Detection and quantification of pulmonary embolism with artificial intelligence: The SFR 2022 artificial intelligence data challenge.

PURPOSE: In 2022, the French Society of Radiology together with the French Society of Thoracic Imaging and CentraleSupelec organized their 13th data challenge. The aim was to aid in the diagnosis of pulmonary embolism, by identifying the presence of pulmonary embolism and by estimating the ratio between right and left ventricular (RV/LV) diameters, and an arterial obstruction index (Qanadli's score) using artificial intelligence. MATERIALS AND METHODS: The data challenge was composed of three tasks: the detection of pulmonary embolism, the RV/LV diameter ratio, and Qanadli's score. Sixteen centers all over France participated in the inclusion of the cases. A health data hosting certified web platform was established to facilitate the inclusion process of the anonymized CT examinations in compliance with general data protection regulation. CT pulmonary angiography images were collected. Each center provided the CT examinations with their annotations. A randomization process was established to pool the scans from different centers. Each team was required to have at least a radiologist, a data scientist, and an engineer. Data were provided in three batches to the teams, two for training and one for evaluation. The evaluation of the results was determined to rank the participants on the three tasks. RESULTS: A total of 1268 CT examinations were collected from the 16 centers following the inclusion criteria. The dataset was split into three batches of 310, 580 and 378 C T examinations provided to the participants respectively on September 5, 2022, October 7, 2022 and October 9, 2022. Seventy percent of the data from each center were used for training, and 30% for the evaluation. Seven teams with a total of 48 participants including data scientists, researchers, radiologists and engineering students were registered for participation. The metrics chosen for evaluation included areas under receiver operating characteristic curves, specificity and sensitivity for the classification task, and the coefficient of determination r2 for the regression tasks. The winning team achieved an overall score of 0.784. CONCLUSION: This multicenter study suggests that the use of artificial intelligence for the diagnosis of pulmonary embolism is possible on real data. Moreover, providing quantitative measures is mandatory for the interpretability of the results, and is of great aid to the radiologists especially in emergency settings.

Generative adversarial networks (GAN)-based data augmentation of rare liver cancers: The SFR 2021 Artificial Intelligence Data Challenge.

PURPOSE: The 2021 edition of the Artificial Intelligence Data Challenge was organized by the French Society of Radiology together with the Centre National d'Études Spatiales and CentraleSupélec with the aim to implement generative adversarial networks (GANs) techniques to provide 1000 magnetic resonance imaging (MRI) cases of macrotrabecular-massive (MTM) hepatocellular carcinoma (HCC), a rare and aggressive subtype of HCC, generated from a limited number of real cases from multiple French centers. MATERIALS AND METHODS: A dedicated platform was used by the seven inclusion centers to securely upload their anonymized MRI examinations including all three cross-sectional images (one late arterial and one portal-venous phase T1-weighted images and one fat-saturated T2-weighted image) in compliance with general data protection regulation. The quality of the database was checked by experts and manual delineation of the lesions was performed by the expert radiologists involved in each center. Multidisciplinary teams competed between October 11th, 2021 and February 13th, 2022. RESULTS: A total of 91 MTM-HCC datasets of three images each were collected from seven French academic centers. Six teams with a total of 28 individuals participated in this challenge. Each participating team was asked to generate one thousand 3-image cases. The qualitative evaluation was performed by three radiologists using the Likert scale on ten randomly selected cases generated by each participant. A quantitative evaluation was also performed using two metrics, the Frechet inception distance and a leave-one-out accuracy of a 1-Nearest Neighbor algorithm. CONCLUSION: This data challenge demonstrates the ability of GANs techniques to generate a large number of images from a small sample of imaging examinations of a rare malignant tumor.

French Imaging Database Against Coronavirus (FIDAC): A large COVID-19 multi-center chest CT database.

PURPOSE: During the first wave of the COVID-19 pandemic, the French Society of Radiology and the French College of Radiology, in partnership with NEHS Digital, have set up a system to collect chest computed tomography (CT) examinations with clinical, virological and radiological metadata, from patients clinically suspected of COVID-19 pneumonia. This allowed the constitution of an anonymized multicenter database, named FIDAC (French Imaging Database Against Coronavirus). The aim of this report was to describe the content of this public database. MATERIALS AND METHODS: Twenty-two French radiology centers participated to the data collection. The data collected were chest CT examinations in DICOM format associated with the following metadata: patient age and sex, originating facility identifier, originating facility region, time from symptom onset to CT examination, indication for CT examination, reverse transcription-polymerase chain reaction (RT-PCR) results and normalized CT report performed by a senior radiologist. All the data were anonymized and sent through a NEHS Digital system to a centralized data center. RESULTS: A total of 5944 patients were included from the 22 centers aggregated into 8 regions with a mean number of patients of 743 ± 603.3 [SD] per region (range: 102-1577 patients). Reasons for CT examination and normalized CT reports were provided for all patients. RT-PCR results were provided in 5574 patients (93.77%) with a positive result of RT-PCR in 44.6% of patients. CONCLUSION: The FIDAC project allowed the creation of a large database of chest CT images and metadata available, under conditions, in open access through the CERF-SFR website.

Validation of a post-mortem computed tomography method for age estimation based on the 4th rib in a French population.

Age estimation is a key factor for identification procedure in forensic context. Based on anthropological findings, degenerative changes of the sternal extremity of the 4th rib are currently used for age estimation. These have been adapted to post-mortem computed tomography (PMCT). The aim of this study was to validate a post-mortem computed tomography method based on a revision of the Iscan's method on a French sample. A total of 250 PMCT (aged from 18-98 years (IQR 36-68 years, median 51 years); 68 (27%) females) from the Medicolegal Institute of Paris (MLIP) were analyzed by two radiologists. The sternal extremity of 4th right rib was scored using method adapted from Iscan et al. Weighted κ was used to evaluate intra- and inter-observer reliability and Spearman correlation was performed to evaluate relationship between age and score. Confidence intervals for individual prediction of age based on 4th rib score and sex were computed with bootstrapping. The intra-observer reliability and inter-observer reliability were almost perfect (weighted κ = 0.85 [95%CI: 0.78-0.93] and 0.82 [95%CI 0.70-0.96] respectively). We confirmed a high correlation between the 4th rib score and subject age (rho = 0.72, p < 0.001), although the confidence intervals for individual age prediction were large, spanning over several decades. This study confirms the high reliability of Iscan method applied to PMCT for age estimation, although future multimodal age prediction techniques may help reducing the span of confidence intervals for individual age estimation.Trial registration: INDS 0,509,211,020, October 2020, retrospectively registered.

Three artificial intelligence data challenges based on CT and ultrasound.

PURPOSE: The 2020 edition of these Data Challenges was organized by the French Society of Radiology (SFR), from September 28 to September 30, 2020. The goals were to propose innovative artificial intelligence solutions for the current relevant problems in radiology and to build a large database of multimodal medical images of ultrasound and computed tomography (CT) on these subjects from several French radiology centers. MATERIALS AND METHODS: This year the attempt was to create data challenge objectives in line with the clinical routine of radiologists, with less preprocessing of data and annotation, leaving a large part of the preprocessing task to the participating teams. The objectives were proposed by the different organizations depending on their core areas of expertise. A dedicated platform was used to upload the medical image data, to automatically anonymize the uploaded data. RESULTS: Three challenges were proposed including classification of benign or malignant breast nodules on ultrasound examinations, detection and contouring of pathological neck lymph nodes from cervical CT examinations and classification of calcium score on coronary calcifications from thoracic CT examinations. A total of 2076 medical examinations were included in the database for the three challenges, in three months, by 18 different centers, of which 12% were excluded. The 39 participants were divided into six multidisciplinary teams among which the coronary calcification score challenge was solved with a concordance index > 95%, and the other two with scores of 67% (breast nodule classification) and 63% (neck lymph node calcifications).

Meningioangiomatosis: Multimodal Analysis and Insights From a Systematic Review.

BACKGROUND: Meningioangiomatosis is a poorly studied, rare, benign, and epileptogenic brain lesion. OBJECTIVE: To demonstrate that surgical resection and a short-time interval to surgery improves epileptic seizure control, we performed a systematic review and meta-analysis of meningioangiomatosis cases. METHODS: Using PRISMA-IPD guidelines, the authors performed a systematic review and meta-analysis of histopathologically-proven meningioangiomatosis cases. Literature search in French and English languages (PubMed, Embase, the Cochrane Library, and the Science Citation Index) including all studies (January 1981 to June 2020) dealing with histopathologically-proven meningioangiomatosis, without age restriction. We assessed clinical, imaging, histomolecular, management, and outcome findings of patients with meningioangiomatosis. RESULTS: Two-hundred and seven cases of meningioangiomatosis from 78 studies were included. Most meningioangiomatosis was sporadic, preferentially concerned male patients, younger than 20 years old, and allowed a functionally independent status. Epileptic seizure was the main symptom, with 81.4% of patients having uncontrolled seizures at the time of surgery. Meningioangiomatosis mainly had frontal (32.3%) or temporal (30.7%) locations. Imaging presentation was heterogeneous, and the diagnosis was often missed preoperatively. The histopathologic pattern was similar whatever the clinical presentation, and immunohistochemistry had limited diagnostic value. On molecular analysis, allelic loss at 22q12 was more frequent in samples of meningioangiomatosis-associated meningioma (37.5%) than in isolated meningioangiomatosis (23.1%). Time interval from diagnosis to surgery (p = 0.011) and lack of surgical resection of the meningioangiomatosis (p = 0.009) were independent predictors of postoperative seizure control. CONCLUSIONS: Owing to low scientific evidence, a multicentric prospective study should help refining the management of meningioangiomatosis.

MRI Atlas of IDH Wild-Type Supratentorial Glioblastoma: Probabilistic Maps of Phenotype, Management, and Outcomes.

Background Tumor location is a main prognostic parameter in patients with glioblastoma. Probabilistic MRI-based brain atlases specifying the probability of tumor location associated with important demographic, clinical, histomolecular, and management data are lacking for isocitrate dehydrogenase (IDH) wild-type glioblastomas. Purpose To correlate glioblastoma location with clinical phenotype, surgical management, and outcomes by using a probabilistic analysis in a three-dimensional (3D) MRI-based atlas. Materials and Methods This retrospective study included all adults surgically treated for newly diagnosed IDH wild-type supratentorial glioblastoma in a tertiary adult surgical neuro-oncology center (2006-2016). Semiautomated tumor segmentation and spatial normalization procedures to build a 3D MRI-based atlas were validated. The authors performed probabilistic analyses by using voxel-based lesion symptom mapping technology. The Liebermeister test was used for binary data, and the generalized linear model was used for continuous data. Results A total of 392 patients (mean age, 61 years ± 13; 233 men) were evaluated. The authors identified the preferential location of glioblastomas according to subventricular zone, age, sex, clinical presentation, revised Radiation Therapy Oncology Group-Recursive Partitioning Analysis class, Karnofsky performance status, O6-methylguanine DNA methyltransferase promoter methylation status, surgical management, and survival. The superficial location distant from the eloquent area was more likely associated with a preserved functional status at diagnosis (348 of 392 patients [89%], P < .05), a large surgical resection (173 of 392 patients [44%], P < .05), and prolonged overall survival (163 of 334 patients [49%], P < .05). In contrast, deep location and location within eloquent brain areas were more likely associated with an impaired functional status at diagnosis (44 of 392 patients [11%], P < .05), a neurologic deficit (282 of 392 patients [72%], P < .05), treatment with biopsy only (183 of 392 patients [47%], P < .05), and shortened overall survival (171 of 334 patients [51%], P < .05). Conclusion The authors identified the preferential location of isocitrate dehydrogenase wild-type glioblastomas according to parameters of interest and provided an image-based integration of multimodal information impacting survival results. This suggests the role of glioblastoma location as a surrogate and multimodal parameter integrating several known prognostic factors. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Huang in this issue.

Optimal 4DFlow MR sequence parameters for the assessment of internal carotid artery stenosis: a simulation study.

PURPOSE: In patients with ICA stenosis, increased peak systolic velocity is a marker of stenosis at risk of ischemic stroke. 4DFlow MRI is a reproducible technique to evaluate velocities in ICA stenosis, although it seems to underestimate velocities as compared with Doppler ultrasonography. The purpose of our study was to confirm that velocities were underestimated on a new set of data acquired with a clinical 4DFlow sequence, and to devise optimal acquisition parameters for ICA stenosis exploration based on a numerical simulation. METHODS: After review board approval, 15 healthy controls and 12 patients presenting ICA stenosis were explored with Doppler ultrasonography and 4DFlow MRI. We created a 2-dimensional simulation of ICA stenosis and its corresponding 4DFlow acquisition, and compared its mean peak systolic velocity underestimation to real MRI and Doppler. We then simulated the acquisition for voxel size ranging from 0.5 to 1.25 mm and number of phases per cardiac cycle ranging from 10 to 25. RESULTS: On acquired data, 4DFlow MR underestimated peak systolic velocities (mean difference between Doppler and 4DFlow: - 35 cm/s), especially high velocities. With spatial and temporal resolutions equivalent to MR acquisition, our simulation yielded similar underestimation (mean difference: - 31 cm/s, P = 0.30). Simulations showed that 0.7-mm resolution and 20 phases per cardiac cycle would be necessary to record peak systolic velocities up to 250 cm/s. CONCLUSION: Higher spatial resolution can provide accurate peak systolic velocities measurement with 4DFlow MRI, thus allowing better ICA stenosis assessment. Further studies are needed to validate the proposed parameters.

Anaesthetic management during intracranial mechanical thrombectomy: systematic review and meta-analysis of current data.

OBJECTIVE: Our aim was to compare the clinical outcome of patients with ischaemic stroke with anterior large vessel occlusion treated with stent retrievers and/or contact aspiration mechanical thrombectomy (MT) under general anaesthesia (GA) or conscious sedation non-GA through a systematic review and meta-analysis. METHODS: The literature was searched using PubMed, Embase and Cochrane databases to identify studies reporting on anaesthesia and MT. Using fixed or random weighted effect, we evaluated the following outcomes: 3-month mortality, modified Rankin Score (mRs) 0-2, recanalisation success (thrombolysis in cerebral infarction (TICI) ≥2b) and symptomatic intracerebral haemorrhagic (sICH) transformation. RESULTS: We identified seven cohorts (including three dedicated randomised controlled trials), totalling 1929 patients (932 with GA). Over the entire sample, mortality, mRs 0-2, TICI≥2b and sICH rates were, respectively 17.5% (99% CI 9.7% to 29.6%; Q-value: 60.1; I2: 93%, 1717 patients), 42.1% (99% CI 33.3% to 51.7%; Q-value: 41.3; I2: 87.9%), 82.9% (99% CI 74.0% to 89.1%; Q-value: 20.7; I2: 80.6%, 1006 patients) and 5.5% (99% CI 2.8% to 10.8%; Q-value: 18.6; I2: 78.5%). MT performed in non-GA patients was associated with better 3-month functional outcome (pooled OR, 1.35; 99% CI 1.04 to 1.76; Q-value: 24.0; I2: 9.2%, 1845 patients) and lower 3-month mortality rate (pooled OR, 0.70; 99% CI 0.49 to 0.98; Q-value: 1.4; I2: 0%, 1717 patients; fixed weighted effect model) compared with GA. MT performed under conscious sedation non-GA had significantly shorter onset-to-recanalisation and onset-to-groin delay compared with GA, and recanalisation success and sICH were similar. CONCLUSION: Non-GA during MT for anterior acute ischaemic stroke with current-generation stent retriever/aspiration devices is associated with better 3-month functional outcome and lower mortality rates. These unadjusted estimates are subject to biases and should be interpreted with caution.

Circumferential Thick Enhancement at Vessel Wall MRI Has High Specificity for Intracranial Aneurysm Instability.

Purpose To identify wall enhancement patterns on vessel wall MRI that discriminate between stable and unstable unruptured intracranial aneurysm (UIA). Materials and Methods Patients were included from November 2012 through January 2016. Vessel wall MR images were acquired at 3 T in patients with stable (incidental and nonchanging over 6 months) or unstable (symptomatic or changing over 6 months) UIA. Each aneurysm was evaluated by using a four-grade classification of enhancement: 0, none; 1, focal; 2, thin circumferential; and 3, thick (>1 mm) circumferential. Inter- and intrareader agreement for the presence and the grade of enhancement were assessed by using κ statistics and 95% confidence interval (CI). The sensitivity, specificity, and negative and positive predictive values of each enhancement grade for differentiating stable from unstable aneurysms was compared. Results The study included 263 patients with 333 aneurysms. Inter- and intrareader agreement was excellent for both the presence of enhancement (κ values, 0.82 [95% CI: 0.67, 0.99] and 0.87 [95% CI: 0.7, 1.0], respectively) and enhancement grade (κ = 0.92 [95% CI: 0.87, 0.95]). In unruptured aneurysms (n = 307), grade 3 enhancement exhibited the highest specificity (84.4%; 233 of 276; 95% CI: 80.1%, 88.7%; P = .02) and negative predictive value (94.3%; 233 of 247) for differentiating between stable and unstable lesions. There was a significant association between grade 3 enhancement and aneurysm instability (P < .0001). Conclusion In patients with intracranial aneurysm, a thick (>1 mm) circumferential pattern of wall enhancement demonstrated the highest specificity for differentiating between stable and unstable aneurysms. © RSNA, 2018.

Independent Factors Affecting Postoperative Complication Rates After Custom-Made Porous Hydroxyapatite Cranioplasty: A Single-Center Review of 109 Cases.

BACKGROUND: Cranioplasties are an important neurosurgical procedure not only for improved cosmesis but also for improved functional recovery after craniectomy with a large cranial defect. The aim of this study was to identify predictive factors of postcranioplasty complications using custom-made porous hydroxyapatite cranioplasty. METHODS: Retrospective review was performed of all patients who underwent a reconstructive cranioplasty using custom-made hydroxyapatite at our institution between February 2008 and September 2017. Postoperative complications considered included bacterial infection, seizures, hydrocephalus requiring ventricular shunt placement, and cranioplasty-to-bone shift. Variables associated at P < 0.1 level in unadjusted analysis were entered into backward stepwise logistic regression models. RESULTS: Of 109 patients included, 15 (13.8%) experienced postoperative infection, with craniectomy performed at an outside institution (adjusted odds ratio [OR] 10.37 [95% confidence interval [CI], 2.03-75.27], P = 0.012) and a previous infection at the surgical site (adjusted OR 6.15 [95%CI, 1.90-19.92], P = 0.003) identified as independent predictors. Six patients (5.5%) experienced postoperative seizures, with stroke (ischemic and hemorrhagic) as a reason for craniectomy (adjusted OR 11.68 [95% CI, 2.56-24.13], P < 0.001) and the presence of seizures in the month before cranioplasty (adjusted OR 9.39 [95% CI, 2.04-127.67], P = 0.002) identified as independent predictors. Four patients (3.7%) experienced postcranioplasty hydrocephalus necessitating shunt placement, and 5 patients (4.6%) experienced cranioplasty-to-bone shift ≥5 mm, but no significant predictive factors were identified for either complication. CONCLUSIONS: This study identified possible predictive factors for postcranioplasty complications to help identify at-risk patients, guide prophylactic care, and improve morbidity of this important surgical procedure.

Predictors of Outcome in Patients with Pediatric Intracerebral Hemorrhage: Development and Validation of a Modified Score.

Purpose To propose and validate a modified pediatric intracerebral hemorrhage (PICH) (mPICH) score and to compare its association with functional outcome to that of the original PICH score. Materials and Methods Data from prospectively included patients were retrospectively analyzed. Consecutive patients with nontraumatic PICH who had undergone clinical follow-up were included. The study population was divided into a development cohort (2008-2012, n = 100) and a validation cohort (2013-2016, n = 43). An mPICH score was developed after variables associated with poor outcome were identified at multivariate analysis (King's Outcome Scale for Childhood Head Injury score < 5a) in the development cohort. The accuracy of the score for prediction of poor outcome was evaluated (sensitivity, specificity). Discrimination and calibration of associations between the mPICH score and poor outcome cohorts were assessed (C statistics, Hosmer-Lemeshow test). Results The mPICH score assessed as follows: brain herniation, four points; altered mental status, three points; hydrocephalus, two points; infratentorial PICH, two points; intraventricular hemorrhage, one point; PICH volume greater than 2% of total brain volume, one point. An mPICH score greater than 5 was associated with severe disability or worse, with sensitivity of 97% (95% confidence interval [CI]: 83%, 100%) and specificity of 61% (95% CI: 49%, 73%). The C statistic was 0.81 (95% CI: 0.73, 0.89). In the validation cohort, sensitivity and specificity were 95.2% (95% CI: 76%, 99%) and 77% (95% CI: 55%, 92%), respectively. There was no significant difference between the observed and predicted risks of poor outcome (P = .46). Conclusion An mPICH score was developed as a simple clinical and imaging grading scale for acute prognosis in patients with PICH. © RSNA, 2017.

Presentation and management of lateral sinus thrombosis following posterior fossa surgery.

OBJECTIVE There are no guidelines for the management of postoperative lateral sinus thrombosis following posterior fossa surgery. Introducing treatment-dose anticoagulant therapy during the immediate postoperative period increases the risk of intracranial bleeding. This study assessed the incidence of and risk factors associated with postoperative lateral sinus thrombosis and the complications related to thrombosis and/or anticoagulation. METHODS This study was a retrospective monocentric analysis of adult patients who underwent surgical removal of a posterior fossa space-occupying lesion with available postoperative imaging. Postoperative lateral sinus thrombosis was defined as a T2* hypointensity within the venous sinus and/or a filling defect on postcontrast MRI or CT scan. RESULTS Among 180 patients, 12 (6.7%; 95% CI 3.0-10.4) were found to have lateral sinus thrombosis on postoperative imaging, none of whom were symptomatic. Unadjusted risk factors for postoperative lateral sinus thrombosis were a history of deep venous thrombosis (p = 0.016), oral contraceptive pill (p = 0.004), midline surgical approach (p = 0.035), and surgical exposure of the sinus (p < 0.001). Seven of the patients (58.3%) with a postoperative lateral sinus thrombosis received immediate treatment-dose anticoagulant therapy. Lateral sinus recanalization occurred radiologically at a mean time of 272 ± 23 days in 85.7% of patients (6 of 7) undergoing treatment-dose anticoagulant therapy and in 20% of patients (1 of 5) not receiving treatment-dose anticoagulant therapy. Postoperative complications occurred in 56.2% of patients (9 of 16) who received treatment-dose curative anticoagulant therapy and in 27% of patients (45 of 164) who did not. CONCLUSIONS Incidental radiological lateral sinus thrombosis following posterior fossa surgery has an incidence of 6.7%. To further define the benefit-to-risk ratio of a treatment-dose anticoagulant therapy, a prospective trial should be considered.

Epileptic seizures in anaplastic gangliogliomas.

AIM: Prevalence and predictors of epileptic seizures are unknown in the malignant variant of ganglioglioma. METHODS: In a retrospective exploratory dataset of 18 supratentorial anaplastic World Health Organization grade III gangliogliomas, we studied: (i) the prevalence and predictors of epileptic seizures at diagnosis; (ii) the evolution of seizures during tumor evolution; (iii) seizure control rates and predictors of epilepsy control after oncological treatments. RESULTS: Epileptic seizures prevalence progresses throughout the natural course of anaplastic gangliogliomas: 44% at imaging discovery, 67% at histopathological diagnosis, 69% following oncological treatment, 86% at tumor progression, and 100% at the end-of-life phase. The medical control of seizures and their refractory status worsened during the tumor's natural course: 25% of uncontrolled seizures at histopathological diagnosis, 40% following oncological treatment, 45.5% at tumor progression, and 45.5% at the end-of-life phase. Predictors of seizures at diagnosis appeared related to the tumor location (i.e. temporal and/or cortical involvement). Prognostic parameters of seizure control after first-line oncological treatment were temporal tumor location, eosinophilic granular bodies, TP53 mutation, and extent of resection. Prognostic parameters of seizure control at tumor progression were a history of epileptic seizures at diagnosis, seizure control after first-line oncological treatment, eosinophilic granular bodies, and TP53 mutation. CONCLUSION: Epileptic seizures are frequently observed in anaplastic gangliogliomas and both prevalence and medically refractory status worsen during the tumor's natural course. Both oncological and antiepileptic treatments should be employed to improve the control of epileptic seizures and the quality of life of patients harboring an anaplastic ganglioglioma.