular contrast agents are reproducible with Gd-EOB-enhanced MRI. • While uncommon, predominant iso- to hyperintensity in the HBP was observed only in NOS, clear cell, and steatohepatitic subtypes. • Gd-EOB-enhanced MRI offers imaging features that are of value for HCC subtype differentiation according to the 5th edition of the WHO Classification of Digestive System Tumors.

Gd-EOB-DTPA-MRCP to localize bile leakage after liver trauma and surgery: impact on treatment and outcome.

OBJECTIVES: Bile leakage (BL) is a challenging complication after hepatobiliary surgery and liver trauma. Gadolinium ethoxybenzyl (Gd-EOB-DTPA)-enhanced magnetic resonance cholangiopancreatography (MRCP) is used to diagnose BL non-invasively. We assessed the value of Gd-EOB-DTPA-MRCP in the detection of postoperative and post-traumatic BL hypothesizing that exact identification of the leakage site is pivotal for treatment planning and outcome. METHODS: We retrospectively enrolled 39 trauma and postoperative patients who underwent Gd-EOB-DTPA-MRCP for suspected BL. Three readers rated the presence of BL and leakage site (intraparenchymal, central, peripheral ± aberrant or disconnected ducts). Imaging findings were compared to subsequent interventional procedures and their complexity and outcome. RESULTS: BL was detected in Gd-EOB-DTPA-MRCP in 25 of patients and was subsequently confirmed. Sites of BL differed significantly between postoperative (central [58%] and peripheral [42%]) and trauma patients (intraparenchymal [100%]; p < 0.001). Aberrant or disconnected ducts were diagnosed in 8%/26% of cases in the postoperative subgroup. Inter-rater agreement for the detection and localization of BL was almost perfect (Κ = 0.85 and 0.88; p < 0.001). Intraparenchymal BL required significantly less complex interventional procedures (p = 0.002), whereas hospitalization and mortality did not differ between the subgroups (p > 0.05). CONCLUSIONS: Gd-EOB-DTPA-MRCP reliably detects and exactly locates BL in postoperative and trauma patients. Exact localization of biliary injuries enables specific treatment planning, as intraparenchymal leakages, which occur more frequently after trauma, require less complex interventions than central or peripheral leaks in the postoperative setting. As a result of specific treatment based on exact BL localization, there was no difference in the duration of hospitalization or mortality. CLINICAL RELEVANCE STATEMENT: Gd-EOB-DTPA-MRCP is a reliable diagnostic tool for exactly localizing iatrogenic and post-traumatic biliary leakage. Its precise localization helps tailor local therapies for different injury patterns, resulting in comparable clinical outcomes despite varying treatments. KEY POINTS: • Gd-EOB-DTPA-MRCP enables adequate detection and localization of bile leakages in both postoperative and post-traumatic patients. • The site of bile leakage significantly impacts the complexity of required additional interventions. • Intraparenchymal bile leakage is commonly seen in patients with a history of liver trauma and requires less complex interventions than postoperative central or peripheral bile leakages, while hospitalization and mortality are similar.

Diagnostic performance of Node Reporting and Data System (Node-RADS) for regional lymph node staging of gastric cancer by CT.

OBJECTIVES: Diagnostic performance of imaging for regional lymph node assessment in gastric cancer is still limited, and there is a lack of consensus on radiological evaluation. At the same time, there is an increasing demand for structured reporting using Reporting and Data Systems (RADS) to standardize oncological imaging. We aimed at investigating the diagnostic performance of Node-RADS compared to the use of various individual criteria for assessing regional lymph nodes in gastric cancer using histopathology as reference. METHODS: In this retrospective single-center study, consecutive 91 patients (median age, 66 years, range 33-91 years, 54 men) with CT scans and histologically proven gastric adenocarcinoma were assessed using Node-RADS assigning scores from 1 to 5 for the likelihood of regional lymph node metastases. Additionally, different Node-RADS criteria as well as subcategories of altered border contour (lobulated, spiculated, indistinct) were assessed individually. Sensitivity, specificity, and Youden's index were calculated for Node-RADS scores, and all criteria investigated. Interreader agreement was calculated using Cohen's kappa. RESULTS: Among all criteria, best performance was found for Node-RADS scores ≥ 3 and ≥ 4 with a sensitivity/specificity/Youden's index of 56.8%/90.7%/0.48 and 48.6%/98.1%/0.47, respectively, both with substantial interreader agreement (κ = 0.73 and 0.67, p < 0.01). Among individual criteria, the best performance was found for short-axis diameter of 10 mm with sensitivity/specificity/Youden's index of 56.8%/87.0%/0.44 (κ = 0.65, p < 0.01). CONCLUSION: This study shows that structured reporting of combined size and configuration criteria of regional lymph nodes in gastric cancer slightly improves overall diagnostic performance compared to individual criteria including short-axis diameter alone. The results show an increase in specificity and unchanged sensitivity. CLINICAL RELEVANCE STATEMENT: The results of this study suggest that Node-RADS may be a suitable tool for structured reporting of regional lymph nodes in gastric cancer. KEY POINTS: • Assessment of lymph nodes in gastric cancer is still limited, and there is a lack of consensus on radiological evaluation. • Node-RADS in gastric cancer improves overall diagnostic performance compared to individual criteria including short-axis diameter. • Node-RADS may be a suitable tool for structured reporting of regional lymph nodes in gastric cancer.

Work expectations, their fulfillment, and exhaustion among radiologists of all career levels: what can be learned from the example of Germany.

OBJECTIVES: To evaluate work expectations of radiologists at different career levels, their fulfillment, prevalence of exhaustion, and exhaustion-associated factors. METHODS: A standardized digital questionnaire was distributed internationally to radiologists of all career levels in the hospital and in ambulatory care via radiological societies and sent manually to 4500 radiologists of the largest German hospitals between December 2020 and April 2021. Statistics were based on age- and gender-adjusted regression analyses of respondents working in Germany (510 out of 594 total respondents). RESULTS: The most frequent expectations were "joy at work" (97%) and a "good working atmosphere" (97%), which were considered fulfilled by at least 78%. The expectation of a "structured residency within the regular time interval" (79%) was more frequently judged fulfilled by senior physicians (83%, odds ratio (OR) 4.31 [95% confidence interval (95% CI) 1.95-9.52]), chief physicians (85%, 6.81 [95% CI 1.91-24.29]), and radiologists outside the hospital (88%, 7.59 [95% CI 2.40-24.03]) than by residents (68%). Exhaustion was most common among residents (physical exhaustion: 38%; emotional exhaustion: 36%), in-hospital specialists (29%; 38%), and senior physicians (30%; 29%). In contrast to paid extra hours, unpaid extra hours were associated with physical exhaustion (5-10 extra hours: OR 2.54 [95% CI 1.54-4.19]). Fewer opportunities to shape the work environment were related to a higher probability of physical (2.03 [95% CI 1.32-3.13]) and emotional (2.15 [95% CI 1.39-3.33]) exhaustion. CONCLUSIONS: While most radiologists enjoy their work, residents wish for more training structure. Ensuring payment of extra hours and employee empowerment may help preventing burnout in high-risk groups. KEY POINTS: • Most important work expectations of radiologists who work in Germany are "joy at work," a "good working atmosphere," "support for further qualification," and a "structured residency within the regular time interval," with the latter containing potential for improvement according to residents. • Physical and emotional exhaustion are common at all career levels except for chief physicians and for radiologists who work outside the hospital in ambulatory care. • Exhaustion as a major burnout criterion is associated with unpaid extra hours and reduced opportunities to shape the work environment.

Can optimized model-based iterative reconstruction improve the contrast of liver lesions in CT?

BACKGROUND: Computed tomography is a standard imaging procedure for the detection of liver lesions, such as metastases, which can often be small and poorly contrasted, and therefore hard to detect. Advances in image reconstruction have shown promise in reducing image noise and improving low-contrast detectability. PURPOSE: To examine a novel, specialized, model-based iterative reconstruction (MBIR) technique for improved low-contrast liver lesion detection. MATERIAL AND METHODS: Patient images with reported poorly contrasted focal liver lesions were retrospectively reconstructed with the low-contrast attenuating algorithm (FIRST-LCD) from primary raw data. Liver-to-lesion contrast, signal-to-noise, and contrast-to-noise ratios for background and liver noise for each lesion were compared for all three FIRST-LCD presets with the established hybrid iterative reconstruction method (AIDR-3D). An additional visual conspicuity score was given by two experienced radiologists for each lesion. RESULTS: A total of 82 lesions in 57 examinations were included in the analysis. All three FIRST-LCD algorithms provided statistically significant increases in liver-to-lesion contrast, with FIRSTMILD showing the largest increase (40.47 HU in AIDR-3D; 45.84 HU in FIRSTMILD; P < 0.001). Substantial improvement was shown in contrast-to-noise metrics. Visual analysis of the lesions shows decreased lesion visibility with all FIRST methods in comparison to AIDR-3D, with FIRSTSTR showing the closest results (P < 0.001). CONCLUSION: Objective image metrics show promise for MBIR methods in improving the detectability of low-contrast liver lesions; however, subjective image quality may be perceived as inferior. Further improvements are necessary to enhance image quality and lesion detection.

Low back pain in adolescent rowers: Association to muscle changes detected by magnetic resonance imaging.

Adult elite rowers are at risk of developing low back pain (LBP). However, LBP data on adolescent elite rowers is currently insufficient. Therefore, the aim of this study was to assess LBP prevalence, LBP intensity and training characteristics in male adolescent elite rowers and a healthy control group. Twenty rowers (mean age 15.8 ± 1.2 years) and a non-athletic control group matched by age and gender (n = 13) were prospectively enrolled and underwent LBP assessment with a validated questionnaire and magnetic resonance imaging (MRI) of the lumbar spine muscles, which included a T2-mapping sequence. From the quantitative image data, T2 relaxation times were calculated. The prevalence of LBP in the last 24 hours and 3 months in the rowing group was 55.0% and 85.0%, respectively, compared to 23.1% and 30.8% in the control group (p < 0.001). Rowers had significantly longer T2 relaxation times of the paraspinal muscles compared to controls (p ≤ 0.041). LBP intensity was associated with longer T2 relaxation times (p < 0.001). Adolescent rowers had a higher prevalence of LBP compared to an age-matched control group. The observed increase in T2 relaxation might be explained by muscle soreness due to strenuous exercise, which is correlated with short-term pain intensity.

Deep Learning Detects Changes Indicative of Axial Spondyloarthritis at MRI of Sacroiliac Joints.

Background MRI is frequently used for early diagnosis of axial spondyloarthritis (axSpA). However, evaluation is time-consuming and requires profound expertise because noninflammatory degenerative changes can mimic axSpA, and early signs may therefore be missed. Deep neural networks could function as assistance for axSpA detection. Purpose To create a deep neural network to detect MRI changes in sacroiliac joints indicative of axSpA. Materials and Methods This retrospective multicenter study included MRI examinations of five cohorts of patients with clinical suspicion of axSpA collected at university and community hospitals between January 2006 and September 2020. Data from four cohorts were used as the training set, and data from one cohort as the external test set. Each MRI examination in the training and test sets was scored by six and seven raters, respectively, for inflammatory changes (bone marrow edema, enthesitis) and structural changes (erosions, sclerosis). A deep learning tool to detect changes indicative of axSpA was developed. First, a neural network to homogenize the images, then a classification network were trained. Performance was evaluated with use of area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. P < .05 was considered indicative of statistically significant difference. Results Overall, 593 patients (mean age, 37 years ± 11 [SD]; 302 women) were studied. Inflammatory and structural changes were found in 197 of 477 patients (41%) and 244 of 477 (51%), respectively, in the training set and 25 of 116 patients (22%) and 26 of 116 (22%) in the test set. The AUCs were 0.94 (95% CI: 0.84, 0.97) for all inflammatory changes, 0.88 (95% CI: 0.80, 0.95) for inflammatory changes fulfilling the Assessment of SpondyloArthritis international Society definition, and 0.89 (95% CI: 0.81, 0.96) for structural changes indicative of axSpA. Sensitivity and specificity on the external test set were 22 of 25 patients (88%) and 65 of 91 patients (71%), respectively, for inflammatory changes and 22 of 26 patients (85%) and 70 of 90 patients (78%) for structural changes. Conclusion Deep neural networks can detect inflammatory or structural changes to the sacroiliac joint indicative of axial spondyloarthritis at MRI. © RSNA, 2022 Online supplemental material is available for this article.

Highly accurate classification of chest radiographic reports using a deep learning natural language model pre-trained on 3.8 million text reports.

MOTIVATION: The development of deep, bidirectional transformers such as Bidirectional Encoder Representations from Transformers (BERT) led to an outperformance of several Natural Language Processing (NLP) benchmarks. Especially in radiology, large amounts of free-text data are generated in daily clinical workflow. These report texts could be of particular use for the generation of labels in machine learning, especially for image classification. However, as report texts are mostly unstructured, advanced NLP methods are needed to enable accurate text classification. While neural networks can be used for this purpose, they must first be trained on large amounts of manually labelled data to achieve good results. In contrast, BERT models can be pre-trained on unlabelled data and then only require fine tuning on a small amount of manually labelled data to achieve even better results. RESULTS: Using BERT to identify the most important findings in intensive care chest radiograph reports, we achieve areas under the receiver operation characteristics curve of 0.98 for congestion, 0.97 for effusion, 0.97 for consolidation and 0.99 for pneumothorax, surpassing the accuracy of previous approaches with comparatively little annotation effort. Our approach could therefore help to improve information extraction from free-text medical reports. Availability and implementationWe make the source code for fine-tuning the BERT-models freely available at https://github.com/fast-raidiology/bert-for-radiology. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Fibrin-targeting molecular MRI in inflammatory CNS disorders.

BACKGROUND: Fibrin deposition is a fundamental pathophysiological event in the inflammatory component of various CNS disorders, such as multiple sclerosis (MS) and Alzheimer's disease. Beyond its traditional role in coagulation, fibrin elicits immunoinflammatory changes with oxidative stress response and activation of CNS-resident/peripheral immune cells contributing to CNS injury. PURPOSE: To investigate if CNS fibrin deposition can be determined using molecular MRI, and to assess its capacity as a non-invasive imaging biomarker that corresponds to inflammatory response and barrier impairment. MATERIALS AND METHODS: Specificity and efficacy of a peptide-conjugated Gd-based molecular MRI probe (EP2104-R) to visualise and quantify CNS fibrin deposition were evaluated. Probe efficacy to specifically target CNS fibrin deposition in murine adoptive-transfer experimental autoimmune encephalomyelitis (EAE), a pre-clinical model for MS (n = 12), was assessed. Findings were validated using immunohistochemistry and laser ablation inductively coupled plasma mass spectrometry. Deposition of fibrin in neuroinflammatory conditions was investigated and its diagnostic capacity for disease staging and monitoring as well as quantification of immunoinflammatory response was determined. Results were compared using t-tests (two groups) or one-way ANOVA with multiple comparisons test. Linear regression was used to model the relationship between variables. RESULTS: For the first time (to our knowledge), CNS fibrin deposition was visualised and quantified in vivo using molecular imaging. Signal enhancement was apparent in EAE lesions even 12-h after administration of EP2104-R due to targeted binding (M ± SD, 1.07 ± 0.10 (baseline) vs. 0.73 ± 0.09 (EP2104-R), p = .008), which could be inhibited with an MRI-silent analogue (M ± SD, 0.60 ± 0.14 (EP2104-R) vs. 0.96 ± 0.13 (EP2104-La), p = .006). CNS fibrin deposition corresponded to immunoinflammatory activity (R2 = 0.85, p < .001) and disability (R2 = 0.81, p < .001) in a model for MS, which suggests a clinical role for staging and monitoring. Additionally, EP2104-R showed substantially higher SNR (M ± SD, 6.6 ± 1 (EP2104-R) vs. 2.7 ± 0.4 (gadobutrol), p = .004) than clinically used contrast media, which increases sensitivity for lesion detection. CONCLUSIONS: Molecular imaging of CNS fibrin deposition provides an imaging biomarker for inflammatory CNS pathology, which corresponds to pathophysiological ECM remodelling and disease activity, and yields high signal-to-noise ratio, which can improve diagnostic neuroimaging across several neurological diseases with variable degrees of barrier impairment.

Feasibility of Intestinal MR Elastography in Inflammatory Bowel Disease.

BACKGROUND: While MR enterography allows detection of inflammatory bowel disease (IBD), the findings continue to be of limited use in guiding treatment-medication vs. surgery. PURPOSE: To test the feasibility of MR elastography of the gut in healthy volunteers and IBD patients. STUDY TYPE: Prospective pilot. POPULATION: Forty subjects (healthy volunteers: n = 20, 37 ± 14 years, 10 women; IBD patients: n = 20 (ulcerative colitis n = 9, Crohn's disease n = 11), 41 ± 15 years, 11 women). FIELD STRENGTH/SEQUENCE: Multifrequency MR elastography using a single-shot spin-echo echo planar imaging sequence at 1.5 T with drive frequencies of 40, 50, 60, and 70 Hz. ASSESSMENT: Maps of shear-wave speed (SWS, in m/s) and loss angle (φ, in rad), representing stiffness and solid-fluid behavior, respectively, were generated using tomoelastography data processing. Histopathological analysis of surgical specimens was used as reference standard in patients. STATISTICAL TESTS: Unpaired t-test, one-way analysis of variance followed by Tukey post hoc analysis, Pearson's correlation coefficient and area under the receiver operating characteristic curve (AUC) with 95%-confidence interval (CI). Significance level of 5%. RESULTS: MR elastography was feasible in all 40 subjects (100% technical success rate). SWS and φ were significantly increased in IBD by 21% and 20% (IBD: 1.45 ± 0.14 m/s and 0.78 ± 0.12 rad; healthy volunteers: 1.20 ± 0.14 m/s and 0.65 ± 0.06 rad), whereas no significant differences were found between ulcerative colitis and Crohn's disease (P = 0.74 and 0.90, respectively). In a preliminary assessment, a high diagnostic accuracy in detecting IBD was suggested by an AUC of 0.90 (CI: 0.81-0.96) for SWS and 0.84 (CI: 0.71-0.95) for φ. DATA CONCLUSION: In this pilot study, our results demonstrated the feasibility of MR elastography of the gut and showed an excellent diagnostic performance in predicting IBD. EVIDENCE LEVEL: 1 TECHNICAL EFFICACY: Stage 1.

Deep learning reconstruction improves radiomics feature stability and discriminative power in abdominal CT imaging: a phantom study.

OBJECTIVES: To compare image quality of deep learning reconstruction (AiCE) for radiomics feature extraction with filtered back projection (FBP), hybrid iterative reconstruction (AIDR 3D), and model-based iterative reconstruction (FIRST). METHODS: Effects of image reconstruction on radiomics features were investigated using a phantom that realistically mimicked a 65-year-old patient's abdomen with hepatic metastases. The phantom was scanned at 18 doses from 0.2 to 4 mGy, with 20 repeated scans per dose. Images were reconstructed with FBP, AIDR 3D, FIRST, and AiCE. Ninety-three radiomics features were extracted from 24 regions of interest, which were evenly distributed across three tissue classes: normal liver, metastatic core, and metastatic rim. Features were analyzed in terms of their consistent characterization of tissues within the same image (intraclass correlation coefficient ≥ 0.75), discriminative power (Kruskal-Wallis test p value < 0.05), and repeatability (overall concordance correlation coefficient ≥ 0.75). RESULTS: The median fraction of consistent features across all doses was 6%, 8%, 6%, and 22% with FBP, AIDR 3D, FIRST, and AiCE, respectively. Adequate discriminative power was achieved by 48%, 82%, 84%, and 92% of features, and 52%, 20%, 17%, and 39% of features were repeatable, respectively. Only 5% of features combined consistency, discriminative power, and repeatability with FBP, AIDR 3D, and FIRST versus 13% with AiCE at doses above 1 mGy and 17% at doses ≥ 3 mGy. AiCE was the only reconstruction technique that enabled extraction of higher-order features. CONCLUSIONS: AiCE more than doubled the yield of radiomics features at doses typically used clinically. Inconsistent tissue characterization within CT images contributes significantly to the poor stability of radiomics features. KEY POINTS: • Image quality of CT images reconstructed with filtered back projection and iterative methods is inadequate for the majority of radiomics features due to inconsistent tissue characterization, low discriminative power, or low repeatability. • Deep learning reconstruction enhances image quality for radiomics and more than doubled the feature yield at doses that are typically used in clinical CT imaging. • Image reconstruction algorithms can optimize image quality for more reliable quantification of tissues in CT images.

Comparison of low-contrast detectability between uniform and anatomically realistic phantoms-influences on CT image quality assessment.

OBJECTIVES: To evaluate the effects of anatomical phantom structure on task-based image quality assessment compared with a uniform phantom background. METHODS: Two neck phantom types of identical shape were investigated: a uniform type containing 10-mm lesions with 4, 9, 18, 30, and 38 HU contrast to the surrounding area and an anatomically realistic type containing lesions of the same size and location with 10, 18, 30, and 38 HU contrast. Phantom images were acquired at two dose levels (CTDIvol of 1.4 and 5.6 mGy) and reconstructed using filtered back projection (FBP) and adaptive iterative dose reduction 3D (AIDR 3D). Detection accuracy was evaluated by seven radiologists in a 4-alternative forced choice experiment. RESULTS: Anatomical phantom structure impaired lesion detection at all lesion contrasts (p < 0.01). Detectability in the anatomical phantom at 30 HU contrast was similar to 9 HU contrast in uniform images (91.1% vs. 89.5%). Detection accuracy decreased from 83.6% at 5.6 mGy to 55.4% at 1.4 mGy in uniform FBP images (p < 0.001), whereas AIDR 3D preserved detectability at 1.4 mGy (80.7% vs. 85% at 5.6 mGy, p = 0.375) and was superior to FBP (p < 0.001). In the assessment of anatomical images, superiority of AIDR 3D was not confirmed and dose reduction moderately affected detectability (74.6% vs. 68.2%, p = 0.027 for FBP and 81.1% vs. 73%, p = 0.018 for AIDR 3D). CONCLUSIONS: A lesion contrast increase from 9 to 30 HU is necessary for similar detectability in anatomical and uniform neck phantom images. Anatomical phantom structure influences task-based assessment of iterative reconstruction and dose effects. KEY POINTS: • A lesion contrast increase from 9 to 30 HU is necessary for similar low-contrast detectability in anatomical and uniform neck phantom images. • Phantom background structure influences task-based assessment of iterative reconstruction and dose effects. • Transferability of CT assessment to clinical imaging can be expected to improve as the realism of the test environment increases.

Prediction of prostate cancer grade using fractal analysis of perfusion MRI: retrospective proof-of-principle study.

OBJECTIVES: Multiparametric MRI has high diagnostic accuracy for detecting prostate cancer, but non-invasive prediction of tumor grade remains challenging. Characterizing tumor perfusion by exploiting the fractal nature of vascular anatomy might elucidate the aggressive potential of a tumor. This study introduces the concept of fractal analysis for characterizing prostate cancer perfusion and reports about its usefulness for non-invasive prediction of tumor grade. METHODS: We retrospectively analyzed the openly available PROSTATEx dataset with 112 cancer foci in 99 patients. In all patients, histological grading groups specified by the International Society of Urological Pathology (ISUP) were obtained from in-bore MRI-guided biopsy. Fractal analysis of dynamic contrast-enhanced perfusion MRI sequences was performed, yielding fractal dimension (FD) as quantitative descriptor. Two-class and multiclass diagnostic accuracy was analyzed using area under the curve (AUC) receiver operating characteristic analysis, and optimal FD cutoffs were established. Additionally, we compared fractal analysis to conventional apparent diffusion coefficient (ADC) measurements. RESULTS: Fractal analysis of perfusion allowed accurate differentiation of non-significant (group 1) and clinically significant (groups 2-5) cancer with a sensitivity of 91% (confidence interval [CI]: 83-96%) and a specificity of 86% (CI: 73-94%). FD correlated linearly with ISUP groups (r2 = 0.874, p < 0.001). Significant groupwise differences were obtained between low, intermediate, and high ISUP group 1-4 (p ≤ 0.001) but not group 5 tumors. Fractal analysis of perfusion was significantly more reliable than ADC in predicting non-significant and clinically significant cancer (AUCFD = 0.97 versus AUCADC = 0.77, p < 0.001). CONCLUSION: Fractal analysis of perfusion MRI accurately predicts prostate cancer grading in low-, intermediate-, and high-, but not highest-grade, tumors. KEY POINTS: • In 112 prostate carcinomas, fractal analysis of MR perfusion imaging accurately differentiated low-, intermediate-, and high-grade cancer (ISUP grade groups 1-4). • Fractal analysis detected clinically significant prostate cancer with a sensitivity of 91% (83-96%) and a specificity of 86% (73-94%). • Fractal dimension of perfusion at the tumor margin may provide an imaging biomarker to predict prostate cancer grading.

Accuracy of fractal analysis and PI-RADS assessment of prostate magnetic resonance imaging for prediction of cancer grade groups: a clinical validation study.

OBJECTIVES: Multiparametric MRI with Prostate Imaging Reporting and Data System (PI-RADS) assessment is sensitive but not specific for detecting clinically significant prostate cancer. This study validates the diagnostic accuracy of the recently suggested fractal dimension (FD) of perfusion for detecting clinically significant cancer. MATERIALS AND METHODS: Routine clinical MR imaging data, acquired at 3 T without an endorectal coil including dynamic contrast-enhanced sequences, of 72 prostate cancer foci in 64 patients were analyzed. In-bore MRI-guided biopsy with International Society of Urological Pathology (ISUP) grading served as reference standard. Previously established FD cutoffs for predicting tumor grade were compared to measurements of the apparent diffusion coefficient (25th percentile, ADC25) and PI-RADS assessment with and without inclusion of the FD as separate criterion. RESULTS: Fractal analysis allowed prediction of ISUP grade groups 1 to 4 but not 5, with high agreement to the reference standard (κFD = 0.88 [CI: 0.79-0.98]). Integrating fractal analysis into PI-RADS allowed a strong improvement in specificity and overall accuracy while maintaining high sensitivity for significant cancer detection (ISUP > 1; PI-RADS alone: sensitivity = 96%, specificity = 20%, area under the receiver operating curve [AUC] = 0.65; versus PI-RADS with fractal analysis: sensitivity = 95%, specificity = 88%, AUC = 0.92, p < 0.001). ADC25 only differentiated low-grade group 1 from pooled higher-grade groups 2-5 (κADC = 0.36 [CI: 0.12-0.59]). Importantly, fractal analysis was significantly more reliable than ADC25 in predicting non-significant and clinically significant cancer (AUCFD = 0.96 versus AUCADC = 0.75, p < 0.001). Diagnostic accuracy was not significantly affected by zone location. CONCLUSIONS: Fractal analysis is accurate in noninvasively predicting tumor grades in prostate cancer and adds independent information when implemented into PI-RADS assessment. This opens the opportunity to individually adjust biopsy priority and method in individual patients. KEY POINTS: • Fractal analysis of perfusion is accurate in noninvasively predicting tumor grades in prostate cancer using dynamic contrast-enhanced sequences (κFD = 0.88). • Including the fractal dimension into PI-RADS as a separate criterion improved specificity (from 20 to 88%) and overall accuracy (AUC from 0.86 to 0.96) while maintaining high sensitivity (96% versus 95%) for predicting clinically significant cancer. • Fractal analysis was significantly more reliable than ADC25 in predicting clinically significant cancer (AUCFD = 0.96 versus AUCADC = 0.75).

Yttrium-90 radioembolization for unresectable hepatocellular carcinoma: predictive modeling strategies to anticipate tumor response and improve patient selection.

OBJECTIVES: This study aims to better characterize potential responders of Y-90-radioembolization at baseline through analysis of clinical variables and contrast enhanced (CE) MRI tumor volumetry in order to adjust therapeutic regimens early on and to improve treatment outcomes. METHODS: Fifty-eight HCC patients who underwent Y-90-radioembolization at our center between 10/2008 and 02/2017 were retrospectively included. Pre- and post-treatment target lesion volumes were measured as total tumor volume (TTV) and enhancing tumor volume (ETV). Survival analysis was performed with Cox regression models to evaluate 65% ETV reduction as surrogate endpoint for treatment efficacy. Univariable and multivariable logistic regression analyses were used to evaluate the combination of baseline clinical variables and tumor volumetry as predictors of ≥ 65% ETV reduction. RESULTS: Mean patients' age was 66 (SD 8.7) years, and 12 were female (21%). Sixty-seven percent of patients suffered from liver cirrhosis. Median survival was 11 months. A threshold of ≥ 65% in ETV reduction allowed for a significant (p = 0.04) separation of the survival curves with a median survival of 11 months in non-responders and 17 months in responders. Administered activity per tumor volume did predict neither survival nor ETV reduction. A baseline ETV/TTV ratio greater than 50% was the most important predictor of arterial devascularization (odds ratio 6.3) in a statistically significant (p = 0.001) multivariable logistic regression model. The effect size was strong with a Cohen's f of 0.89. CONCLUSION: We present a novel approach to identify promising candidates for Y-90 radioembolization at pre-treatment baseline MRI using tumor volumetry and clinical baseline variables. KEY POINTS: • A decrease of 65% enhancing tumor volume (ETV) on follow-up imaging 2-3 months after Y-90 radioembolization of HCC enables the early prediction of significantly improved median overall survival (11 months vs. 17 months, p = 0.04). • Said decrease in vascularization is predictable at baseline: an ETV greater than 50% is the most important variable in a multivariable logistic regression model that predicts responders at a high level of significance (p = 0.001) with an area under the curve of 87%.