Clinical and angiographic outcomes in patients with intracranial aneurysms treated with the pipeline embolization device: intra-procedural technical difficulties, major morbidity, and neurological mortality decrease significantly with increased operator experience in device deployment and patient management.

PURPOSE: Flow diversion constitutes a pivotal advancement in endovascular intracranial aneurysm treatment, but requires development of a new skill set. The aim of this study is to determine whether outcomes after treatment with the Pipeline Embolization Device improve with experience. METHODS: We retrospectively reviewed all patients with intracranial aneurysms treated with Pipeline at two centers over a 4.5-year period. Baseline patient and aneurysm characteristics, complications, and angiographic outcomes were analyzed. RESULTS: One hundred forty patients underwent 150 Pipeline procedures to treat 167 intracranial aneurysms during the study period, 109 women, mean age 55.4 years. One hundred twenty-six aneurysms were ICA, mean size 10.2 mm and mean neck 6.4 mm. Intra-procedural technical difficulties were higher during the first 75 procedures compared with the subsequent 75 (13.3 vs 2.7%; p = 0.03), as combined major morbidity and neurological mortality (14.7 vs 4%; p = 0.046). In multivariate regression analysis, increased operator experience with Pipeline remained an independent predictor of intra-procedural technical difficulties (p = 0.02, odds ratio (OR) 0.015, 95% CI 0.0004-0.55) and combined major morbidity and neurological mortality (p = 0.03, OR 0.16, 95% CI 0.03-0.84). At last follow-up, 123 aneurysms were completely occluded (81.5%, mean 24 months). In our cohort, age ≤ 53 years was an independent predictor of complete aneurysm occlusion at last follow-up (p = 0.001, OR 0.92, 95% CI 0.88-0.97). Five aneurysms were retreated (3.3%). CONCLUSION: The Pipeline embolization device is an effective treatment for intracranial aneurysms. The risk of intra-procedural technical difficulties and combined major morbidity and neurological mortality decreases significantly with increased operator experience in Pipeline deployment and patient management.

Hematoma Mimicking Breast Cancer on CT Scan and Breast Ultrasound.

There are many benign breast lesions that mimic breast cancer on breast imaging. Postlumpectomy scar, hematoma, fat necrosis, diabetic mastopathy, and granulomatous mastitis are examples of benign breast lesions that have suspicious breast imaging findings. Mammogram and breast ultrasound are the imaging studies to evaluate breast findings. CT scan is not used to evaluate breast findings because it delivers high radiation dose to the breast, and breast tissue is often confused as breast masses on CT scan. The following case demonstrates an incidentally detected breast mass on CT scan performed to assess for pulmonary embolism. The CT scan and subsequent breast ultrasound both demonstrated suspicious breast imaging findings. Final pathology from ultrasound-guided biopsy revealed hematoma. This benign finding was concordant with the patient's medical history of cirrhosis with low platelet count and medication history of warfarin.

MRI Compatibility: Automatic Brain Shunt Valve Recognition using Feature Engineering and Deep Convolutional Neural Networks.

The aim of this study is to evaluate whether we could develop a machine learning method to distinguish models of cerebrospinal fluid shunt valves (CSF-SVs) from their appearance in clinical X-rays. This is an essential component of an automatic MRI safety system based on X-ray imaging. To this end, a retrospective observational study using 416 skull X-rays from unique subjects retrieved from a clinical PACS system was performed. Each image included a CSF-SV representing the most common brands of programmable shunt valves currently used in US which were split into five different classes. We compared four machine learning pipelines: two based on engineered image features (Local Binary Patterns and Histogram of Oriented Gradients) and two based on features learned by a deep convolutional neural network architecture. Performance is evaluated using accuracy, precision, recall and f1-score. Confidence intervals are computed with non-parametric bootstrap procedures. Our best performing method identified the valve type correctly 96% [CI 94-98%] of the time (CI: confidence intervals, precision 0.96, recall 0.96, f1-score 0.96), tested using a stratified cross-validation approach to avoid chances of overfitting. The machine learning pipelines based on deep convolutional neural networks showed significantly better performance than the ones based on engineered image features (mean accuracy 95-96% vs. 56-64%). This study shows the feasibility of automatically distinguishing CSF-SVs using clinical X-rays and deep convolutional neural networks. This finding is the first step towards an automatic MRI safety system for implantable devices which could decrease the number of patients that experience denials or delays of their MRI examinations.

The mapping of the visual field onto the dorso-lateral tectum of the pigeon (Columba livia) and its relations with retinal specializations.

Most of the physiological studies of the pigeon retino-tectal visual pathway have investigated the accessible tectum, a small dorso-lateral tectal section that can be easily accessed by a simple craniotomy. However, at present we lack a detailed study of the topographical arrangement between the visual field, the retina and the accessible tectum. In particular, it is not known which section of the visual field is mapped onto the accessible tectum, and which of the specialized retinal areas mediates this projection. Here we determined, using local field potential (LFP) recordings and reverse retinoscopy, the shape, size and position in the visual space of the portion of the visual field mapped onto the accessible tectum (called here the accessible visual field, or AVF). Using this data and the mapping of Nalbach et al. [Vis. Res. 30 (4) (1990) 529], the retinal area corresponding to the AVF was determined. Such retinal area was also directly delimited by means of retrograde transport of DiI. The results indicate that the AVF is a triangular perifoveal zone encompassing only 15% of total visual field. The retinal region corresponding to the AVF has the shape of an elongated triangle that runs parallel to the visual equator and contains the fovea, the tip of the pecten, a perifoveal region of the yellow field and a small crescent of the red field. In agreement with this anatomical heterogeneity, visual evoked potentials measured in different parts of the accessible tectum present steep variations in shape and size. These results are helpful to better design and interpret anatomical and physiological experiments involving the pigeon's visual system.