Person:

Romero, Javier

Purpose We sought to develop a standardized curriculum capable of assessing key competencies in Interventional Neuroradiology by the use of models and simulators in an objective, quantitative, and efficient way. In this evaluation we analyzed the associations between the practical experience, theoretical knowledge, and the skills lab performance of interventionalists. Materials and Methods We evaluated the endovascular skills of 26 participants of the Advanced Course in Endovascular Interventional Neuroradiology of the European Society of Neuroradiology with a set of three tasks (aneurysm coiling and thrombectomy in a virtual simulator and placement of an intra-aneurysmal flow disruptor in a flow model). Practical experience was assessed by a survey. Participants completed a written and oral examination to evaluate theoretical knowledge. Bivariate and multivariate analyses were performed. Results: In multivariate analysis knowledge of materials and techniques in Interventional Neuroradiology was moderately associated with skills in aneurysm coiling and thrombectomy. Experience in mechanical thrombectomy was moderately associated with thrombectomy skills, while age was negatively associated with thrombectomy skills. We found no significant association between age, sex, or work experience and skills in aneurysm coiling. Conclusion: Our study gives an example of how an integrated curriculum for reasonable and cost-effective assessment of key competences of an interventional neuroradiologist could look. In addition to traditional assessment of theoretical knowledge practical skills are measured by the use of endovascular simulators yielding objective, quantitative, and constructive data for the evaluation of the current performance status of participants as well as the evolution of their technical competency over time.

Owing to improvements in image recognition via deep learning, machine-learning algorithms could eventually be applied to automated medical diagnoses that can guide clinical decision-making. However, these algorithms remain a 'black box' in terms of how they generate the predictions from the input data. Also, high-performance deep learning requires large, high-quality training datasets. Here, we report the development of an understandable deep-learning system that detects acute intracranial haemorrhage (ICH) and classifies five ICH subtypes from unenhanced head computed-tomography scans. By using a dataset of only 904 cases for algorithm training, the system achieved a performance similar to that of expert radiologists in two independent test datasets containing 200 cases (sensitivity of 98% and specificity of 95%) and 196 cases (sensitivity of 92% and specificity of 95%). The system includes an attention map and a prediction basis retrieved from training data to enhance explainability, and an iterative process that mimics the workflow of radiologists. Our approach to algorithm development can facilitate the development of deep-learning systems for a variety of clinical applications and accelerate their adoption into clinical practice.