Reproducibility of artificial intelligence models in computed tomography of the head: a quantitative analysis.

When developing artificial intelligence (AI) software for applications in radiology, the underlying research must be transferable to other real-world problems. To verify to what degree this is true, we reviewed research on AI algorithms for computed tomography of the head. A systematic review was conducted according to the preferred reporting items for systematic reviews and meta-analyses. We identified 83 articles and analyzed them in terms of transparency of data and code, pre-processing, type of algorithm, architecture, hyperparameter, performance measure, and balancing of dataset in relation to epidemiology. We also classified all articles by their main functionality (classification, detection, segmentation, prediction, triage, image reconstruction, image registration, fusion of imaging modalities). We found that only a minority of authors provided open source code (10.15%, n 0 7), making the replication of results difficult. Convolutional neural networks were predominantly used (32.61%, n = 15), whereas hyperparameters were less frequently reported (32.61%, n = 15). Data sets were mostly from single center sources (84.05%, n = 58), increasing the susceptibility of the models to bias, which increases the error rate of the models. The prevalence of brain lesions in the training (0.49 ± 0.30) and testing (0.45 ± 0.29) datasets differed from real-world epidemiology (0.21 ± 0.28), which may overestimate performances. This review highlights the need for open source code, external validation, and consideration of disease prevalence.

FDA-approved deep learning software application versus radiologists with different levels of expertise: detection of intracranial hemorrhage in a retrospective single-center study.

PURPOSE: To assess an FDA-approved and CE-certified deep learning (DL) software application compared to the performance of human radiologists in detecting intracranial hemorrhages (ICH). METHODS: Within a 20-week trial from January to May 2020, 2210 adult non-contrast head CT scans were performed in a single center and automatically analyzed by an artificial intelligence (AI) solution with workflow integration. After excluding 22 scans due to severe motion artifacts, images were retrospectively assessed for the presence of ICHs by a second-year resident and a certified radiologist under simulated time pressure. Disagreements were resolved by a subspecialized neuroradiologist serving as the reference standard. We calculated interrater agreement and diagnostic performance parameters, including the Breslow-Day and Cochran-Mantel-Haenszel tests. RESULTS: An ICH was present in 214 out of 2188 scans. The interrater agreement between the resident and the certified radiologist was very high (κ = 0.89) and even higher (κ = 0.93) between the resident and the reference standard. The software has delivered 64 false-positive and 68 false-negative results giving an overall sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of 68.2%, 96.8%, 69.5%, 96.6%, and 94.0%, respectively. Corresponding values for the resident were 94.9%, 99.2%, 93.1%, 99.4%, and 98.8%. The accuracy of the DL application was inferior (p < 0.001) to that of both the resident and the certified neuroradiologist. CONCLUSION: A resident under time pressure outperformed an FDA-approved DL program in detecting ICH in CT scans. Our results underline the importance of thoughtful workflow integration and post-approval validation of AI applications in various clinical environments.

Blake's Pouch Cysts and Differential Diagnoses in Prenatal and Postnatal MRI : A Pictorial Review.

PURPOSE: The clinical variability of Blake's pouch cysts (BPC) may range from asymptomatic via ataxia to sequelae of decompensated hydrocephalus. On the other hand, Dandy-Walker malformation (DWM) and cerebellar vermis hypoplasia generally correlate with less favorable neurologic development. The aim was to illustrate the potential of prenatal and postnatal neuroimaging to distinguish a BPC or persistent BP from other posterior fossa malformations. METHODS: This pictorial review addresses the inconsistent nomenclature, clinical features, and magnetic resonance imaging (MRI) patterns of BPC and five differential diagnoses. The MRI findings of 11 patients, acquired at up to 3 T in 3 institutions, are demonstrated. Furthermore, the literature was searched for recent improvements in genetic and embryological background knowledge. RESULTS: Posterior fossa malformations often resemble each other and may even be imitated by sequelae of hemorrhagic, ischemic or infectious disruptions, i.e. congenital anomalies of morphology despite normal developmental potential. Hydrocephalus is a typical, albeit not always congenital finding in BPC. It is frequently associated with cerebellar disruptions and DWM; however, it is also a rare complication of posterior fossa arachnoid cysts. A moderately elevated vermis needs follow-up to confirm persistent BP versus vermian hypoplasia or DWM. The fetal cerebellar tail, previously assumed to be specific for DWM, may be imitated in cases of persistent BP. CONCLUSION: The accurate diagnosis of isolated BPC is not always straightforward, which is especially critical in the context of fetomaternal medicine. A detailed description of posterior fossa malformations is to be preferred over unspecific terminology.

Delayed fenestration of Blake's pouch with or without vermian hypoplasia: fetal MRI at 3 tesla versus 1.5 tesla.

BACKGROUND: Fetal magnetic resonance imaging (MRI), mainly performed at standard field strength, plays a role in the classification of posterior fossa malformations. In the context of early second-trimester screening, upward rotation of the cerebellar vermis per se is usually compatible with a more favorable outcome than Dandy-Walker malformation and profound vermian hypoplasia. Delayed fenestration of Blake's pouch may either mimic vermian hypoplasia by compression or be associated with it in individual cases. To increase specificity, there is a growing interest in the use of high-field MRI which is believed to be safe as long as the specific absorption rate is kept within accepted limits. We aim to illustrate its added value during the second and third trimester. CASE PRESENTATION: In the first case, fetal MRI at 1.5 Tesla was performed at 21 and 27 weeks' gestation with sonographic follow up postnataly. In the second case, 3 Tesla MR images were acquired at 21 and 34 weeks' gestation as well as in the neonatal period. CONCLUSIONS: This pictorial case vignette supports the suggestion that mid-gestational MRI at 3 Tesla has the potential to exclude pronounced vermian hypoplasia with higher confidence than at 1.5 Tesla. However, the discrimination of mild hypoplasia from slight deformation of the cerebellar vermis will likely remain challenging.

Microvessels may Confound the "Swallow Tail Sign" in Normal Aged Midbrains: A Postmortem 7 T SW-MRI Study.

BACKGROUND AND PURPOSE: Susceptibility weighted imaging (SWI) plays a role in the differential diagnosis of Parkinson's disease, but lacks widespread acceptance in clinical routine. In a descriptive pilot study, we assessed hypointense microstructures of the normal substantia nigra pars compacta at ultrahigh-field strength for interpretation of the "swallow tail sign." METHODS: Magnetic resonance imaging at 7 Tesla was performed in five postmortem samples obtained from subjects not affected by Parkinson's disease. Susceptibility weighted images, including minimum intensity projections, were created followed by consensus assessment for microvascular confound. Histological workup in this case-control study included iron and myelin staining. Seven Tesla SWI images from the reference cohort of nine living subjects, all of which showed a positive "swallow tail sign" in their midbrains, were assessed visually. RESULTS: All specimens showed microvessels running through the dorsal pars compacta and along the caudolateral circumference of the red nucleus. Hypointense imaging patterns in the medial part of the "swallow tail" were due to susceptibility effects of iron deposits and microvessels. In eight out of nine control subjects, one or more microvessels were detected medial to the dorsolateral nigral hyperintensity or at least unilaterally in the medial part of the "swallow tail." One microvessel crossing nigrosome 1 was found in two in-vivo cases. CONCLUSION: Both iron deposits and microvessels contribute to the hyposignal surrounding nigrosome 1 in susceptibility weighted imaging of normal aged midbrains at ultrahigh-field strength. When assessing the substantia nigra for the presence or absence of the "swallow tail sign," intrinsic vessels may be a sporadic confounder.

Behavioral disorders and cognitive impairment associated with cerebellar lesions.

In the last decade evidence has accumulated that suggests that the cerebellum is involved not only in motor but also in behavioral and cognitive functions. A myriad of anatomical, clinical and imaging studies support that assumption. The lengthened survival of patients with cerebellar tumors has also brought an increased awareness of neurocognitive deficits to the neurooncological community. Although evidence from neurosurgical case series exists that clearly demonstrates that patients afflicted from posterior fossa tumors are at high risk for long-term cognitive or adaptive deficits, there is still a lack of systematic translational review on this issue. Accordingly a systematic review was conducted to summarize the impact of cerebellar lesions on behavior and cognition. The findings and clinical implications are discussed in the light of the recent advances in neuroimaging techniques.

Mechanical thrombectomy with stent retrievers in acute ischemic stroke.

Mechanical thrombectomy (MTE) in patients with acute ischemic infarct caused by large-vessel occlusion is becoming used with increasing frequency in many stroke centers. With the introduction of stent retrievers, recanalization rates >80 % are reached by most operators. However, although the technical success rate of MTE has been increased, clinical results have not improved to the same degree. In this review, the indications for MTE, the technique, and the technical and clinical outcomes are discussed. Complications and predictors for good clinical outcome are described based on recent data from the literature.

The "central vein sign": is there a place for susceptibility weighted imaging in possible multiple sclerosis?

OBJECTIVES: Susceptibility weighted imaging (SWI) may have the potential to depict the perivenous extent of white matter lesions (WMLs) in multiple sclerosis (MS). We aimed to assess the discriminatory value of the "central vein sign" (CVS). METHODS: In a 3-T magnetic resonance imaging (MRI) study, 28 WMLs in 14 patients with at least one circumscribed lesion >5 mm and not more than eight non-confluent lesions >3 mm were prospectively included. Only WMLs in FLAIR images with a maximum diameter of >5 mm were correlated to their SWI equivalent for CVS evaluation. RESULTS: Five patients fulfilled the revised McDonald criteria for MS and nine patients were given alternative diagnoses. Nineteen MS-WMLs and nine non-MS-WMLs >5 mm were detected. Consensus reading found a central vein in 16 out of 19 MS-WMLs (84 %) and in one out of nine non-MS-WMLs (11 %), respectively. The CVS proved to be a highly significant discriminator (P < 0.001) between MS-WMLs and non-MS-WMLs with a sensitivity, specificity, positive and negative predictive value and accuracy of 84 %, 89 %, 94 %, 73 % and 86 %, respectively. Inter-rater agreement was good (κ = 0.77). CONCLUSIONS: Even though the CVS is not exclusively found in MS-WMLs, SWI may be a useful adjunct in patients with possible MS. KEY POINTS: • MRI continues to yield further information concerning MS lesions. • SWI adds diagnostic information in patients with possible MS. • The "central vein sign" was predominantly seen in MS lesions. • The "central vein sign" helps discriminate between MS and non-MS lesions.

Chudley-McCullough syndrome: case report and review of the neuroimaging spectrum.

We report on a child with Chudley-McCullough syndrome and re-evaluate the spectrum of imaging findings (in 15 previously reported patients) which appear to be variable and, to some extent, ambiguous in the literature. Magnetic resonance imaging of the brain revealed asymmetric colpocephaly with agenesis of the splenium corporis callosi, ribbon-like subcortical gray matter heterotopia along the cingulate gyri, malrotation of both hippocampi, and dysplasia of the cerebellum. Macrocrania together with sensorineural hearing loss, colpocephaly, and posterior or complete agenesis of the corpus callosum can be considered the hallmarks of the autosomal recessive Chudley-McCullough syndrome. These may be variably associated with interhemispheric arachnoid cyst, cortical dysplasia, gray matter heterotopia, and cerebellar dysplasia. While early support with hearing aids may lead to improved language and cognitive outcome, shunting of ventricular dilatation is not indicated in the Chudley-McCullough syndrome.

MRI of the temporo-mandibular joint: which sequence is best suited to assess the cortical bone of the mandibular condyle? A cadaveric study using micro-CT as the standard of reference.

OBJECTIVE: To determine the best suited sagittal MRI sequence out of a standard temporo-mandibular joint (TMJ) imaging protocol for the assessment of the cortical bone of the mandibular condyles of cadaveric specimens using micro-CT as the standard of reference. METHODS: Sixteen TMJs in 8 human cadaveric heads (mean age, 81 years) were examined by MRI. Upon all sagittal sequences, two observers measured the cortical bone thickness (CBT) of the anterior, superior and posterior portions of the mandibular condyles (i.e. objective analysis), and assessed for the presence of cortical bone thinning, erosions or surface irregularities as well as subcortical bone cysts and anterior osteophytes (i.e. subjective analysis). Micro-CT of the condyles was performed to serve as the standard of reference for statistical analysis. RESULTS: Inter-observer agreements for objective (r = 0.83-0.99, P < 0.01) and subjective (κ = 0.67-0.88) analyses were very good. Mean CBT measurements were most accurate, and cortical bone thinning, erosions, surface irregularities and subcortical bone cysts were best depicted on the 3D fast spoiled gradient echo recalled sequence (3D FSPGR). CONCLUSION: The most reliable MRI sequence to assess the cortical bone of the mandibular condyles on sagittal imaging planes is the 3D FSPGR sequence. KEY POINTS: MRI may be used to assess the cortical bone of the TMJ. • Depiction of cortical bone is best on 3D FSPGR sequences. • MRI can assess treatment response in patients with TMJ abnormalities.

Increased cerebellar volume in the early stage of fucosidosis: a case control study.

INTRODUCTION: Yet unreported in this lysosomal storage disease, we aimed to quantify our observation that patients with fucosidosis may show abnormally increased cerebellar volumes during early childhood. METHODS: Five normocephalic fucosidosis patients (age range 2-25 months, three males) were included in this retrospective case control study. The control cohort consisted of 25 children (age range 0-36 months, 15 males). Image postprocessing was performed independently by two radiologists. Using validated software, manual tracing of contours on contiguous sagittal magnetic resonance images was allowed for cerebellar volumetry. We tested the null hypothesis that mean cerebellar volumes of four fucosidosis patients (age 16, 20, 21, and 25 months) and of an age-matched control cohort (n = 8, age range 13-26 months) were equal based on a two-tailed unpaired t-test. RESULTS: Interobserver agreement was excellent (R = 1, p < 0.01). A rough trajectory of normal cerebellar development appeared to flatten around the age of 1 year. With mean volumes of 121.36 and 102.30 ml, respectively, cerebellar volumes of fucosidosis patients with a mean age of 21 months were significantly increased compared to age-matched controls (p < 0.05). In a single patient, longer-term follow-up with MRI at the age of 47 months was available and showed cerebellar atrophy. CONCLUSION: Increased cerebellar volume was shown to be an additional feature in the early stage of fucosidosis. The combination with a confirmed tendency toward atrophy of the cerebellum during later course of the disease is probably unique in the context of metabolic disorders of the brain.

Dual-energy CT angiography in peripheral arterial occlusive disease-accuracy of maximum intensity projections in clinical routine and subgroup analysis.

OBJECTIVE: To evaluate the accuracy of dual-energy CT angiography (DE-CTA) maximum intensity projections (MIPs) in symptomatic peripheral arterial occlusive disease (PAOD). METHODS: In 58 patients, DE-CTA of the lower extremities was performed on dual-source CT. In a maximum of 35 arterial segments, severity of the most stenotic lesion was graded (<10%, 10-49% and 50-99% luminal narrowing or occlusion) independently by two radiologists, with DSA serving as the reference standard. RESULTS: In DSA, 52.3% of segments were significantly stenosed or occluded. Agreement of DE-CTA MIPs with DSA was good in the aorto-iliac and femoro-popliteal regions (κ = 0.72; κ = 0.66), moderate in the crural region (κ = 0.55), slight in pedal arteries (κ = 0.10) and very good in bypass segments (κ = 0.81). Accuracy was 88%, 78%, 74%, 55% and 82% for the respective territories and moderate (75%) overall, with good sensitivity (84%) and moderate specificity (67%). Sensitivity and specificity was 82% and 76% in claudicants and 84% and 61% in patients with critical limb ischaemia. CONCLUSION: While correlating well with DSA above the knee, accuracy of DE-CTA MIPs appeared to be moderate in the calf and largely insufficient in calcified pedal arteries, especially in patients with critical limb ischaemia.

Impact of PVCs noted during coronary calcium scan on image quality and accuracy in subsequent coronary dual-source CT angiography.

The purpose of this study was to investigate the impact of the presence of one or more premature ventricular contractions (PVCs) during coronary calcium score scan (CS) on image quality and accuracy of subsequent dual-source coronary CT angiography (DS CCTA). Fifty-three out of 502 patients showed one or more PVCs during CS and built the study group. Sixty consecutively registered patients with sinus rhythm formed the control group. Two independent, blinded readers classified 1,723 coronary artery segments as being of diagnostic or non-diagnostic image quality. All segments with diagnostic image quality were assessed for the presence for significant coronary artery stenosis. Accuracy was calculated using conventional angiography as standard of reference. The percentage of DS CCTA data sets with diagnostic image quality was significantly lower in the study group (79.2%) compared to the control group (90.6%, P < 0.05). The rate of coronary segment with diagnostic image quality was highly significant lower in the study group (89.5%; n = 723/808) compared to the control group (99.4%; n = 908/915; P < 0.001). However, accuracy did not differ in both groups. Significant correlation between non-diagnostic image quality and mean HR was only found in the study group (P = 0.019). The presence of one or more PVCs during CS does not affect accuracy but impairs image quality of subsequent DS CCTA. Mean HR is a predictor of non-diagnostic data sets only in the study group, suggesting more aggressive HR control could be beneficial.

Transsphenoidal extension of heterotopic glioneuronal tissue: pathoanatomic considerations in symptomatic neonates.

PURPOSE: In this clinical investigation, we aimed (1) to re-evaluate the nature of glioneuronal tissue with transsphenoidal extension and how it fits into the nomenclature of midline malformations and mass lesions; (2) to find out if our imaging findings support current pathoanatomic concepts of clefts and canals in the sphenoid body of newborns. METHODS: In two neonates with respiratory distress due to nasopharyngeal masses, 3T MRI was performed, and CT in one of them. Imaging features were analyzed in consensus by two pediatric neuroradiologists with histological reports being available. An interdisciplinary panel compared the findings to those of case publications and differential entities from our institutional case collection. RESULTS: Referring to our rare case of transsphenoidal cerebral heterotopia and unique case of hypothalamic hamartoma with transsphenoidal herniation, glioneuronal heterotopia may definitely extend through the sphenoid bone. Consequently, there is reason for brain heterotopias to be labeled as such also in case of an intracranial component. Connection between heterotopic glioneuronal tissue in the nasopharynx and a hypothalamic hamartoma may go along with indistinct margins to normal brain. Neither extension through a transsphenoidal cleft nor association with a cleft palate are specific for cerebral heterotopia. Our findings support the hypothesis that transsphenoidal cerebral heterotopias do not or at least not invariably follow the route of Rathke's pouch, known as the craniopharyngeal canal. CONCLUSION: Transsphenoidal glioneuronal heterotopia should be the top differential diagnosis in MR imaging if a non-enhancing nasopharyngeal mass of an infant extends through a craniopharyngeal cleft within the intersphenoid synchondrosis.

Fetal MRI of a hypothalamic hamartoma in Pallister-Hall syndrome.

Fetal magnetic resonance imaging is increasingly being used as an adjunct to ultrasound. It allows for better visualization of in utero brain development and intracranial abnormalities (especially cerebral malformations). Hypothalamic hamartoma is a nonneoplastic malformation resembling normal hypothalamic tissue both histologically and on magnetic resonance imaging. Although it is rare, this entity is important to recognize for appropriate management and genetic counseling. We describe a unique patient in whom magnetic resonance imaging of the fetal brain allowed a prenatal diagnosis of Pallister-Hall syndrome.

Feasibility and potential value of flat-panel detector-based computed tomography in myelography after spinal surgery.

OBJECT: The purpose of this study was to assess the value of myelography using flat-panel detector-based computed tomography (fpCT) in 5 patients in whom the image quality of multislice CT (MSCT) or MR imaging was limited by metal artifacts. METHODS: The application of fpCT to myelographic imaging of the lumbar spine and cervicothoracic junction after surgery was feasible. Multiplanar, preferably sagittal, and 3D reconstructions adequately depicted disc space implants and provided high resolution images of osseous structures. RESULTS: The images obtained with fpCT allowed evaluation of anatomical details such as single nerve roots and proved especially valuable in a patient with impaired MR imaging results caused by metal artifacts from an intraoperative abrasion. In a case of recurrent disc herniation, imaging results of myelographic fpCT and MSCT scanning were in good agreement. CONCLUSIONS: The novel imaging technique the authors describe yielded adequate results in patients with a history of spinal surgery, may be superior to MSCT scanning in depicting osseous structures and metallic implants, and has the potential to provide multilevel spinal images. Myelographic fpCT scanning may be the preferred modality in patients with expected or known metal artifacts on myelographic MSCT scans and/or MR images.

FDG PET/CT in the detection of recurrent rectal cancer.

The purpose of this study was to assess the additional value of combined fluorine 18 (18F)-fluorodeoxyglucose (FDG) positron emission tomography (PET)/computed tomography (CT) in the follow-up of rectal cancer after surgery. Forty-eight examinations in 30 patients were evaluated retrospectively. CT and PET components were interpreted separately, and this was followed by a consensus reading. Sites of increased FDG uptake as well as PET/CT findings were categorized as benign (1), equivocal (2), or malignant (3). The standard of reference was histology or clinical and imaging follow-up for at least 6 months. Sensitivity, specificity, positive and negative predictive values, and accuracy for differentiating benign (14/31) from malignant (17/31) uptake sites in the small pelvis were 100%, 64%, 77%, 100%, and 84% for PET/CT, and 100%, 29%, 63%, 100%, and 68% for PET, respectively. Regarding extrapelvic abnormalities, PET/CT was able to distinguish benign (31/88) from malignant (57/88) with a sensitivity, specificity, positive and negative predictive values, and accuracy of 100%, 87%, 93%, 100%, and 95%, compared with 96%, 68%, 85%, 91%, and 86% for PET. The rare case of an FDG uptake of adrenal adenoma is documented. PET/CT is valuable in the staging of rectal cancer, particularly for excluding recurrent disease suspected by PET interpretation alone in a considerable number of patients.