Search

With the rapid integration of artificial intelligence into medical practice, there has been an exponential increase in the number of scientific papers and industry players offering models designed for various tasks. Understanding these, however, is difficult for a radiologist in practice, given the core mathematical principles and complicated terminology involved. This review aims to elucidate the core mathematical concepts of both machine learning and deep learning models, explaining the various steps and common terminology in common layman language. Thus, by the end of this article, the reader should be able to understand the basics of how prediction models are built and trained, including challenges faced and how to avoid them. The reader would also be equipped to adequately evaluate various models, and take a decision on whether a model is likely to perform adequately in the real-world setting.

No This article is protected by copyright. All rights reserved.

Background: Primary extramedullary plasmacytomas (EMP) are rare; however, secondary forms may be seen in ~10-15% of patients with systemic multiple myeloma (MM). The diagnosis of EMP is based on the demonstration of monoclonal plasma cells in the lesion, which requires tissue sampling. Case Presentation: We present a case of a 38 year old female with MM who underwent diagnostic US at our institute. Multiple focal liver lesions were detected, which were suspicious for EMP. She underwent fine needle aspiration cytology (FNAC) for diagnosis, following which she developed hemoperitoneum secondary to deranged clotting parameters (prothrombin time and platelet count). CT angiography revealed active hepatic capsular bleed. She was taken up for percutaneous embolisation, and the supplying vessel successfully embolised using gel foam particles. Conclusion: Complications may rarely occur in interventional procedures, particularly in patients with comorbidities. However, prompt diagnosis and management help prevent adverse outcomes.

Ultrasonography (US) forms the mainstay of imaging in children; however, in the chest, its use has traditionally been limited to evaluation of pleural pathology. US techniques such as endobronchial and endoscopic ultrasound, which are commonly used for detection of mediastinal lymphadenopathy are invasive, aerosol generating, and often require sedation. Transcutaneous mediastinal sonography (TMUS) offers a useful alternative, which is easier to perform and overcomes these limitations. In this review, we summarize the technique, as well as imaging appearances of lymph nodes on TMUS. We also list common problems faced by operators and suggest troubleshooting methods for these.

This study aims to evaluate the utility of computed tomography angiography (CTA) signs of vascular injury in the differentiation of vessel transection from pure thrombosis with intact vessel wall.Retrospective analysis was done on 146 consecutive patients who had undergone CTA and surgical exploration from January 2015 to September 2019. Twelve imaging parameters were assessed. Chi-square was used to test the difference between groups. In addition, a scoring system was devised where one point each was added for the presence of 5 signs and absence of 3 signs. ROC analysis was done for the variables which had shown significant difference between groups and for the composite score.On surgical exploration, 87 patients had transection of vessel, while 59 had thrombosis. Significant difference was found among the two groups in non-opacification, pseudoaneurysm, extravasation (p = 0.04 each), thrombosed cord (p 0.001), collaterals (p = 0.001) and hematoma (p = 0.002), while other signs did not show significant difference. The AUC value for each of these variables was 0.650, while for the score, AUC was .843(.773-.913). A cut-off value of ≥ 1.5 gave 83.1% sensitivity and 70% specificity for diagnosing transection.CTA is a useful tool to classify the nature of vascular injury. It is advisable to use a composite score for maximum diagnostic value.

Ultrasonography (US) is the workhorse of pediatric imaging; however, lung US is only a recently developed application. US of the lung is based predominantly on the imaging of chest wall-air-fluid interfaces. In this review, we summarize the available literature on applications of lung US in neonatal as well as pediatric care. We describe the imaging appearance of various commonly encountered pathologies including pneumonia and respiratory distress syndrome, among others, and provide illustrative images. Finally, we describe the limitations of the technique that are essential knowledge for radiologists, critical care physicians, sonographers and technologists attempting to use lung US effectively for diagnosis and management.

To determine high resolution CT (HRCT) patterns of pulmonary fibrosis (PF) in children; and their etiological correlates.This was a retrospective study involving 149 children with diffuse lung disease (DLD). Patterns of involvement were classified based on dominant lung finding as ground glass opacity (GGO) dominant, nodule dominant, cystic lung disease, or PF. Patterns of PF were classified based on distribution and morphology into airway centric fibrosis (ACF), subpleural fibrosis (SPF), progressive massive fibrosis (PMF) and fibrocavitary. A comparison was made between the two dominant groups for apicobasal distribution, associated findings (GGO, nodules, cysts), and pulmonary artery hypertension (PAH).Nineteen patients showed PF on HRCT. ACF was commonest (52.6%), followed by SPF (42.1%). The common etiology was sarcoidosis (30%) in ACF, and connective tissue disorders (CTD) (50%) in SPF. Significant difference was found between ACF and SPF in apicobasal distribution (p = 0.04), presence of nodules (p = 0.03), and cysts (p = 0.02).PF may present as an end stage of several childhood lung diseases. PF on imaging has discernible morphological patterns that correlate with underlying etiology.

Purpose: To assess differentiating features between bacterial, Aspergillus, and Mucor skull base osteomyelitis (SBO) with regard to clinical presentation and imaging appearances. Material and methods: A literature search was performed in April 2020 for studies on SBO with a minimum sample size of 10 patients. Studies that reported presenting symptoms, cross-sectional imaging findings, complications, and mortality were included in the analysis. The quality of included articles was tested using the Quality Assessment of Diagnostic Accuracy Studies 2 tool. A data extraction form was used to retrieve relevant parameters from each of the articles. Results: Thirteen articles were included in the final analysis. Diabetes mellitus was the most common predisposing factor (12.5-91.0%). Presenting complaints in all bacterial SBO studies were otogenic, while fungal SBO patients had nasal/ocular complaints. Rates of mortality and surgical intervention in the fungal group were 50-100% and 50%, respectively, as compared to the bacterial group - 7-87% and 10%, respectively. On imaging, the site of initial infection in bacterial SBO was the external auditory canal, while in fungal SBO it was the paranasal sinus. The incidenceof orbital extension was < 5% in bacterial and 44-70% in fungal SBO, among which Mucor had rates of 65-70%. Bone erosion was less extensive in bacterial SBO, and the patterns differed. The highest incidence of vascular involvement and non-enhancing lesions (23-36%) was seen in Mucor. Aspergillus showed highest sino-cranial extension (52-55%) and homogenous bright enhancement. Conclusions: Systematic analysis of the clinico-radiological parameters in each of the studies revealed differences in presentation, clinical course, extension, bone erosion, and enhancement.

Background Pelvic US remains the workhorse for detection and characterization of adnexal masses in most centers worldwide. While the differentiation of benign from malignant masses remains the foremost concern, it is imperative to narrow the differential diagnosis for management of benign masses as well as prognostication of malignant masses. The IOTA group as well as ACR have described a five category classification system for adnexal lesions based on morphological patterns. In addition, a six category risk stratification has been proposed, incorporating the probability of malignancy as well as management recommendations. Learning objectives 1) Understand pattern based approach to adnexal lesion classification and the possible entities fitting into each pattern with the help of illustrations. 2) Classify lesions into appropriate risk categories based on diagnostic algorithms provided at the end of each section.