Reproducible Radiomics Features from Multi-MRI-Scanner Test-Retest-Study: Influence on Performance and Generalizability of Models.

BACKGROUND: Radiomics models trained on data from one center typically show a decline of performance when applied to data from external centers, hindering their introduction into large-scale clinical practice. Current expert recommendations suggest to use only reproducible radiomics features isolated by multiscanner test-retest experiments, which might help to overcome the problem of limited generalizability to external data. PURPOSE: To evaluate the influence of using only a subset of robust radiomics features, defined in a prior in vivo multi-MRI-scanner test-retest-study, on the performance and generalizability of radiomics models. STUDY TYPE: Retrospective. POPULATION: Patients with monoclonal plasma cell disorders. Training set (117 MRIs from center 1); internal test set (42 MRIs from center 1); external test set (143 MRIs from center 2-8). FIELD STRENGTH/SEQUENCE: 1.5T and 3.0T; T1-weighted turbo spin echo. ASSESSMENT: The task for the radiomics models was to predict plasma cell infiltration, determined by bone marrow biopsy, noninvasively from MRI. Radiomics machine learning models, including linear regressor, support vector regressor (SVR), and random forest regressor (RFR), were trained on data from center 1, using either all radiomics features, or using only reproducible radiomics features. Models were tested on an internal (center 1) and a multicentric external data set (center 2-8). STATISTICAL TESTS: Pearson correlation coefficient r and mean absolute error (MAE) between predicted and actual plasma cell infiltration. Fisher's z-transformation, Wilcoxon signed-rank test, Wilcoxon rank-sum test; significance level P < 0.05. RESULTS: When using only reproducible features compared with all features, the performance of the SVR on the external test set significantly improved (r = 0.43 vs. r = 0.18 and MAE = 22.6 vs. MAE = 28.2). For the RFR, the performance on the external test set deteriorated when using only reproducible instead of all radiomics features (r = 0.33 vs. r = 0.44, P = 0.29 and MAE = 21.9 vs. MAE = 20.5, P = 0.10). CONCLUSION: Using only reproducible radiomics features improves the external performance of some, but not all machine learning models, and did not automatically lead to an improvement of the external performance of the overall best radiomics model. TECHNICAL EFFICACY: Stage 2.

Potential of ChatGPT and GPT-4 for Data Mining of Free-Text CT Reports on Lung Cancer.

Background The latest large language models (LLMs) solve unseen problems via user-defined text prompts without the need for retraining, offering potentially more efficient information extraction from free-text medical records than manual annotation. Purpose To compare the performance of the LLMs ChatGPT and GPT-4 in data mining and labeling oncologic phenotypes from free-text CT reports on lung cancer by using user-defined prompts. Materials and Methods This retrospective study included patients who underwent lung cancer follow-up CT between September 2021 and March 2023. A subset of 25 reports was reserved for prompt engineering to instruct the LLMs in extracting lesion diameters, labeling metastatic disease, and assessing oncologic progression. This output was fed into a rule-based natural language processing pipeline to match ground truth annotations from four radiologists and derive performance metrics. The oncologic reasoning of LLMs was rated on a five-point Likert scale for factual correctness and accuracy. The occurrence of confabulations was recorded. Statistical analyses included Wilcoxon signed rank and McNemar tests. Results On 424 CT reports from 424 patients (mean age, 65 years ± 11 [SD]; 265 male), GPT-4 outperformed ChatGPT in extracting lesion parameters (98.6% vs 84.0%, P < .001), resulting in 96% correctly mined reports (vs 67% for ChatGPT, P < .001). GPT-4 achieved higher accuracy in identification of metastatic disease (98.1% [95% CI: 97.7, 98.5] vs 90.3% [95% CI: 89.4, 91.0]) and higher performance in generating correct labels for oncologic progression (F1 score, 0.96 [95% CI: 0.94, 0.98] vs 0.91 [95% CI: 0.89, 0.94]) (both P < .001). In oncologic reasoning, GPT-4 had higher Likert scale scores for factual correctness (4.3 vs 3.9) and accuracy (4.4 vs 3.3), with a lower rate of confabulation (1.7% vs 13.7%) than ChatGPT (all P < .001). Conclusion When using user-defined prompts, GPT-4 outperformed ChatGPT in extracting oncologic phenotypes from free-text CT reports on lung cancer and demonstrated better oncologic reasoning with fewer confabulations. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Hafezi-Nejad and Trivedi in this issue.

CT Angiography Clot Burden Score from Data Mining of Structured Reports for Pulmonary Embolism.

Background Many studies emphasize the role of structured reports (SRs) because they are readily accessible for further automated analyses. However, using SR data obtained in clinical routine for research purposes is not yet well represented in literature. Purpose To compare the performance of the Qanadli scoring system with a clot burden score mined from structured pulmonary embolism (PE) reports from CT angiography. Materials and Methods In this retrospective study, a rule-based text mining pipeline was developed to extract descriptors of PE and right heart strain from SR of patients with suspected PE between March 2017 and February 2020. From standardized PE reporting, a pulmonary artery obstruction index (PAOI) clot burden score (PAOICBS) was derived and compared with the Qanadli score (PAOIQ). Scoring time and confidence from two independent readings were compared. Interobserver and interscore agreement was tested by using the intraclass correlation coefficient (ICC) and Bland-Altman analysis. To assess conformity and diagnostic performance of both scores, areas under the receiver operating characteristic curve (AUCs) were calculated to predict right heart strain incidence, as were optimal cutoff values for maximum sensitivity and specificity. Results SR content authored by 67 residents and signed off by 32 consultants from 1248 patients (mean age, 63 years ± 17 [standard deviation]; 639 men) was extracted accurately and allowed for PAOICBS calculation in 304 of 357 (85.2%) PE-positive reports. The PAOICBS strongly correlated with the PAOIQ (r = 0.94; P < .001). Use of PAOICBS yielded overall time savings (1.3 minutes ± 0.5 vs 3.0 minutes ± 1.7), higher confidence levels (4.2 ± 0.6 vs 3.6 ± 1.0), and a higher ICC (ICC, 0.99 vs 0.95), respectively, compared with PAOIQ (each, P < .001). AUCs were similar for PAOICBS (AUC, 0.75; 95% CI: 0.70, 0.81) and PAOIQ (AUC, 0.77; 95% CI: 0.72, 0.83; P = .68), with cutoff values of 27.5% for both scores. Conclusion Data mining of structured reports enabled the development of a CT angiography scoring system that simplified the Qanadli score as a semiquantitative estimate of thrombus burden in patients with pulmonary embolism. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Hunsaker in this issue.

Presentation and outcome of mixed neuroendocrine non-neuroendocrine neoplasms of the pancreas.

BACKGROUND/OBJECTIVES: Mixed neuroendocrine non-neuroendocrine neoplasms (MiNEN) of the pancreas and periampullary region are extremely rare and heterogeneous malignancies. Literature is sparse, clinical management is not standardized and little is known about survival outcomes. The aim of this study was to identify pathological and radiological features of MiNEN and assess the outcome of surgical management. METHODS: Patients undergoing surgery for pancreatic and periampullary MiNEN between 2001 and 2019 were retrospectively analysed based on a prospective database. Histological, radiological and clinical features were assessed. Survival was analysed in a nested case-control study and matched-pair analyses with pure neuroendocrine neoplasms (pNEN) and ductal adeno- or acinar cell carcinomas of the pancreas. A literature review with focus on survival after surgical resection was additionally performed. RESULTS: Of 13 patients with MiNEN, 5 had acinar-MiNEN and 8 adeno-MiNEN. Two of 5 (40%) acinar-MiNEN and one adeno-MiNEN patients had liver metastases. All but one adeno-MiNEN (88%) showed preoperative radiological features of pancreatic adenocarcinoma, 3 of 5 (60%) acinar-MiNEN exhibited mainly neuroendocrine features. No surgical mortality was observed. The 5-year overall survival rate in all MiNEN was 40%. Five-year survival rate was 58% in adeno-MiNEN and comparable to that of matched ductal adenocarcinomas (36%) and pNEN (48%). Five-year overall survival rate was 20% in acinar-MiNEN, compared to 39% in acinar carcinoma patients and 59% in matched pNEN patients. CONCLUSIONS: MiNEN are rare and difficult to distinguish from pure adenocarcinoma or neuroendocrine neoplasm preoperatively. Surgical resection would therefore be the treatment of choice in localized tumors.

[Diagnostic imaging of malignant lymphomas]. / Bildgebende Diagnostik von malignen Lymphomen.

BACKROUND: Malignant lymphomas represent approximately 5% of all cancers. Imaging procedures play a crucial role concerning initial staging and assessment of the response to treatment. OBJECTIVE: This article gives an overview of the significance of imaging procedures in the treatment of patients with malignant lymphomas at various times during treatment. These include the initial assessment of the extent of the disease and staging during and after treatment under consideration of the current classification systems. MATERIAL AND METHODS: A selective literature search was carried out with analysis of dedicated original research articles and reviews as well as a discussion of the clinical guidelines. RESULTS: Computed tomography (CT) is the basic diagnostic tool in patients with malignant lymphomas. Particularly important is fluorodeoxyglucose (FDG) positron emission tomography (PET) CT, which enables a more accurate stage definition and a better assessment of the response to treatment in FDG-avid lymphoma subtypes. Using the FDG-PET/CT-based Deauville score persisting disease activity can be identified in residual masses and refractory disease can be distinguished from complete metabolic remission. The use of magnetic resonance imaging (MRI) with diffusion-weighted imaging can represent a future alternative but is, however, not yet sufficiently standardized and validated. CONCLUSION: The standardized analysis and reporting of purely morphological and metabolic imaging procedures is the backbone of treatment decisions in patients with malignant lymphomas.

Characterization of arthralgia induced by PD-1 antibody treatment in patients with metastasized cutaneous malignancies.

BACKGROUND: PD-1 antibodies (PD1ab) are increasingly used in metastatic melanoma and other malignancies. Arthralgia is an underestimated side effect of PD-1 antibody treatment with unknown cause. Our aim was to characterize PD1ab-induced arthralgia. PATIENTS AND METHODS: We retrospectively included patients with metastatic cutaneous malignancies treated with pembrolizumab or nivolumab ± ipilimumab at the National Center for Tumor Diseases (Heidelberg) between 01/2013 and 09/2016. Arthralgia was characterized by laboratory diagnostics, imaging, and if indicated, rheumatologic consultation. RESULTS: 26 of 195 patients (13.3%) developed arthralgia. The median onset of symptoms was 100 days (7-780 days). Most frequently, arthralgia involved large joints (shoulders, knees) in a predominantly symmetrical pattern. Only two patients were seropositive for rheumatoid factor and/or anti-citrullinated protein antibodies. Ten patients developed the clinical picture of arthritis, with seven of them showing synovitis in MRI or PET/CT. Five patients showed inflammation in joints pre-damaged by osteoarthritis. In 11 patients arthralgia could not be specified. The majority of patients was satisfactorily treated with non-steroidal anti-inflammatory drugs (NSAIDs), 23.1% required additional low-dose corticosteroids and only 7.6% of our patients received further immunosuppressive treatment. Patients with arthralgia showed a better treatment response and improved PFS and OS. CONCLUSION: Arthralgia is frequent during PD1ab treatment. The clinical picture varies between synovitis of predominantly large joints, progressive osteoarthritis and arthralgia without evident joint damage. Vast majority of cases can be satisfactorily managed by NSAID and/or low-dose corticosteroids.

Clinical management of cystic echinococcosis: state of the art and perspectives.

PURPOSE OF REVIEW: This review draws attention to patients with cystic echinococcosis migrating from highly endemic to non-endemic countries with limited experience in cystic echinococcosis management, to ultrasound-based cyst staging, and to the WHO cyst classification as a powerful, to date underused tool to triage patients into the four currently available treatment modalities. RECENT FINDINGS: In non- and low-prevalence countries, cystic echinococcosis is often misclassified. Differential diagnoses, such as simple cysts and other benign and malignant space-occupying lesions, have similar appearances on imaging. Serology is confirmatory but often disappointing due to sensitivity and specificity problems. There is increasing confidence in assigning uncomplicated cystic echinococcosis cysts to the four treatment modalities [drug treatment (benzimidazoles), percutaneous methods, surgery, watch and wait] on the basis of cyst stage (WHO cyst classification), size and location. However, current best practice is still not widely implemented outside cystic echinococcosis treatment centres, and further consolidation is needed by well-designed clinical trials. Recently published long-term follow-up studies have shown that patients with inactive cyst stages CE4 and CE5 benefit, especially since they do not need any treatment at all if they have not received prior benzimidazole or percutaneous therapy. Instead, cysts that have reached cyst stages CE4 and CE5 through drug treatment do need careful follow-up as they often relapse. Surgical procedures in which the cysts are opened and percutaneous approaches require very careful control of cyst content spillage to the peritoneum, pleura and intravascularly to prevent dissemination - a still too often neglected issue with severe consequences for patients. SUMMARY: Though a neglected disease with predominantly expert-opinion-based treatment strategies, cystic echinococcosis patients can greatly benefit from interdisciplinary management in cystic echinococcosis treatment centres and cyst-stage-based assignment of treatment modalities. Increased suspicion for cystic echinococcosis in migrants from endemic to non-endemic countries is urgently needed at the current level of global mobility.

Frequency of abdominal aortic expansion after thoracic endovascular repair of type B aortic dissection.

PURPOSE: To determine abdominal aortic expansion after thoracic endovascular aortic repair (TEVAR) in patients with aortic dissection type B and 36 months minimum follow-up. METHODS: Retrospective study of 18 TEVAR patients with follow-up >36 months. Abdominal aortic diameters at celiac trunk (location B) and infrarenal aorta (location C) were recorded on the first and last imaging after TEVAR. False lumen thrombosis was determined at level of endograft (A) and at B and C. Aortic expansion was defined as diameter increase of 5 mm or 15%. Correlation analyses were performed to investigate potential determinants of expansion. RESULTS: Median follow-up was 75.2 months. Sixteen of 18 patients (88.9%) demonstrated abdominal expansion. Mean expansion was 9.9 ± 6.1 mm at B and 11.7 ± 6.5 mm at C, without a difference between acute and chronic dissections. Critical diameters of 55 mm were reached in two patients treated for chronic dissection (11.1%). Annual diameter increase was significantly greater at locations with baseline diameters >30 mm (2.1 ± 1.1 mm vs. 1.0 ± 0.6 mm, p = 0.009). Baseline diameters were greater in patients with chronic dissections. CONCLUSION: Abdominal aortic expansion can be frequently recognized after TEVAR for aortic dissection type B and occurs independently from thoracic false lumen thrombosis. Clinical significant abdominal aortic expansion may occur more frequently in patients treated with TEVAR for chronic dissection.

Imaging differentiation of solid pseudopapillary neoplasms and neuroendocrine neoplasms of the pancreas.

Purpose: The present study aimed to compare the computed tomography (CT) and magnetic resonance imaging (MRI) features of solid pseudopapillary neoplasms (SPNs) and pancreatic neuroendocrine neoplasms (pNENs). Method: Lesion imaging features of 39 patients with SPNs and 127 patients with pNENs were retrospectively extracted from 104 CT and 91 MRI scans. Results: Compared to pNEN patients, SPN patients were significantly younger (mean age 51.8 yrs versus 32.7 yrs) and more often female (female: male ratio, 5.50:1 versus 1.19:1). Most SPNs and pNENs presented as well-defined lesions with an expansive growth pattern. SPNs more often appeared as round or ovoid lesions, compared to pNENs which showed a lobulated or irregular shape in more than half of cases (p<0.01). A surrounding capsule was detected in the majority of SPNs, but only in a minority of pNENs (<0.01). Hemorrhage occurred non-significantly more often in SPNs (p=0.09). Signal inhomogeneity in T1-fat-saturated (p<0.01) and T2-weighted imaging (p=0.046) as well as cystic degeneration (p<0.01) were more often observed in SPNs. Hyperenhancement in the arterial and portal-venous phase was more common in pNENs (p<0.01). Enlargement of locoregional lymph nodes (p<0.01) and liver metastases (p=0.03) were observed in some pNEN patients, but not in SPN patients. Multivariate logistic regression identified the presence of a capsule (p<0.01), absence of arterial hyperenhancement (p<0.01), and low patient age (p<0.01), as independent predictors for SPN. Conclusions: The present study provides three key features for differentiating SPNs from pNENs extracted from a large patient cohort: presence of a capsule, absence of arterial hyperenhancement, and low patient age.

Improvement of Breast Cancer Detection Using Dual-Layer Spectral CT.

This study aimed to investigate the diagnostic performance of breast mass detection on monoenergetic image data at 40 keV (MonoE40) and on iodine maps (IM) compared with conventional image data (CI). In this prospective single-center case-control study, 50 breast cancer patients were examined using contrast-enhanced dual-layer spectral CT. For qualitative and quantitative comparison of MonoE40 and IM with CI image data, four blinded, independent readers assessed 300 randomized single slices (two slices for each imaging type per case) with or without cancerous lesions for the presence of a breast mass. Detection sensitivity and specificity were calculated and readers rated their subjective diagnostic certainty. For statistical analysis of sensitivity and specificity, a paired t-test and ANOVA were used (significance level p = 0.05). A total of 50 female patients (median age 51 years, range 28-83 years) participated. IM had the highest overall scores in sensitivity and specificity for breast cancer detection, with 0.97 ± 0.06 and 0.95 ± 0.07, respectively, compared with 0.90 ± 0.04 and 0.92 ± 0.06 in CI. MonoE40 yielded a sensitivity of 0.96 ± 0.02 and specificity of 0.94 ± 0.08. All differences in sensitivity and specificity between MonoE or IM and CI were statistically significant (p < 0.001). The superiority of IM sensitivity and specificity was most pronounced in patients with dense breasts. Spectral CT improved the detection of breast cancer with higher sensitivity and specificity compared to conventional image data in our study.

Calcium-Based Imaging of the Spine at Dual-Layer CT and Evaluation of Vertebral Fractures in Multiple Myeloma.

PURPOSE: To evaluate the prediction of vertebral fractures in plasma cell dyscrasias using dual-layer CT (DLCT) with quantitative assessment of conventional CT image data (CI), calcium suppressed image data (CaSupp), and calculation of virtual calcium-only (VCa) image data. MATERIAL AND METHODS: Patients (n = 81) with the diagnosis of a plasma cell dyscrasia and whole-body DLCT at the time of diagnosis and follow-up were retrospectively enrolled. CI, CaSupp25, and CaSupp100 were quantitatively analyzed using regions of interest in the lumbar vertebral bodies and fractured vertebral bodies on baseline or follow-up imaging. VCa were calculated by subtraction (CaSupp100-CaSupp25), delineating bone only. Logistic regression analyses were performed to assess the possibility of imminent spine fractures. RESULTS: In 24 patients, new vertebral fractures were observed in the follow-up imaging. The possibility of new vertebral fractures was significant for baseline assessment of CT numbers in CI, CaSupp25, and VCa (p = 0.01, respectively), with a higher risk for new fractures in the case of lower CT numbers in CI (Odds ratio = [0.969; 0.994]) and VCa (Odds ratio = [0.978; 0.995]) and in the case of higher CT numbers in CaSupp 25 (Odds ratio 1.015 [1.006; 1.026]). Direct model comparisons implied that CT numbers in CaSupp 25 and VCa might show better fracture prediction than those in CI (R2 = 0.18 both vs. 0.15; AICc = 91.95, 91.79 vs. 93.62), suggesting cut-off values for CI at 103 HU (sensitivity: 54.2%; specificity: 82.5; AUC: 0.69), for VCa at 129 HU (sensitivity: 41.7%; specificity: 94.7; AUC: 0.72). CONCLUSIONS: Quantitative assessment with CaSupp and calculation of VCa is feasible to predict the vertebral fracture risk in MM patients. DLCT may prove useful in detecting imminent fractures.

Feasibility of Reduced Iodine Loads for Vascular Assessment Prior to Transcatheter Aortic Valve Implantation (TAVI) Using Spectral Detector CT.

Reduced iodine loads for computed tomography (CT)-based vascular assessment prior to transcatheter aortic valve implantation (TAVI) may be feasible in conjunction with a spectral detector CT scanner. This prospective single-center study considered 100 consecutive patients clinically referred for pre-TAVI CT. They were examined on a dual-layer detector CT scanner to obtain an ECG-gated cardiac scan and a non-ECG-gated aortoiliofemoral scan. Either a standard contrast media (SCM) protocol using 80 mL Iohexol 350 mgI/mL (iodine load: 28 gI) or a body-mass-index adjusted reduced contrast media (RCM) protocol using 40-70 mL Iohexol 350 mgI/mL (iodine load: 14-24.5 gI) were employed. Conventional images and virtual monoenergetic images at 40-80 keV were reconstructed. A threshold of 250 HU was set for sufficient attenuation along the arterial access pathway. A qualitative assessment used a five-point Likert scale. Sufficient attenuation in the thoracic aorta was observed for all patients in both groups using conventional images. In the abdominal, iliac, and femoral segments, sufficient attenuation was observed for the majority of patients when using virtual monoenergetic images (SCM: 96-100% of patients, RCM: 88-94%) without statistical difference between both groups. Segments with attenuation measurements below the threshold remained qualitatively assessable as well. Likert scores were 'excellent' for virtual monoenergetic images 50 keV and 55 keV in both groups (RCM: 1.2-1.4, SCM: 1.2-1.3). With diagnostic image quality maintained, it can be concluded that reduced iodine loads of 14-24.5 gI are feasible for pre-TAVI vascular assessment on a spectral detector CT scanner.

Successful treatment with nab-paclitaxel and gemcitabine after FOLFIRINOX failure in a patient with metastasized pancreatic adenocarcinoma.

BACKGROUND: Advanced pancreatic adenocarcinoma still remains associated with a desperate prognosis. Nevertheless, treatment options for patients with metastasized disease have improved considerably over the last few years. Recently, cytotoxic combination therapies such as the FOLFIRINOX regimen and combined nab-paclitaxel/gemcitabine have shown improved overall survival compared to gemcitabine alone. There is still no standard of care in second-line therapy for patients with disease progression. CASE REPORT: We report the case of a 47-year-old patient who dramatically responded to second-line treatment with nab-paclitaxel and gemcitabine after primary progression to the FOLFIRINOX protocol. CONCLUSION: Second-line treatment after FOLFIRINOX is feasible for patients with good performance status. Our case report supports preclinical findings that suggest that pancreatic cancer is a heterogeneous disease. Further studies that characterize possible subgroups and identify predictive molecular markers to guide therapy are warranted.

Vascular pathology in patients with alveolar echinococcosis: framework for assessment and clinical management - a retrospective case series.

OBJECTIVE: Alveolar echinococcosis (AE) is a parasitic liver disease with infiltrative growth similar to solid organ malignancies. Major vascular damage is frequent and often remains untreated until catastrophic events precipitate. Detailed clinical and radiological assessment is required to guide individualised treatment decisions. Standardised radiological reporting templates of malignancies with profiles resembling AE are candidates for adaptation. Our objectives are to describe vascular pathology in AE and establish a framework for structured evaluation as the basis for treatment decisions and monitoring. DESIGN: Retrospective case series. RESULTS: 69 patients (37.1%) had vascular involvement: portal vein (PV) 24.7%, hepatic vein (HV) 22.6% inferior vena cava (IVC) 13.4%. Significant stenosis/occlusion of vessels was present in 15.1% of PV, in 13.4% of HV and in 7.5% of IVC involvement. Vascular pathology needing specific treatment or monitoring was present in 8.6% of patients. The most frequent clinical presentation was high grade IVC stenosis or occlusion which was seen in 11 patients of the cohort. CONCLUSION: Advanced AE requires early multidisciplinary assessment to prevent progressive impairment of liver function due to vascular damage. The focus at first presentation is on complete evaluation of vascular (and biliary) involvement. The focus in non-resectable AE is on prevention of vascular (and biliary) complications while suppressing growth of AE lesions by benzimidazole treatment to improve the quality of life of patients. We developed a framework for standardised vascular assessment and follow-up of patients with AE to recognise and treat complications early.

In Vivo Repeatability and Multiscanner Reproducibility of MRI Radiomics Features in Patients With Monoclonal Plasma Cell Disorders: A Prospective Bi-institutional Study.

OBJECTIVES: Despite the extensive number of publications in the field of radiomics, radiomics algorithms barely enter large-scale clinical application. Supposedly, the low external generalizability of radiomics models is one of the main reasons, which hinders the translation from research to clinical application. The objectives of this study were to investigate reproducibility of radiomics features (RFs) in vivo under variation of patient positioning, magnetic resonance imaging (MRI) sequence, and MRI scanners, and to identify a subgroup of RFs that shows acceptable reproducibility across all different acquisition scenarios. MATERIALS AND METHODS: Between November 30, 2020 and February 16, 2021, 55 patients with monoclonal plasma cell disorders were included in this prospective, bi-institutional, single-vendor study. Participants underwent one reference scan at a 1.5 T MRI scanner and several retest scans: once after simple repositioning, once with a second MRI protocol, once at another 1.5 T scanner, and once at a 3 T scanner. Radiomics feature from the bone marrow of the left hip bone were extracted, both from original scans and after different image normalizations. Intraclass correlation coefficient (ICC) was used to assess RF repeatability and reproducibility. RESULTS: Fifty-five participants (mean age, 59 ± 7 years; 36 men) were enrolled. For T1-weighted images after muscle normalization, in the simple test-retest experiment, 110 (37%) of 295 RFs showed an ICC ≥0.8: 54 (61%) of 89 first-order features (FOFs), 35 (95%) of 37 volume and shape features, and 21 (12%) of 169 texture features (TFs). When the retest was performed with different technical settings, even after muscle normalization, the number of FOF/TF with an ICC ≥0.8 declined to 58/13 for the second protocol, 29/7 for the second 1.5 T scanner, and 49/7 for the 3 T scanner, respectively. Twenty-five (28%) of the 89 FOFs and 6 (4%) of the 169 TFs from muscle-normalized T1-weighted images showed an ICC ≥0.8 throughout all repeatability and reproducibility experiments. CONCLUSIONS: In vivo, only few RFs are reproducible with different MRI sequences or different MRI scanners, even after application of a simple image normalization. Radiomics features selected by a repeatability experiment only are not necessarily suited to build radiomics models for multicenter clinical application. This study isolated a subset of RFs, which are robust to variations in MRI acquisition observed in scanners from 1 vendor, and therefore are candidates to build reproducible radiomics models for monoclonal plasma cell disorders for multicentric applications, at least when centers are equipped with scanners from this vendor.

Prediction of Bone Marrow Biopsy Results From MRI in Multiple Myeloma Patients Using Deep Learning and Radiomics.

OBJECTIVES: In multiple myeloma and its precursor stages, plasma cell infiltration (PCI) and cytogenetic aberrations are important for staging, risk stratification, and response assessment. However, invasive bone marrow (BM) biopsies cannot be performed frequently and multifocally to assess the spatially heterogenous tumor tissue. Therefore, the goal of this study was to establish an automated framework to predict local BM biopsy results from magnetic resonance imaging (MRI). MATERIALS AND METHODS: This retrospective multicentric study used data from center 1 for algorithm training and internal testing, and data from center 2 to 8 for external testing. An nnU-Net was trained for automated segmentation of pelvic BM from T1-weighted whole-body MRI. Radiomics features were extracted from these segmentations, and random forest models were trained to predict PCI and the presence or absence of cytogenetic aberrations. Pearson correlation coefficient and the area under the receiver operating characteristic were used to evaluate the prediction performance for PCI and cytogenetic aberrations, respectively. RESULTS: A total of 672 MRIs from 512 patients (median age, 61 years; interquartile range, 53-67 years; 307 men) from 8 centers and 370 corresponding BM biopsies were included. The predicted PCI from the best model was significantly correlated ( P ≤ 0.01) to the actual PCI from biopsy in all internal and external test sets (internal test set: r = 0.71 [0.51, 0.83]; center 2, high-quality test set: r = 0.45 [0.12, 0.69]; center 2, other test set: r = 0.30 [0.07, 0.49]; multicenter test set: r = 0.57 [0.30, 0.76]). The areas under the receiver operating characteristic of the prediction models for the different cytogenetic aberrations ranged from 0.57 to 0.76 for the internal test set, but no model generalized well to all 3 external test sets. CONCLUSIONS: The automated image analysis framework established in this study allows for noninvasive prediction of a surrogate parameter for PCI, which is significantly correlated to the actual PCI from BM biopsy.

Deep Learning-based Assessment of Oncologic Outcomes from Natural Language Processing of Structured Radiology Reports.

Purpose: To train a deep natural language processing (NLP) model, using data mined structured oncology reports (SOR), for rapid tumor response category (TRC) classification from free-text oncology reports (FTOR) and to compare its performance with human readers and conventional NLP algorithms. Materials and Methods: In this retrospective study, databases of three independent radiology departments were queried for SOR and FTOR dated from March 2018 to August 2021. An automated data mining and curation pipeline was developed to extract Response Evaluation Criteria in Solid Tumors-related TRCs for SOR for ground truth definition. The deep NLP bidirectional encoder representations from transformers (BERT) model and three feature-rich algorithms were trained on SOR to predict TRCs in FTOR. Models' F1 scores were compared against scores of radiologists, medical students, and radiology technologist students. Lexical and semantic analyses were conducted to investigate human and model performance on FTOR. Results: Oncologic findings and TRCs were accurately mined from 9653 of 12 833 (75.2%) queried SOR, yielding oncology reports from 10 455 patients (mean age, 60 years ± 14 [SD]; 5303 women) who met inclusion criteria. On 802 FTOR in the test set, BERT achieved better TRC classification results (F1, 0.70; 95% CI: 0.68, 0.73) than the best-performing reference linear support vector classifier (F1, 0.63; 95% CI: 0.61, 0.66) and technologist students (F1, 0.65; 95% CI: 0.63, 0.67), had similar performance to medical students (F1, 0.73; 95% CI: 0.72, 0.75), but was inferior to radiologists (F1, 0.79; 95% CI: 0.78, 0.81). Lexical complexity and semantic ambiguities in FTOR influenced human and model performance, revealing maximum F1 score drops of -0.17 and -0.19, respectively. Conclusion: The developed deep NLP model reached the performance level of medical students but not radiologists in curating oncologic outcomes from radiology FTOR.Keywords: Neural Networks, Computer Applications-Detection/Diagnosis, Oncology, Research Design, Staging, Tumor Response, Comparative Studies, Decision Analysis, Experimental Investigations, Observer Performance, Outcomes Analysis Supplemental material is available for this article. © RSNA, 2022.

The Endurant Stent Graft System: 15-month follow-up report in patients with challenging abdominal aortic anatomies.

OBJECTIVE: The objective of this study is to report a 15-month follow-up with the Endurant Stent Graft System in patients with challenging aortic anatomies. METHODS: At three German clinics, a consecutive series of 50 patients underwent endovascular abdominal aortic repair (EVAR) for challenging abdominal aortic aneurysm with the Endurant stent graft between November 2008 and May 2009. EVAR was elective in 48 cases and emergent in two. Patients had short (≤15 mm) aortic necks, severe suprarenal/infrarenal angulation, and/or small (<8 mm), calcified, severely angulated, or tortuous iliac or femoral access vessels. Additionally, a cohort of 40 patients without challenging anatomies were retrospectively analysed to clarify differences concerning technical success, mortality, and morbidity between these groups. RESULTS: The primary technical success rate was 92% (46/50). The 30-day mortality rate was 2% (1/50), the death due to multiorgan failure. Intraoperative angiograms revealed three type I endoleaks (2 proximal and 1 distal), and one of those was persisting at 30 days (30-day rate, 2%). Postoperative imaging discovered no further type I or type III endoleaks. The 30-day rate of the type II endoleak was 6% (3/50). There were two cases of graft limb occlusion, both requiring reintervention within 30 days. Follow-up was available in all of the 50 patients (100%) over a median of 15 months (1-25). During this time, seven patients died (overall mortality, 16%; 8/50), besides the above-described patient, all of them unrelated to the procedure. Compared to the 30-day results with the Endurant stent graft in non-challenging anatomies (no type I endoleak; no graft limb occlusion; all-cause mortality, 0%), procedure-related complications in challenging anatomies are increasing. CONCLUSION: Early and 15-month results with the Endurant stent graft in patients with challenging aortic anatomies are encouraging.

Consolidated lung on contrast-enhanced chest CT: the use of spectral-detector computed tomography parameters in differentiating atelectasis and pneumonia.

OBJECTIVES: To investigate the value of spectral-detector computed tomography (SDCT) parameters for the quantitative differentiation between atelectasis and pneumonia on contrast-enhanced chest CT. MATERIAL AND METHODS: Sixty-three patients, 22 clinically diagnosed with pneumonia and 41 with atelectasis, underwent contrast-enhanced SDCT scans during the venous phase. CT numbers (Hounsfield Units [HU]) were measured on conventional reconstructions (CON120kVp) and the iodine concentration (Ciodine, [mg/ml]), and effective atomic number (Zeff) on spectral reconstructions, using region-of-interest (ROI) analysis. Receiver operating characteristics (ROC) and contrast-to-noise ratios (CNRs) were calculated to assess each reconstruction's potential to differentiate between atelectasis and pneumonia. RESULTS: On contrast-enhanced SDCT, the difference between atelectasis and pneumonia was significant on CON120kVp, Ciodine, and Zeff images (p < 0.001). On CON120kVp images, a threshold of 81 HU achieved a sensitivity of 93 % and a specificity of 95 % for identifying pneumonia, while Ciodine and Zeff images reached the same sensitivity but lower specificities of 85 % and 83 %. CON120kVp images showed significantly higher CNRs between normal lung and atelectasis or pneumonia with 30.63 and 27.69 compared to Ciodine images with 3.54 and 1.27 and Zeff images with 4.22 and 7.63 (p < 0.001). None of the parameters could differentiate atelectasis and pneumonia without contrast media. CONCLUSIONS: Contrast-enhanced SDCT can differentiate atelectasis and pneumonia based on the spectral parameters Ciodine, and Zeff. However, they had no added value compared to CT number measurement on CON120kVp images. Furthermore, contrast media is still needed for a differentiation based on quantitative SDCT parameters.