Improved Productivity Using Deep Learning-assisted Reporting for Lumbar Spine MRI.

Background Lumbar spine MRI studies are widely used for back pain assessment. Interpretation involves grading lumbar spinal stenosis, which is repetitive and time consuming. Deep learning (DL) could provide faster and more consistent interpretation. Purpose To assess the speed and interobserver agreement of radiologists for reporting lumbar spinal stenosis with and without DL assistance. Materials and Methods In this retrospective study, a DL model designed to assist radiologists in the interpretation of spinal canal, lateral recess, and neural foraminal stenoses on lumbar spine MRI scans was used. Randomly selected lumbar spine MRI studies obtained in patients with back pain who were 18 years and older over a 3-year period, from September 2015 to September 2018, were included in an internal test data set. Studies with instrumentation and scoliosis were excluded. Eight radiologists, each with 2-13 years of experience in spine MRI interpretation, reviewed studies with and without DL model assistance with a 1-month washout period. Time to diagnosis (in seconds) and interobserver agreement (using Gwet κ) were assessed for stenosis grading for each radiologist with and without the DL model and compared with test data set labels provided by an external musculoskeletal radiologist (with 32 years of experience) as the reference standard. Results Overall, 444 images in 25 patients (mean age, 51 years ± 20 [SD]; 14 women) were evaluated in a test data set. DL-assisted radiologists had a reduced interpretation time per spine MRI study, from a mean of 124-274 seconds (SD, 25-88 seconds) to 47-71 seconds (SD, 24-29 seconds) (P < .001). DL-assisted radiologists had either superior or equivalent interobserver agreement for all stenosis gradings compared with unassisted radiologists. DL-assisted general and in-training radiologists improved their interobserver agreement for four-class neural foraminal stenosis, with κ values of 0.71 and 0.70 (with DL) versus 0.39 and 0.39 (without DL), respectively (both P < .001). Conclusion Radiologists who were assisted by deep learning for interpretation of lumbar spinal stenosis on MRI scans showed a marked reduction in reporting time and superior or equivalent interobserver agreement for all stenosis gradings compared with radiologists who were unassisted by deep learning. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Hayashi in this issue.

Effect of Training Data Volume on Performance of Convolutional Neural Network Pneumothorax Classifiers.

Large datasets with high-quality labels required to train deep neural networks are challenging to obtain in the radiology domain. This work investigates the effect of training dataset size on the performance of deep learning classifiers, focusing on chest radiograph pneumothorax detection as a proxy visual task in the radiology domain. Two open-source datasets (ChestX-ray14 and CheXpert) comprising 291,454 images were merged and convolutional neural networks trained with stepwise increase in training dataset sizes. Model iterations at each dataset volume were evaluated on an external test set of 525 emergency department chest radiographs. Learning curve analysis was performed to fit the observed AUCs for all models generated. For all three network architectures tested, model AUCs and accuracy increased rapidly from 2 × 103 to 20 × 103 training samples, with more gradual increase until the maximum training dataset size of 291 × 103 images. AUCs for models trained with the maximum tested dataset size of 291 × 103 images were significantly higher than models trained with 20 × 103 images: ResNet-50: AUC20k = 0.86, AUC291k = 0.95, p < 0.001; DenseNet-121 AUC20k = 0.85, AUC291k = 0.93, p < 0.001; EfficientNet AUC20k = 0.92, AUC 291 k = 0.98, p < 0.001. Our study established learning curves describing the relationship between dataset training size and model performance of deep learning convolutional neural networks applied to a typical radiology binary classification task. These curves suggest a point of diminishing performance returns for increasing training data volumes, which algorithm developers should consider given the high costs of obtaining and labelling radiology data.

Deep Learning Model for Automated Detection and Classification of Central Canal, Lateral Recess, and Neural Foraminal Stenosis at Lumbar Spine MRI.

Background Assessment of lumbar spinal stenosis at MRI is repetitive and time consuming. Deep learning (DL) could improve -productivity and the consistency of reporting. Purpose To develop a DL model for automated detection and classification of lumbar central canal, lateral recess, and neural -foraminal stenosis. Materials and Methods In this retrospective study, lumbar spine MRI scans obtained from September 2015 to September 2018 were included. Studies of patients with spinal instrumentation or studies with suboptimal image quality, as well as postgadolinium studies and studies of patients with scoliosis, were excluded. Axial T2-weighted and sagittal T1-weighted images were used. Studies were split into an internal training set (80%), validation set (9%), and test set (11%). Training data were labeled by four radiologists using predefined gradings (normal, mild, moderate, and severe). A two-component DL model was developed. First, a convolutional neural network (CNN) was trained to detect the region of interest (ROI), with a second CNN for classification. An internal test set was labeled by a musculoskeletal radiologist with 31 years of experience (reference standard) and two subspecialist radiologists (radiologist 1: A.M., 5 years of experience; radiologist 2: J.T.P.D.H., 9 years of experience). DL model performance on an external test set was evaluated. Detection recall (in percentage), interrater agreement (Gwet κ), sensitivity, and specificity were calculated. Results Overall, 446 MRI lumbar spine studies were analyzed (446 patients; mean age ± standard deviation, 52 years ± 19; 240 women), with 396 patients in the training (80%) and validation (9%) sets and 50 (11%) in the internal test set. For internal testing, DL model and radiologist central canal recall were greater than 99%, with reduced neural foramina recall for the DL model (84.5%) and radiologist 1 (83.9%) compared with radiologist 2 (97.1%) (P < .001). For internal testing, dichotomous classification (normal or mild vs moderate or severe) showed almost-perfect agreement for both radiologists and the DL model, with respective κ values of 0.98, 0.98, and 0.96 for the central canal; 0.92, 0.95, and 0.92 for lateral recesses; and 0.94, 0.95, and 0.89 for neural foramina (P < .001). External testing with 100 MRI scans of lumbar spines showed almost perfect agreement for the DL model for dichotomous classification of all ROIs (κ, 0.95-0.96; P < .001). Conclusion A deep learning model showed comparable agreement with subspecialist radiologists for detection and classification of central canal and lateral recess stenosis, with slightly lower agreement for neural foraminal stenosis at lumbar spine MRI. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Hayashi in this issue.

Cumulative iodinated contrast exposure for computed tomography during acute kidney injury and major adverse kidney events.

OBJECTIVES: To determine if contrast-enhanced CT imaging performed in patients during their episode of AKI contributes to major adverse kidney events (MAKE). METHODS: A propensity score-matched analysis of 1127 patients with AKI defined by KDIGO criteria was done. Their mean age was 63 ± 16 years with 56% males. A total of 419 cases exposed to CT contrast peri-AKI were matched with 798 non-exposed controls for 14 covariates including comorbidities, acute illnesses, and initial AKI severity; outcomes including MAKE and renal recovery in hospital were compared using bivariate analysis and logistic regression. MAKE was a composite of mortality, renal replacement therapy, and doubling of serum creatinine on discharge over baseline; renal recovery was classified as early versus late based on a 7-day timeline from AKI onset to nadir creatinine or cessation of renal replacement therapy in survivors. RESULTS: Sixty-two patients received cumulatively > 100 mL of CT contrast, 143 patients had > 50-100 mL, and 214 patients had 50 mL or less; MAKE occurred in 34%, 17%, and 21%, respectively, as compared with 20% in non-exposed controls (p = 0.008 for patients with > 100 mL contrast versus none). More contrast-exposed patients experienced late renal recovery (27% versus 20%) and longer hospital days (median 10 versus 8) than non-exposed patients (all p < 0.01). On multivariate analysis, cumulative CT contrast > 100 mL was independently associated with MAKE (odds ratio 2.39 versus non-contrast, adjusted for all confounders, p = 0.005); cumulative CT contrast under 100 mL was not associated with MAKE. CONCLUSIONS: High cumulative volume of CT contrast administered to patients with AKI is associated with worse short-term renal outcomes and delayed renal recovery. KEY POINTS: • Cumulative intravenous iodinated contrast for CT imaging of more than 100 mL, during an episode of acute kidney injury, was independently associated with worse renal outcomes and less renal recovery. • These adverse outcomes including renal replacement therapy were not more frequent in similar patients who received cumulatively 100 mL or less of CT contrast, compared with non-exposed patients. • More patients with CT contrast exposure during acute kidney injury experienced delayed renal recovery.

Value of a molecular screening program to support clinical trial enrollment in Asian cancer patients: The Integrated Molecular Analysis of Cancer (IMAC) Study.

The value of precision oncology initiatives in Asian contexts remains unresolved. Here, we review the institutional implementation of prospective molecular screening to facilitate accrual of patients into biomarker-driven clinical trials, and to explore the mutational landscape of advanced tumors occurring in a prospective cohort of Asian patients (n = 396) with diverse cancer types. Next-generation sequencing (NGS) and routine clinicopathological assays, such as immunohistochemistry, copy number analysis and in situ hybridization tests, were performed on tumor samples. Actionable biomarker results were used to identify eligibility for early-phase, biomarker-driven clinical trials. Overall, NGS was successful in 365 of 396 patients (92%), achieving a mean depth of 1,943× and coverage uniformity of 96%. The median turnaround time from sample receipt to return of genomic results was 26.0 days (IQR, 19.0-39.0 days). Reportable mutations were found in 300 of 365 patients (82%). Ninety-one percent of patients at study enrollment indicated consent to receive incidental findings and willingness to undergo genetic counseling if required. The most commonly mutated oncogenes included KRAS (19%), PIK3CA (16%), EGFR (5%), BRAF (3%) and KIT (3%); while the most frequently mutated tumor suppressor genes included TP53 (40%), SMARCB1 (12%), APC (8%), PTEN (6%) and SMAD4 (5%). Among 23 patients enrolled in genotype-matched trials, median progression-free survival was 2.9 months (IQR, 1.5-4.0 months). Nine of 20 evaluable patients (45%; 95% CI, 23.1-68.5%) derived clinical benefit, including 3 partial responses and 6 with stable disease lasting ≥ 8 weeks.

Pushing the limits of immune-related response: a case of "extreme pseudoprogression".

The advent of immune checkpoint targeted immunotherapy has seen a spectrum of immune-related phenomena in both tumor responses and toxicities. We describe a case of pseudoprogression that pushes the limits of immune-related response criteria and challenges the boundaries and definitions set by trial protocols. A middle-aged man with conventional clear cell renal cell carcinoma (RCC) had received multiple prior systemic treatments including vascular endothelial growth factor receptor tyrosine kinase inhibitors, as well as multiple surgeries and radiotherapy treatments. He was eventually started on nivolumab-the anti-programmed death receptor-1 monoclonal antibody approved for the treatment of advanced RCC. Clinical deterioration was observed soon after a 100 mg dose of nivolumab, with onset of acute renal failure and declining performance status. Radiologic progression was documented in multiple sites including worsening tumor infiltration of his residual kidney. The patient was on palliative treatment and visited by the home hospice team in an end-of-life situation. The patient unexpectedly improved and went on to achieve a durable tumor response. The case is illustrative of an extreme manifestation of pseudoprogression, and impels us to probe the assumptions and controversies surrounding this phenomenon.

MR Evaluation of an Unusual Intruder of the Bladder: A Bladder Fibroid Case Report.

INTRODUCTION: Mesenchymal tumours of the bladder are benign but rare occurrences and represent approximately 1% of all bladder tumours. CASE REPORT: We report a case of a large bladder leiomyoma in an asymptomatic patient. A large pelvic mass was discovered incidentally on the bedside ultrasound scan during a review at the gynecology clinic. Intra-operatively, no mass was seen in the pelvis, and cystoscopy demonstrated an intravesical mass. It was further evaluated with cystoscopy. MR imaging demonstrated typical features of a bladder leiomyoma. Subsequently, the patient underwent partial cystectomy, and the mass was removed, which was histologically proven leiomyoma. CONCLUSION: Awareness of this rare clinical entity and identification of its typical radiological features on MR imaging can aid with accurate diagnosis and preclude unnecessary radical surgery.

Comprehensive characterization of genomic features and clinical outcomes following targeted therapy and secondary cytoreductive surgery in OCCC: a single center experience.

OBJECTIVE: Ovarian clear cell carcinoma (OCCC) is associated with chemoresistance. Limited data exists regarding the efficacy of targeted therapies such as immune checkpoint inhibitors (ICI) and bevacizumab, and the role of secondary cytoreductive surgery (SCS). METHODS: We retrospectively analyzed genomic features and treatment outcomes of 172 OCCC patients treated at our institution from January 2000 to May 2022. Next-generation sequencing (NGS) was performed where sufficient archival tissue was available. RESULTS: 64.0% of patients were diagnosed at an early stage, and 36.0% at an advanced stage. Patients with advanced/relapsed OCCC who received platinum-based chemotherapy plus bevacizumab followed by maintenance bevacizumab had a median first-line progression-free survival (PFS) of 12.2 months, compared with 9.3 months for chemotherapy alone (hazard ratio=0.69; 95% confidence interval [CI]=0.33, 1.45). In 27 patients who received an ICI, the overall response rate was 18.5% and median duration of response was 7.4 months (95% CI=6.5, 8.3). In 17 carefully selected patients with fewer than 3 sites of relapse, median PFS was 35 months (95% CI=0, 73.5) and median overall survival was 96.8 months (95% CI=44.6, 149.0) after SCS. NGS on 58 tumors revealed common mutations in ARID1A (48.3%), PIK3CA (46.6%), and KRAS (20.7%). Pathogenic alterations in PIK3CA, FGFR2, and NBN were associated with worse survival outcomes. Median tumor mutational burden was 3.78 (range, 0-16). All 26 patients with available loss of heterozygosity (LOH) scores had LOH <16%. CONCLUSION: Our study demonstrates encouraging outcomes with bevacizumab and ICI, and SCS in select relapsed OCCC patients. Prospective trials are warranted.

Detection of Pneumothorax with Deep Learning Models: Learning From Radiologist Labels vs Natural Language Processing Model Generated Labels.

RATIONALE AND OBJECTIVES: To compare the performance of pneumothorax deep learning detection models trained with radiologist versus natural language processing (NLP) labels on the NIH ChestX-ray14 dataset. MATERIALS AND METHODS: The ChestX-ray14 dataset consisted of 112,120 frontal chest radiographs with 5302 positive and 106, 818 negative labels for pneumothorax using NLP (dataset A). All 112,120 radiographs were also inspected by 4 radiologists leaving a visually confirmed set of 5,138 positive and 104,751 negative for pneumothorax (dataset B). Datasets A and B were used independently to train 3 convolutional neural network (CNN) architectures (ResNet-50, DenseNet-121 and EfficientNetB3). All models' area under the receiver operating characteristic curve (AUC) were evaluated with the official NIH test set and an external test set of 525 chest radiographs from our emergency department. RESULTS: There were significantly higher AUCs on the NIH internal test set for CNN models trained with radiologist vs NLP labels across all architectures. AUCs for the NLP/radiologist-label models were 0.838 (95%CI:0.830, 0.846)/0.881 (95%CI:0.873,0.887) for ResNet-50 (p = 0.034), 0.839 (95%CI:0.831,0.847)/0.880 (95%CI:0.873,0.887) for DenseNet-121, and 0.869 (95%CI: 0.863,0.876)/0.943 (95%CI: 0.939,0.946) for EfficientNetB3 (p ≤0.001). Evaluation with the external test set also showed higher AUCs (p <0.001) for the CNN models trained with radiologist versus NLP labels across all architectures. The AUCs for the NLP/radiologist-label models were 0.686 (95%CI:0.632,0.740)/0.806 (95%CI:0.758,0.854) for ResNet-50, 0.736 (95%CI:0.686, 0.787)/0.871 (95%CI:0.830,0.912) for DenseNet-121, and 0.822 (95%CI: 0.775,0.868)/0.915 (95%CI: 0.882,0.948) for EfficientNetB3. CONCLUSION: We demonstrated improved performance and generalizability of pneumothorax detection deep learning models trained with radiologist labels compared to models trained with NLP labels.

Cytokine Release Syndrome in Cancer Patients Receiving Immune Checkpoint Inhibitors: A Case Series of 25 Patients and Review of the Literature.

Cytokine release syndrome (CRS) is a phenomenon of immune hyperactivation described in the setting of immunotherapy. Unlike other immune-related adverse events, CRS triggered by immune checkpoint inhibitors (ICIs) is not well described. The clinical characteristics and course of 25 patients with ICI-induced CRS from 2 tertiary hospitals were abstracted retrospectively from the medical records and analyzed. CRS events were confirmed by 2 independent reviewers and graded using the Lee et al. scale. The median duration of CRS was 15.0 days (Q1; Q3 6.3; 29.8) and 10 (40.0%) had multiple episodes of CRS flares. Comparing the clinical factors and biomarkers in Grades 1-2 and 3-5 CRS, we found that patients with Grades 3-5 CRS had following: (i) had longer time to fever onset [25.0 days (Q1; Q3 13.0; 136.5) vs. 3.0 days (Q1; Q3 0.0; 18.0), p=0.027]; (ii) more cardiovascular (p=0.002), neurologic (p=0.001), pulmonary (p=0.044) and rheumatic (p=0.037) involvement; (iii) lower platelet count (p=0.041) and higher urea (p=0.041) at presentation compared to patients with Grades 1-2 CRS. 7 patients (28.0%) with Grades 1-2 CRS were rechallenged using ICIs without event. 9 patients (36.0%) were treated with pulse methylprednisolone and 6 patients (24.0%) were treated with tocilizumab. Despite this, 3 patients (50%) who received tocilizumab had fatal (Grade 5) outcomes from ICI-induced CRS. Longer time to fever onset, lower platelet count and higher urea at presentation were associated with Grade 3-5 CRS. These parameters may be used to predict which patients are likely to develop severe CRS.

Application of Artificial Intelligence Methods for Imaging of Spinal Metastasis.

Spinal metastasis is the most common malignant disease of the spine. Recently, major advances in machine learning and artificial intelligence technology have led to their increased use in oncological imaging. The purpose of this study is to review and summarise the present evidence for artificial intelligence applications in the detection, classification and management of spinal metastasis, along with their potential integration into clinical practice. A systematic, detailed search of the main electronic medical databases was undertaken in concordance with the PRISMA guidelines. A total of 30 articles were retrieved from the database and reviewed. Key findings of current AI applications were compiled and summarised. The main clinical applications of AI techniques include image processing, diagnosis, decision support, treatment assistance and prognostic outcomes. In the realm of spinal oncology, artificial intelligence technologies have achieved relatively good performance and hold immense potential to aid clinicians, including enhancing work efficiency and reducing adverse events. Further research is required to validate the clinical performance of the AI tools and facilitate their integration into routine clinical practice.

Diagnostic Accuracy of CT for Metastatic Epidural Spinal Cord Compression.

BACKGROUND: Early diagnosis of metastatic epidural spinal cord compression (MESCC) is vital to expedite therapy and prevent paralysis. Staging CT is performed routinely in cancer patients and presents an opportunity for earlier diagnosis. METHODS: This retrospective study included 123 CT scans from 101 patients who underwent spine MRI within 30 days, excluding 549 CT scans from 216 patients due to CT performed post-MRI, non-contrast CT, or a gap greater than 30 days between modalities. Reference standard MESCC gradings on CT were provided in consensus via two spine radiologists (11 and 7 years of experience) analyzing the MRI scans. CT scans were labeled using the original reports and by three radiologists (3, 13, and 14 years of experience) using dedicated CT windowing. RESULTS: For normal/none versus low/high-grade MESCC per CT scan, all radiologists demonstrated almost perfect agreement with kappa values ranging from 0.866 (95% CI 0.787-0.945) to 0.947 (95% CI 0.899-0.995), compared to slight agreement for the reports (kappa = 0.095, 95%CI -0.098-0.287). Radiologists also showed high sensitivities ranging from 91.51 (95% CI 84.49-96.04) to 98.11 (95% CI 93.35-99.77), compared to 44.34 (95% CI 34.69-54.31) for the reports. CONCLUSION: Dedicated radiologist review for MESCC on CT showed high interobserver agreement and sensitivity compared to the current standard of care.

Comparison of metrics for the evaluation of medical segmentations using prostate MRI dataset.

Nine previously proposed segmentation evaluation metrics, targeting medical relevance, accounting for holes, and added regions or differentiating over- and under-segmentation, were compared with 24 traditional metrics to identify those which better capture the requirements for clinical segmentation evaluation. Evaluation was first performed using 2D synthetic shapes to highlight features and pitfalls of the metrics with known ground truths (GTs) and machine segmentations (MSs). Clinical evaluation was then performed using publicly-available prostate images of 20 subjects with MSs generated by 3 different deep learning networks (DenseVNet, HighRes3DNet, and ScaleNet) and GTs drawn by 2 readers. The same readers also performed the 2D visual assessment of the MSs using a dual negative-positive grading of -5 to 5 to reflect over- and under-estimation. Nine metrics that correlated well with visual assessment were selected for further evaluation using 3 different network ranking methods - based on a single metric, normalizing the metric using 2 GTs, and ranking the network based on a metric then averaging, including leave-one-out evaluation. These metrics yielded consistent ranking with HighRes3DNet ranked first then DenseVNet and ScaleNet using all ranking methods. Relative volume difference yielded the best positivity-agreement and correlation with dual visual assessment, and thus is better for providing over- and under-estimation. Interclass Correlation yielded the strongest correlation with the absolute visual assessment (0-5). Symmetric-boundary dice consistently yielded good discrimination of the networks for all three ranking methods with relatively small variations within network. Good rank discrimination may be an additional metric feature required for better network performance evaluation.

Infiltrative pattern of metastatic invasive lobular breast carcinoma in the abdomen: a pictorial review.

Invasive lobular carcinoma (ILC) has a greater tendency to metastasize to the peritoneum, retroperitoneum, and gastrointestinal (GI) tract as compared to invasive carcinoma of no special type (NST). Like primary ILC in the breast, ILC metastases are frequently infiltrative and hypometabolic, rather than mass forming and hypermetabolic in nature. This renders them difficult to detect on conventional and metabolic imaging studies. As a result, intra-abdominal ILC metastases are often detected late, with patients presenting with clinical complications such as liver failure, hydronephrosis, or bowel obstruction. In patients with known history of ILC, certain imaging features are very suggestive of infiltrative metastatic ILC. These include retroperitoneal or peritoneal nodularity and linitis plastica appearance of the bowel. Recognition of linitis plastica on imaging should prompt deep or repeat biopsies. In this pictorial review, the authors aim to familiarize readers with imaging features and pitfalls for evaluation of intra-abdominal metastatic ILC. Awareness of these will allow the radiologist to assess these patients with a high index of suspicion and aid detection of metastatic disease. Also, this can direct histopathology and immunohistochemical staining to obtain the correct diagnosis in suspected metastatic disease.

Atypical Response Patterns in Renal Cell Carcinoma Treated with Immune Checkpoint Inhibitors-Navigating the Radiologic Potpourri.

BACKGROUND: Atypical response patterns have been a topic of increasing relevance since the advent of immune checkpoint inhibitors (ICIs), challenging the traditional RECIST (Response Evaluation Criteria in Solid Tumors) method of tumor response assessment. Newer immune-related response criteria can allow for the evolution of radiologic pseudoprogression, but still fail to capture the full range of atypical response patterns encountered in clinical reporting. METHODS: We did a detailed lesion-by-lesion analysis of the serial imaging of 46 renal cell carcinoma (RCC) patients treated with ICIs with the aim of capturing the full range of radiologic behaviour. RESULTS: Atypical response patterns observed included pseudoprogression (n = 15; 32.6%), serial pseudoprogression (n = 4; 8.7%), dissociated response (n = 22; 47.8%), abscopal response (n = 9; 19.6%), late response (n = 5; 10.9%), and durable response after cessation of immunotherapy (n = 2; 4.3%). Twenty-four of 46 patients (52.2%) had at least one atypical response pattern and 18 patients (39.1%) had multiple atypical response patterns. CONCLUSIONS: There is a high incidence of atypical response patterns in RCC patients receiving ICIs and the study contributes to the growing literature on the abscopal effect. The recognition of these interesting and overlapping radiologic patterns challenges the oncologist to tweak treatment options such that the clinical benefits of ICIs are potentially maximized.

PD-L1 Expressing Recurrent Clear Cell Carcinoma of the Vulva with Durable Partial Response to Pembrolizumab: A Case Report.

BACKGROUND: The optimal treatment and molecular landscape of recurrent clear cell carcinoma of the vulva (VCCC) are unknown. No reported data exist regarding the efficacy of anti-programmed death 1 (PD-1) immune checkpoint inhibition in VCCC. We report on a patient with chemotherapy-refractory recurrent VCCC, who was found to have high tumor programmed death-ligand 1 (PD-L1) combined positive score (CPS), and subsequently experienced a durable partial response (PR), after treatment with off-label fifth-line pembrolizumab. CASE PRESENTATION: A forty-year-old Filipino woman presented to our center with recurrent VCCC that had progressed on multiple prior lines of cytotoxic chemotherapy. She had a large 25 cm fungating left groin tumor causing marked lower limb lymphedema, pain and limited mobility. PD-L1 CPS by immunohistochemistry was 45. She was treated with off-label pembrolizumab monotherapy and had a dramatic clinical, biochemical and radiological partial response. The progression-free survival of this patient's VCCC after treatment with pembrolizumab, defined as the time from initiation of pembrolizumab until disease progression (by Response Evaluation Criteria in Solid Tumors (version 1.1)), was 8 months. While receiving pembrolizumab, she was diagnosed with concurrent secondary myelodysplastic syndrome with excess blasts (MDS-EB), thought to be related to her prior exposure to multiple lines of cytotoxic chemotherapy. This eventually progressed to acute myeloid leukemia (AML), leading to her demise. Overall survival from time of initiation of pembrolizumab till death was 16 months. CONCLUSION: Pembrolizumab was active in this patient with chemotherapy-refractory VCCC which harbored high PD-L1 CPS. Further studies to determine the role of immune check-point blockade in the treatment of VCCC are warranted.

Deep Learning Systems for Pneumothorax Detection on Chest Radiographs: A Multicenter External Validation Study.

PURPOSE: To assess the generalizability of a deep learning pneumothorax detection model on datasets from multiple external institutions and examine patient and acquisition factors that might influence performance. MATERIALS AND METHODS: In this retrospective study, a deep learning model was trained for pneumothorax detection by merging two large open-source chest radiograph datasets: ChestX-ray14 and CheXpert. It was then tested on six external datasets from multiple independent institutions (labeled A-F) in a retrospective case-control design (data acquired between 2016 and 2019 from institutions A-E; institution F consisted of data from the MIMIC-CXR dataset). Performance on each dataset was evaluated by using area under the receiver operating characteristic curve (AUC) analysis, sensitivity, specificity, and positive and negative predictive values, with two radiologists in consensus being used as the reference standard. Patient and acquisition factors that influenced performance were analyzed. RESULTS: The AUCs for pneumothorax detection for external institutions A-F were 0.91 (95% CI: 0.88, 0.94), 0.97 (95% CI: 0.94, 0.99), 0.91 (95% CI: 0.85, 0.97), 0.98 (95% CI: 0.96, 1.0), 0.97 (95% CI: 0.95, 0.99), and 0.92 (95% CI: 0.90, 0.95), respectively, compared with the internal test AUC of 0.93 (95% CI: 0.92, 0.93). The model had lower performance for small compared with large pneumothoraces (AUC, 0.88 [95% CI: 0.85, 0.91] vs AUC, 0.96 [95% CI: 0.95, 0.97]; P = .005). Model performance was not different when a chest tube was present or absent on the radiographs (AUC, 0.95 [95% CI: 0.92, 0.97] vs AUC, 0.94 [95% CI: 0.92, 0.05]; P > .99). CONCLUSION: A deep learning model trained with a large volume of data on the task of pneumothorax detection was able to generalize well to multiple external datasets with patient demographics and technical parameters independent of the training data.Keywords: Thorax, Computer Applications-Detection/DiagnosisSee also commentary by Jacobson and Krupinski in this issue.Supplemental material is available for this article.©RSNA, 2021.

Phase 1 Study of Low-Dose Fractionated Whole Abdominal Radiation Therapy in Combination With Weekly Paclitaxel for Platinum-Resistant Ovarian Cancer (GCGS-01).

PURPOSE: Low-dose fractionated whole abdominal radiation therapy (LDFWART) has synergistic activity with paclitaxel in preclinical models. The aim of this phase 1 trial was to determine the recommended phase 2 dose and preliminary activity of weekly paclitaxel (wP) concurrent with LDFWART in patients with platinum-resistant ovarian cancer (PROC). METHODS AND MATERIALS: Patients were enrolled at de-escalating dose levels of wP (part A), starting at 80 mg/m2, concurrent with fixed-dose LDFWART delivered in 60 cGy fractions twice-daily, 2 days per week, for 6 continuous weeks. After completing the 6-week course of wP + LDFWART, patients received wP until disease progression. Dose-limiting toxicity was evaluated during the first 3 weeks of wP + LDFWART. At wP (80 mg/m2) + LDFWART, no dose-limiting toxicities were observed; this was the established maximum tolerated dose. The trial was expanded (part B) with 7 additional patients with platinum-resistant, high-grade serous ovarian cancer to confirm toxicity and activity. RESULTS: A total of 10 heavily pretreated patients were recruited (3 patients to part A, 7 patients to part B). They had received a median of 5 prior lines of therapy, and 70% of patients had received prior wP; 60% of patients completed 6 weeks of wP + LDFWART. Common related grade ≥3 adverse events were neutropenia (60%) and anemia (30%). Median progression-free survival was 3.2 months, and overall survival was 13.5 months. Of patients evaluable for response, 33% (3 of 9) achieved confirmed biochemical response (CA125 decrease >50% from baseline), 11% (1) achieved a partial response, and 5 patients had stable disease, giving a disease control rate of 66.7% (6 of 9). Four patients had durable disease control of ≥12 weeks, completing 12 to 21 weeks of wP. CONCLUSIONS: The recommended phase 2 dose of wP + LDFWART for 6 weeks is 80 mg/m2. Encouraging efficacy in heavily pretreated PROC patients was observed, suggesting that further development of this therapeutic strategy in PROC should be considered.

Evaluation of Multimodal Algorithms for the Segmentation of Multiparametric MRI Prostate Images.

Prostate segmentation in multiparametric magnetic resonance imaging (mpMRI) can help to support prostate cancer diagnosis and therapy treatment. However, manual segmentation of the prostate is subjective and time-consuming. Many deep learning monomodal networks have been developed for automatic whole prostate segmentation from T2-weighted MR images. We aimed to investigate the added value of multimodal networks in segmenting the prostate into the peripheral zone (PZ) and central gland (CG). We optimized and evaluated monomodal DenseVNet, multimodal ScaleNet, and monomodal and multimodal HighRes3DNet, which yielded dice score coefficients (DSC) of 0.875, 0.848, 0.858, and 0.890 in WG, respectively. Multimodal HighRes3DNet and ScaleNet yielded higher DSC with statistical differences in PZ and CG only compared to monomodal DenseVNet, indicating that multimodal networks added value by generating better segmentation between PZ and CG regions but did not improve the WG segmentation. No significant difference was observed in the apex and base of WG segmentation between monomodal and multimodal networks, indicating that the segmentations at the apex and base were more affected by the general network architecture. The number of training data was also varied for DenseVNet and HighRes3DNet, from 20 to 120 in steps of 20. DenseVNet was able to yield DSC of higher than 0.65 even for special cases, such as TURP or abnormal prostate, whereas HighRes3DNet's performance fluctuated with no trend despite being the best network overall. Multimodal networks did not add value in segmenting special cases but generally reduced variations in segmentation compared to the same matched monomodal network.