Journal of the Belgian Society of Radiology最新文献

Teaching Point: Retinal detachment is a rare initial clinical manifestation of lung cancer with intraorbital metastases, early diagnosis on magnetic resonance imaging is important for therapeutic implications.

Teaching point: Benign hyperostosis of the rib is a benign entity consisting of a stress phenomenon that should not be confused with Paget, fibrous dysplasia, or osteoblastic metastasis.

Teaching point: Small bowel diverticulitis, much less common than its colonic counterpart, is a diagnosis that must be considered in the presence of abdominal pain, especially in an elderly person.

Teaching point: Although rare, an intra-mammary metastasis from extramammary cancer should be considered in a patient with an oncological history.

Teaching Point: Hepatic alveolar echinococcosis can mimic a slow-growing tumor, and multi-organ involvement is rare; imaging has a crucial role in diagnosing this zoonosis that is endemic in the southern part of Belgium.

This is a case of barotrauma imaging (Macklin effect) after invasive mechanical ventilation in a 14-week-old newborn with complicated bronchiolitis. Teaching point: Imaging could help us improve defining the anatomical boundaries of the Macklin effect, an incompletely known anatomo-physiological entity.

Large gastric hernias are common and usually cause minor symptoms. Rarely, complete intrathoracic herniation of the stomach is complicated by strangulation. The underlying mechanism can be gastric volvulus or the less recognized phenomenon of gastric fundus redescent. We describe a case where this rare but potentially lethal complication of gastric herniation is present. Additionally, we show that gastric pneumatosis, a sign associated with ischemia, can be initially visualized on a plain chest radiograph in this setting. Teaching point: Redescent of the fundus is a possible, but unrecognized cause of gastric strangulation in intrathoracic stomachs.

Teaching point: Magnetic resonance imaging (MRI) has significantly improved the evaluation of brachial plexus injuries, offering new possibilities for microsurgical repair and contributing to the functional prognosis.

Aneurysmal dilatations can affect any aortic segment and represent the result of various causes, atherosclerotic disease being the most common and frequently involved. We hereby illustrate a case of a patient with thoracic aortic aneurysm rupture due to extensive atherosclerotic disease, with multiple complex penetrating ulcerated atherosclerotic plaques located in the descending aorta. CT angiography evaluation included a comprehensive description of imaging features and extent of the thoracic aortic aneurysm, the presence of thrombus, relationship to adjacent structures and branches, associated complications. Teaching Point: Thoracic aortic aneurysm rupture due to extensive atherosclerotic disease with multiple penetrating ulcers.

Objectives: To evaluate the performances of machine learning using semantic and radiomic features from magnetic resonance imaging data to distinguish cystic pituitary adenomas (CPA) from Rathke's cleft cysts (RCCs).

Materials and methods: The study involved 65 patients diagnosed with either CPA or RCCs. Multiple observers independently assessed the semantic features of the tumors on the magnetic resonance images. Radiomics features were extracted from T2-weighted, T1-weighted, and T1-contrast-enhanced images. Machine learning models, including Support Vector Machines (SVM), Logistic Regression (LR), and Light Gradient Boosting (LGB), were then trained and validated using semantic features only and a combination of semantic and radiomic features. Statistical analyses were carried out to compare the performance of these various models.

Results: Machine learning models that combined semantic and radiomic features achieved higher levels of accuracy than models with semantic features only. Models with combined semantic and T2-weighted radiomics features achieved the highest test accuracies (93.8%, 92.3%, and 90.8% for LR, SVM, and LGB, respectively). The SVM model combined semantic features with T2-weighted radiomics features had statistically significantly better performance than semantic features only (p = 0.019).

Conclusion: Our study demonstrates the significant potential of machine learning for differentiating CPA from RCCs.