RESUMO
The successful integration of neural networks in a clinical setting is still uncommon despite major successes achieved by artificial intelligence in other domains. This is mainly due to the black box characteristic of most optimized models and the undetermined generalization ability of the trained architectures. The current work tackles both issues in the radiology domain by focusing on developing an effective and interpretable cardiomegaly detection architecture based on segmentation models. The architecture consists of two distinct neural networks performing the segmentation of both cardiac and thoracic areas of a radiograph. The respective segmentation outputs are subsequently used to estimate the cardiothoracic ratio, and the corresponding radiograph is classified as a case of cardiomegaly based on a given threshold. Due to the scarcity of pixel-level labeled chest radiographs, both segmentation models are optimized in a semi-supervised manner. This results in a significant reduction in the costs of manual annotation. The resulting segmentation outputs significantly improve the interpretability of the architecture's final classification results. The generalization ability of the architecture is assessed in a cross-domain setting. The assessment shows the effectiveness of the semi-supervised optimization of the segmentation models and the robustness of the ensuing classification architecture.
Assuntos
Inteligência Artificial , Cardiomegalia , Humanos , Cardiomegalia/diagnóstico por imagem , Generalização Psicológica , Coração , Processamento de Imagem Assistida por Computador , Redes Neurais de ComputaçãoRESUMO
CT protocols that diagnose COVID-19 vary in regard to the associated radiation exposure and the desired image quality (IQ). This study aims to evaluate CT protocols of hospitals participating in the RACOON (Radiological Cooperative Network) project, consolidating CT protocols to provide recommendations and strategies for future pandemics. In this retrospective study, CT acquisitions of COVID-19 patients scanned between March 2020 and October 2020 (RACOON phase 1) were included, and all non-contrast protocols were evaluated. For this purpose, CT protocol parameters, IQ ratings, radiation exposure (CTDIvol), and central patient diameters were sampled. Eventually, the data from 14 sites and 534 CT acquisitions were analyzed. IQ was rated good for 81% of the evaluated examinations. Motion, beam-hardening artefacts, or image noise were reasons for a suboptimal IQ. The tube potential ranged between 80 and 140 kVp, with the majority between 100 and 120 kVp. CTDIvol was 3.7 ± 3.4 mGy. Most healthcare facilities included did not have a specific non-contrast CT protocol. Furthermore, CT protocols for chest imaging varied in their settings and radiation exposure. In future, it will be necessary to make recommendations regarding the required IQ and protocol parameters for the majority of CT scanners to enable comparable IQ as well as radiation exposure for different sites but identical diagnostic questions.
RESUMO
BACKGROUND: Arterial spin labeling (ASL) is a magnetic resonance imaging (MRI)-based technique using labeled blood-water of the brain-feeding arteries as an endogenous tracer to derive information about brain perfusion. It enables the assessment of cerebral blood flow (CBF). METHOD: This review aims to provide a methodological and technical overview of ASL techniques, and to give examples of clinical use cases for various diseases affecting the central nervous system (CNS). There is a special focus on recent developments including super-selective ASL (ssASL) and time-resolved ASL-based magnetic resonance angiography (MRA) and on diseases commonly not leading to characteristic alterations on conventional structural MRI (e.âg., concussion or migraine). RESULTS: ASL-derived CBF may represent a clinically relevant parameter in various pathologies such as cerebrovascular diseases, neoplasms, or neurodegenerative diseases. Furthermore, ASL has also been used to investigate CBF in mild traumatic brain injury or migraine, potentially leading to the establishment of imaging-based biomarkers. Recent advances made possible the acquisition of ssASL by selective labeling of single brain-feeding arteries, enabling spatial perfusion territory mapping dependent on blood flow of a specific preselected artery. Furthermore, ASL-based MRA has been introduced, providing time-resolved delineation of single intracranial vessels. CONCLUSION: Perfusion imaging by ASL has shown promise in various diseases of the CNS.âGiven that ASL does not require intravenous administration of a gadolinium-based contrast agent, it may be of particular interest for investigations in pediatric cohorts, patients with impaired kidney function, patients with relevant allergies, or patients that undergo serial MRI for clinical indications such as disease monitoring. KEY POINTS: · ASL is an MRI technique that uses labeled blood-water as an endogenous tracer for brain perfusion imaging.. · It allows the assessment of CBF without the need for administration of a gadolinium-based contrast agent.. · CBF quantification by ASL has been used in several pathologies including brain tumors or neurodegenerative diseases.. · Vessel-selective ASL methods can provide brain perfusion territory mapping in cerebrovascular diseases.. · ASL may be of particular interest in patient cohorts with caveats concerning gadolinium administration..
Assuntos
Transtornos Cerebrovasculares , Transtornos de Enxaqueca , Doenças Neurodegenerativas , Humanos , Criança , Meios de Contraste , Marcadores de Spin , Gadolínio , Imageamento por Ressonância Magnética/métodos , Artérias , Angiografia por Ressonância Magnética/métodos , Transtornos Cerebrovasculares/diagnóstico por imagem , ÁguaRESUMO
PURPOSE: Technical feasibility of CT-based calculation of fractional flow reserve (cFFR) using a 128-row computed tomography scanner in an everyday routine setting. Post-processing and everyday practicability should be analyzed on the scanner on-site in connection with clinical parameters. MATERIALS AND METHODS: This single-center retrospective analysis included 230 patients (74 female; mean age 63.8 years) with CCTA within 21 months between 01/2018 and 09/2019 without non-pathological examinations. cFFR values were obtained using a deep learning-based non-commercial research prototype (cFFR Version3.5.0; Siemens Healthineers GmbH, Erlangen). cFFR values were evaluated at two points: at the maximum point of the stenosis and 1.0âcm distal to the stenosis. Comparison with invasive coronary angiography in 57/230 patients (24.7â%) was performed. CT parameters and quality were evaluated. Further subgroup classification concerning criteria of technical postprocessing was performed: no changes necessary, minor corrections necessary, major corrections necessary, and no evaluation was possible. The required time from starting the software to the final result was evaluated. RESULTS: A total of 116/448 (25.9â%) mild, 223/448 (49.8â%) moderate, and 109/448 (24.3â%) obstructive stenoses was found. The mean cFFR at the maximum point of the stenosis was 0.92â±â0.09 and significantly higher than the cFRR value of 0.89â±â0.13 distal to the stenosis (pâ<â0.001*). The mean degree of stenosis was 44.02â±â26.99â% (range: 1-99â%) with an area of 5.39â±â3.30âmm2. In a total of 45 patients (19.1â%), a relevant reduction in cFFR below 0.80 was determined. Overall, in 57/230 patients (24.8â%), catheter angiography was performed. No significant difference in the degree of maximal stenosis (CAD-RADS 0-2/3/4) was detected between the classification of CCTA and ICA (pâ=â0.171). The mean post-processing time varied significantly with 8.34â±â4.66âmin. in single-vessel CADâvs. 12.91â±â3.92âmin. in two-vessel CADâvs. 21.80â± 5.94âmin. in three-vessel CADâ(each pâ<â0.001). CONCLUSION: Noninvasive onsite quantification of cFFR is feasible with minimal observer interaction in a routine real-world setting on a 128-row scanner. Deep learning-based algorithms allow a robust and semi-automatic on-site determination of cFFR based on data from standard CT scanners. KEY POINTS: · Non-invasive on-site quantification of cFFR is feasible with minimal observer interaction.. · Deep-learning based algorithms allow robust and semi-automatic on-site determination of cFFR.. · The mean follow-up time varied significantly with the extent of vascular CAD..
Assuntos
Doença da Artéria Coronariana , Estenose Coronária , Reserva Fracionada de Fluxo Miocárdico , Humanos , Feminino , Pessoa de Meia-Idade , Doença da Artéria Coronariana/diagnóstico por imagem , Estudos Retrospectivos , Estenose Coronária/diagnóstico por imagem , Constrição Patológica , Estudos de Viabilidade , Angiografia por Tomografia Computadorizada/métodos , Valor Preditivo dos Testes , Angiografia Coronária/métodosRESUMO
Imaging of the temporal bone and middle ear is challenging for radiologists due to the abundance of distinct anatomical structures and the plethora of possible pathologies. The basis for a precise diagnosis is knowledge of the underlying anatomy as well as the clinical presentation and the individual patient's otological status. In this article, we aimed to summarize the most common inflammatory lesions of the temporal bone and middle ear, describe their specific imaging characteristics, and highlight their differential diagnoses. First, we introduce anatomical and imaging fundamentals. Additionally, a point-to-point comparison of the radiological and histological features of the wide spectrum of inflammatory diseases of the temporal bone and middle ear in context with a review of the current literature and current trends is given.
Assuntos
Otopatias , Humanos , Otopatias/diagnóstico por imagem , Otopatias/patologia , Tomografia Computadorizada por Raios X/métodos , Orelha Média/diagnóstico por imagem , Osso Temporal/diagnóstico por imagem , Osso Temporal/patologiaRESUMO
In medical imaging, deep learning models serve as invaluable tools for expediting diagnoses and aiding specialized medical professionals in making clinical decisions. However, effectively training deep learning models typically necessitates substantial quantities of high-quality data, a resource often lacking in numerous medical imaging scenarios. One way to overcome this deficiency is to artificially generate such images. Therefore, in this comparative study we train five generative models to artificially increase the amount of available data in such a scenario. This synthetic data approach is evaluated on a a downstream classification task, predicting four causes for pneumonia as well as healthy cases on 1082 chest X-ray images. Quantitative and medical assessments show that a Generative Adversarial Network (GAN)-based approach significantly outperforms more recent diffusion-based approaches on this limited dataset with better image quality and pathological plausibility. We show that better image quality surprisingly does not translate to improved classification performance by evaluating five different classification models and varying the amount of additional training data. Class-specific metrics like precision, recall, and F1-score show a substantial improvement by using synthetic images, emphasizing the data rebalancing effect of less frequent classes. However, overall performance does not improve for most models and configurations, except for a DreamBooth approach which shows a +0.52 improvement in overall accuracy. The large variance of performance impact in this study suggests a careful consideration of utilizing generative models for limited data scenarios, especially with an unexpected negative correlation between image quality and downstream classification improvement.
RESUMO
Since the beginning of the COVID-19 pandemic, many different machine learning models have been developed to detect and verify COVID-19 pneumonia based on chest X-ray images. Although promising, binary models have only limited implications for medical treatment, whereas the prediction of disease severity suggests more suitable and specific treatment options. In this study, we publish severity scores for the 2358 COVID-19 positive images in the COVIDx8B dataset, creating one of the largest collections of publicly available COVID-19 severity data. Furthermore, we train and evaluate deep learning models on the newly created dataset to provide a first benchmark for the severity classification task. One of the main challenges of this dataset is the skewed class distribution, resulting in undesirable model performance for the most severe cases. We therefore propose and examine different augmentation strategies, specifically targeting majority and minority classes. Our augmentation strategies show significant improvements in precision and recall values for the rare and most severe cases. While the models might not yet fulfill medical requirements, they serve as an appropriate starting point for further research with the proposed dataset to optimize clinical resource allocation and treatment.
Assuntos
COVID-19 , Pandemias , Humanos , Benchmarking , Aprendizado de Máquina , Rememoração MentalRESUMO
In medical imaging, deep learning models can be a critical tool to shorten time-to-diagnosis and support specialized medical staff in clinical decision making. The successful training of deep learning models usually requires large amounts of quality data, which are often not available in many medical imaging tasks. In this work we train a deep learning model on university hospital chest X-ray data, containing 1082 images. The data was reviewed, differentiated into 4 causes for pneumonia, and annotated by an expert radiologist. To successfully train a model on this small amount of complex image data, we propose a special knowledge distillation process, which we call Human Knowledge Distillation. This process enables deep learning models to utilize annotated regions in the images during the training process. This form of guidance by a human expert improves model convergence and performance. We evaluate the proposed process on our study data for multiple types of models, all of which show improved results. The best model of this study, called PneuKnowNet, shows an improvement of + 2.3% points in overall accuracy compared to a baseline model and also leads to more meaningful decision regions. Utilizing this implicit data quality-quantity trade-off can be a promising approach for many scarce data domains beyond medical imaging.
Assuntos
Aprendizado Profundo , Pneumonia , Humanos , Curadoria de Dados , Pneumonia/diagnóstico por imagem , Diagnóstico por ImagemRESUMO
PURPOSE: To implement the technical feasibility of an AI-based software prototype optimized for the detection of COVID-19 pneumonia in CT datasets of the lung and the differentiation between other etiologies of pneumonia. METHODS: This single-center retrospective case-control-study consecutively yielded 144 patients (58 female, mean age 57.72 ± 18.25 y) with CT datasets of the lung. Subgroups including confirmed bacterial (n = 24, 16.6%), viral (n = 52, 36.1%), or fungal (n = 25, 16.6%) pneumonia and (n = 43, 30.7%) patients without detected pneumonia (comparison group) were evaluated using the AI-based Pneumonia Analysis prototype. Scoring (extent, etiology) was compared to reader assessment. RESULTS: The software achieved an optimal sensitivity of 80.8% with a specificity of 50% for the detection of COVID-19; however, the human radiologist achieved optimal sensitivity of 80.8% and a specificity of 97.2%. The mean postprocessing time was 7.61 ± 4.22 min. The use of a contrast agent did not influence the results of the software (p = 0.81). The mean evaluated COVID-19 probability is 0.80 ± 0.36 significantly higher in COVID-19 patients than in patients with fungal pneumonia (p < 0.05) and bacterial pneumonia (p < 0.001). The mean percentage of opacity (PO) and percentage of high opacity (PHO ≥ -200 HU) were significantly higher in COVID-19 patients than in healthy patients. However, the total mean HU in COVID-19 patients was -679.57 ± 112.72, which is significantly higher than in the healthy control group (p < 0.001). CONCLUSION: The detection and quantification of pneumonia beyond the primarily trained COVID-19 datasets is possible and shows comparable results for COVID-19 pneumonia to an experienced reader. The advantages are the fast, automated segmentation and quantification of the pneumonia foci.
RESUMO
Magnetic resonance imaging (MRI) in therapy-naïve intracranial glioma is paramount for neuro-oncological diagnostics, and it provides images that are helpful for surgery planning and intraoperative guidance during tumor resection, including assessment of the involvement of functionally eloquent brain structures. This study reviews emerging MRI techniques to depict structural information, diffusion characteristics, perfusion alterations, and metabolism changes for advanced neuro-oncological imaging. In addition, it reflects current methods to map brain function close to a tumor, including functional MRI and navigated transcranial magnetic stimulation with derived function-based tractography of subcortical white matter pathways. We conclude that modern preoperative MRI in neuro-oncology offers a multitude of possibilities tailored to clinical needs, and advancements in scanner technology (e.âg., parallel imaging for acceleration of acquisitions) make multi-sequence protocols increasingly feasible. Specifically, advanced MRI using a multi-sequence protocol enables noninvasive, image-based tumor grading and phenotyping in patients with glioma. Furthermore, the add-on use of preoperatively acquired MRI data in combination with functional mapping and tractography facilitates risk stratification and helps to avoid perioperative functional decline by providing individual information about the spatial location of functionally eloquent tissue in relation to the tumor mass. KEY POINTS:: · Advanced preoperative MRI allows for image-based tumor grading and phenotyping in glioma.. · Multi-sequence MRI protocols nowadays make it possible to assess various tumor characteristics (incl. perfusion, diffusion, and metabolism).. · Presurgical MRI in glioma is increasingly combined with functional mapping to identify and enclose individual functional areas.. · Advancements in scanner technology (e.âg., parallel imaging) facilitate increasing application of dedicated multi-sequence imaging protocols.. CITATION FORMAT: · Sollmann N, Zhang H, Kloth C etâal. Modern preoperative imaging and functional mapping in patients with intracranial glioma. Fortschr Röntgenstr 2023; 195: 989â-â1000.
Assuntos
Neoplasias Encefálicas , Glioma , Humanos , Neoplasias Encefálicas/diagnóstico por imagem , Neoplasias Encefálicas/cirurgia , Imagem de Tensor de Difusão/métodos , Mapeamento Encefálico/métodos , Glioma/diagnóstico por imagem , Glioma/cirurgia , Encéfalo/patologia , Imageamento por Ressonância Magnética/métodosRESUMO
The imaging evaluation of computed tomography (CT), CT angiography (CTA), and CT perfusion (CTP) is of crucial importance in the setting of each emergency department for suspected cerebrovascular impairment. A fast and clear assignment of characteristic imaging findings of acute stroke and its differential diagnoses is essential for every radiologist. Different entities can mimic clinical signs of an acute stroke, thus the knowledge and fast identification of stroke mimics is important. A fast and clear assignment is necessary for a correct diagnosis and a rapid initiation of appropriate therapy. This pictorial review describes the most common imaging findings in CTP with clinical signs for acute stroke or other acute neurological disorders. The knowledge of these pictograms is therefore essential and should also be addressed in training and further education of radiologists.
RESUMO
Due to the increasing use of cross-sectional imaging techniques and new technical possibilities, the number of incidentally detected cystic lesions of the pancreas is rapidly increasing in everyday radiological routines. Precise and rapid classification, including targeted therapeutic considerations, is of essential importance. The new European guideline should also support this. This review article provides information on the spectrum of cystic pancreatic lesions, their appearance, and a comparison of morphologic and histologic characteristics. This is done in the context of current literature and clinical value. The recommendations of the European guidelines include statements on conservative management as well as relative and absolute indications for surgery in cystic lesions of the pancreas. The guidelines suggest surgical resection for mucinous cystic neoplasm (MCN) ≥ 40 mm; furthermore, for symptomatic MCN or imaging signs of malignancy, this is recommended independent of its size (grade IB recommendation). For main duct IPMNs (intraductal papillary mucinous neoplasms), surgical therapy is always recommended; for branch duct IPMNs, a number of different risk criteria are applicable to evaluate absolute or relative indications for surgery. Based on imaging characteristics of the most common cystic pancreatic lesions, a precise diagnostic classification of the tumor, as well as guidance for further treatment, is possible through radiology.
RESUMO
In recent years, several deep learning approaches have been successfully applied in the field of medical image analysis. More specifically, different deep neural network architectures have been proposed and assessed for the detection of various pathologies based on chest X-ray images. While the performed assessments have shown very promising results, most of them consist in training and evaluating the performance of the proposed approaches on a single data set. However, the generalization of such models is quite limited in a cross-domain setting, since a significant performance degradation can be observed when these models are evaluated on data sets stemming from different medical centers or recorded under different protocols. The performance degradation is mostly caused by the domain shift between the training set and the evaluation set. To alleviate this problem, different unsupervised domain adaptation approaches are proposed and evaluated in the current work, for the detection of cardiomegaly based on chest X-ray images, in a cross-domain setting. The proposed approaches generate domain invariant feature representations by adapting the parameters of a model optimized on a large set of labeled samples, to a set of unlabeled images stemming from a different data set. The performed evaluation points to the effectiveness of the proposed approaches, since the adapted models outperform optimized models which are directly applied to the evaluation sets without any form of domain adaptation.
RESUMO
BACKGROUND: Structured reporting allows a high grade of standardization and thus a safe and unequivocal report communication. In the past years, the radiological societies have started several initiatives to base radiological reports on structured reporting rather than free text reporting. METHODS: Upon invitation of the working group for Cardiovascular Imaging of the German Society of Radiology, in 2018 an interdisciplinary group of Radiologists, Cardiologists, Pediatric Cardiologists and Cardiothoracic surgeons -all experts on the field of cardiovascular MR and CT imaging- met for interdisciplinary consensus meetings at the University Hospital Cologne. The aim of these meetings was to develop and consent templates for structured reporting in cardiac MR and CT of various cardiovascular diseases. RESULTS: Two templates for structured reporting of CMR in ischemia imaging and vitality imaging and two templates for structured reporting of CT imaging for planning Transcatheter Aortic Valve Implantation (TAVI; pre-TAVI-CT) and coronary CT were discussed, consented and transferred to a HTML 5/IHR MRRT compatible format. The templates were made available for free use on the website www.befundung.drg.de. CONCLUSION: This paper suggests consented templates in German language for the structured reporting of cross-sectional CMR imaging of ischemia and vitality as well as reporting of CT imaging pre-TAVI and coronary CT. The implementation of these templates is aimed at providing a constant level of high reporting quality and increasing the efficiency of report generation as well as a clinically based communication of imaging results. KEY POINTS: · Structured reporting offers a constant level of high reporting quality and increases the efficiency of report generation as well as a clinically based communication of imaging results.. · For the first time templates in German language for the structured reporting of CMR imaging of ischemia and vitality and CT imaging pre-TAVI and coronary CT are reported.. · These templates will be made available on the website www.befundung.drg.de and can be commented via strukturierte-befundung@drg.de.. ZITIERWEISE: · Soschynski M, Bunck AC, Beer M etâal. Structured Reporting in Cross-Sectional Imaging of the Heart: Reporting Templates for CMR Imaging of Ischemia and Myocardial Viability and for Cardiac CT Imaging of Coronary Heart Disease and TAVI Planning. Fortschr Röntgenstr 2023; 195: 293â-â296.