Five simultaneous artificial intelligence data challenges on ultrasound, CT, and MRI.

PURPOSE: The goal of this data challenge was to create a structured dynamic with the following objectives: (1) teach radiologists the new rules of General Data Protection Regulation (GDPR), while building a large multicentric prospective database of ultrasound, computed tomography (CT) and MRI patient images; (2) build a network including radiologists, researchers, start-ups, large companies, and students from engineering schools, and; (3) provide all French stakeholders working together during 5 data challenges with a secured framework, offering a realistic picture of the benefits and concerns in October 2018. MATERIALS AND METHODS: Relevant clinical questions were chosen by the Société Francaise de Radiologie. The challenge was designed to respect all French ethical and data protection constraints. Multidisciplinary teams with at least one radiologist, one engineering student, and a company and/or research lab were gathered using different networks, and clinical databases were created accordingly. RESULTS: Five challenges were launched: detection of meniscal tears on MRI, segmentation of renal cortex on CT, detection and characterization of liver lesions on ultrasound, detection of breast lesions on MRI, and characterization of thyroid cartilage lesions on CT. A total of 5,170 images within 4 months were provided for the challenge by 46 radiology services. Twenty-six multidisciplinary teams with 181 contestants worked for one month on the challenges. Three challenges, meniscal tears, renal cortex, and liver lesions, resulted in an accuracy>90%. The fourth challenge (breast) reached 82% and the lastone (thyroid) 70%. CONCLUSION: Theses five challenges were able to gather a large community of radiologists, engineers, researchers, and companies in a very short period of time. The accurate results of three of the five modalities suggest that artificial intelligence is a promising tool in these radiology modalities.

Psoas proximal insertion as a simple and reliable landmark for numbering lumbar vertebrae on MRI of the lumbar spine.

OBJECTIVE: To evaluate the value of psoas muscle proximal insertion for correct numbering of the lumbar vertebrae in MRI, in particular in case of lumbosacral transitional vertebra (LSTV). METHODS: Two radiologists assessed 477 MRI scans of the lumbar spine with a sagittal localizer sequence on the whole spine for numbering vertebrae caudally from C2. Proximal insertion of the psoas was determined as the most proximal vertebra with psoas over half of its body on coronal T2 STIR sequence. The last lumbar vertebra was named considering both its number and the presence or absence of LSTV according to Castellvi classification. These same parameters were also assessed on 207 PET-CT scans of another cohort including the whole spine. RESULTS: Proximal insertion of the psoas was L1 in 94.1% of cases: 98.5% in case of modal anatomy, 81.4% in case of LSTV, and 51.7% in case of missing or supernumerary lumbar vertebra without LSTV. There was no statistically significant difference between MRI and CT data. The inter-reader agreement for determination of psoas proximal insertion was excellent (kappa = 0.96). CONCLUSION: Proximal insertion of the psoas muscle is a helpful marker for correct numbering of the lumbar vertebrae in MRI and to detect a complete lumbosacral segmentation anomaly. KEY POINTS: • Proximal insertion of the psoas muscle can be easily identified on a coronal T2 STIR sequence. • Psoas proximal insertion on the spine almost always designates the first lumbar vertebra and is helpful to accurately number all lumbar vertebrae, especially in case of lumbosacral transitional vertebra. • Conversely, when psoas muscle does not insert five lumbar bodies above the apparent lumbosacral joint, the probability of variation in the number of lumbar vertebrae is high.

4D in vivo quantification of ankle joint space width using dynamic MRI.

Spatio-temporal evolution of joint space width (JSW) during motion is of great importance to help with making early treatment plans for degenerative joint diseases like osteoarthritis (OA). These diseases can affect people of all ages leading to an acceleration of joint degeneration and to limitations in the activities of daily living. However, only a few studies have attempted to quantify the JSW from moving joints. In this paper, we present a generic pipeline to accurately determine the changes of the JSW during the joint motion cycle. The key idea is to combine spatial information of static MRI with temporal information of low-resolution (LR) dynamic MRI sequences via an intensity-based registration framework, leading to a high-resolution (HR) temporal reconstruction of the joint. This allows the temporal JSW to be measured in the HR domain using an Eulerian approach for solving partial differential equations (PDEs) inside a deforming inter-bone area where the HR reconstructed bone segmentations are considered as temporal Dirichlet boundaries. The proposed approach has been applied and evaluated on in vivo MRI data of five healthy children to non-invasively quantify the spatio-temporal evolution of the JSW of the ankle (tibiotalar joint) during the entire dorsi-plantar flexion motion cycle. Promising results were obtained, showing that this pipeline can be useful to perform large-scale studies containing subjects with OA for different joints like ankle and knee.

Temporal resolution enhancement of dynamic MRI sequences within a motion-based framework.

Dynamic MRI has made it possible to non-invasively capture the moving human joints in vivo. Real-time Fast Field Echo (FFE) sequences have the potential to reduce the effect of motion artifacts by acquiring the image data within a few milliseconds. However, the short acquisition times affect the temporal resolution of the acquired sequences. In this paper, we propose a post-processing technique to reconstruct the missing frames of the sequence given the reduced amount of acquired data, which leads to recover the entire joint trajectory outside the MR scanner. To do this, we generalize the Log-Euclidean polyrigid registration framework to deal with dynamic three-dimensional articulated structures by adding the time as fourth dimension : we first estimate the rigid motion of each bone from the acquired data using linear intensity-based registration. Then, we fuse these local transformations to compute the non-linear joint deformations between successive images using a spatio-temporal log-euclidean polyrigid framework. The idea is to reconstruct the missing time frames by interpolating the realistic joint deformation fields in the domain of matrix logarithms assuming the motion to be consistent over a short period of time. The algorithm has been applied and validated using dynamic data from five children performing passive ankle dorsi-plantar flexion.

Contribution of brain CT in the diagnosis of tuberculous meningitis: a case report from Djibouti.

Tuberculous meningitis, a serious disease with high mortality and morbidity, remains frequent in countries with endemic tuberculosis. Its non-specific presentation often delays the introduction of appropriate treatment. Its definitive diagnosis requires isolation of Mycobacterium tuberculosis from cerebrospinal fluid, although this test may be negative without conclusively ruling out this diagnosis. A presumptive diagnosis should be reached as soon as possible through a body of clinical evidence, including the lumbar puncture findings. Brain computed tomography (CT) with and without contrast medium injection is helpful for the diagnosis of tuberculous meningitis and its complications. We discuss the features of CT and their value in relation to a case of tuberculous meningitis in Djibouti, as well as the role of CT in managing this disease.

In vivo skin moisturizing measurement by high-resolution 3 Tesla magnetic resonance imaging.

BACKGROUND: Magnetic resonance imaging (MRI) is rarely used for the exploration of skin, even if studies have validated both feasibility of skin MRI and its interest for anatomical, physiological, and biochemical study of the skin. The purpose of this study is to explore moisturizing of the different skin layers using 3-T scan. METHODS: An MRI of the heel's skin was performed using a 23 mm coil diameter on a 3T scan with a FFE (Fast Field Echo) 3D T1-weighted sequence and a TSE (Turbo Spin Echo) calculation T2-weighted sequence (pixels size of respectively 60 and 70 µm). This study was conducted on 35 healthy volunteers, who were scanned before applying moisturizer topic and 1 h after applying it. Region of interest in the stratum corneum, the epidermis and the dermis were generated on the T2 mapping. The thickness of each layer was measured. The T1 sequence allowed accurate cross-examination repositioning to ensure the comparability of the measurements. RESULTS: Among the 35 cases, two were excluded from the analysis because of movement artifacts. Measurements before and after moisturizer topic application displayed a T2 increase of 48.94% (P < 0.0001) in the stratum corneum and of 5.45% (P < 0.0001) in the epidermis yet without significant difference in the dermis. There was no significant link between the thickness of the stratum corneum and the T2 increase. However, there was a strong correlation between the thickness of the stratum corneum and the thickness of the epidermis (P < 0.001; rhô=0.72). CONCLUSION: High-resolution MRI allows fine exploration of anatomical and physiological properties of the skin and can further be used to extend the studies of skin hydration.

Postmortem computed tomography findings in suicide victims.

Suicide is the eighth cause of mortality in France and the leading cause in people aged between 25 and 34 years. The most common methods of suicide are hanging, self-poisoning with medicines and firearms. Postmortem computed tomography (CT) is a useful adjunct to autopsy to confirm suicide and exclude other causes of death. At autopsy, fractures of the hyoid bone or thyroid cartilage, or both, are found in more than 50% of suicidal hangings. Cervical vertebra fractures are rare and only seen in suicide victims jumping from a great height. Three-dimensional reconstructions from CT data are useful to visualize the ligature mark on the neck. In suicides by firearm, postmortem CT shows entry and exit wounds, parenchymal lesions along the bullet path, as well as projectiles in case of penetrating trauma. However, in the chest and abdomen it is more difficult to identify the path of the projectile. Postmortem CT also shows specific features of suicide by drowning or stabbing, but its use is limited in cases of self-poisoning. The use of postmortem CT is also limited by decomposition and change of body position. This article presents the imaging features seen on postmortem CT according to the method of suicide.

Diagnostic accuracy of diffusion-weighted MR imaging versus delayed gadolinium enhanced T1-weighted imaging in middle ear recurrent cholesteatoma: A retrospective study of 39 patients.

OBJECTIVE: MR imaging using diffusion-weighted (DW) images and delayed gadolinium-enhanced T1-weighted images is evolving into an alternative to second look surgery in detection of recurrent cholesteatomas. The aim of this study was to retrospectively compare the DW images, the post-gadolinium T1-weighted images and the combination of both methods in this indication. PATIENTS AND METHODS: We retrospectively evaluated the MR examination of 39 patients clinically suspected for a recurrent cholesteatoma. Patients in the study underwent DW sequences, delayed gadolinium enhanced T1-weighted sequences as well as standard uninjected protocol using T1 and T2 sequences. Three blinded radiologists evaluated three data sets: a set of post-gadolinium T1-weighted images, a set of DW images and a set of the combination of both methods. The interobserver agreement was evaluated and the diagnostic accuracy of each method was described by sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV). The performances of the three techniques were also evaluated using ROC curves, from which the AUC were compared. Results were compared with surgical results or a two-year follow-up. RESULTS: The overall sensitivity and specificity were respectively 63% and 71% for the post-gadolinium T1-weighted images, 88% and 75% for the DW images and 84% and 75% for the combined images. The PPV and NPV were respectively 89% and 33% for the post-gadolinium T1-weighted images, 93% and 62% for the DW images and 93% and 55% for the combined images. The sensitivity and the NPV were significantly different between the three methods (P<0.0001 and P=0.027). There was no statistically significant difference in specificity or PPV between the three methods (P=0.931 and P=0.650). The diagnostic accuracy evaluated with the AUC showed no statistically significant difference between the DW images and the combined images (P=0.433). CONCLUSION: MR imaging reliably identifies those patients suspected for recurrent cholesteatoma who require a surgical second look by using DW MR imaging. The combination with delayed gadolinium enhanced T1-weighted sequences does not significantly increase the diagnostic accuracy of the examination.

Biometry of the normal stapes using stapes axial plane, high-resolution computed tomography.

OBJECTIVE: The stapes is difficult to analyse on computed tomography because of the small size of its components and its oblique orientation. The stapes axial plane, parallel to the superstructure, seems optimal for this purpose. The present study assessed the position of the stapes axial plane with respect to the usual axial plane including the lateral semicircular canal, and sought to measure the main dimensions of the stapes. METHODS: This retrospective study comprised 208 computed tomography scans of normal ears. Stapes length and width, footplate thickness and incudostapedial joint width were measured. RESULTS: The stapes axial plane was directed upward, outward (44°) and forward (12°) with respect to the lateral semicircular canal plane. Mean head-to-footplate distance was 3.7 mm and mean superstructure width was 2.7 mm. Mean footplate thickness was 0.27 mm on stapes axial plane versus 0.48 mm on lateral semicircular canal plane. Incudostapedial joint width was systematically less than 0.7 mm. CONCLUSION: Stapes dimensions on stapes axial plane were close to anatomical data, particularly for footplate thickness.

Clinical-radiological correlation after functional endoscopic sinus surgery in patients with chronic rhinosinusitis: interest of a sinonasal aerial volumetry.

BACKGROUND: Although a CT scan is often performed after functional endoscopic sinonasal surgery (FESS) in patients with chronic rhinosinusitis, its role hasn`t been firmly established. The goal of this study is to investigate the correlation between symptoms and CT findings before and after FESS for chronic rhinosinusitis. In addition, the interobserver agreement for both sinonasal aerial volumetry and CT score is assessed. METHODS: Thirty-three patients surgically treated for chronic rhinosinusitis were included in this prospective study. Conventional and modified Lund-Mackay scores and sinonasal volumetry were determined by two radiologists before (M0), at 3 months (M3) and 1 year (M12) after surgery. The symptoms were evaluated by the 22-item SinoNasal Outcome Test (SNOT-22). RESULTS: Change of SNOT-22 and air volume were significantly correlated between M0 and M12, but not between M0 and M3, for both readers. Compared to other scores, volume had the best intraclass correlation coefficient and reproducibility, according to the Bland-Altman analysis. No correlation was found between SNOT-22 and CT scores before and after surgery, except between M12 and M0 for one reader. CONCLUSION: The correlation between CT scan and symptoms is low or absent. The measurement of sinonasal air volume is best correlated with the symptoms after surgery, with the best inter-observer agreement.