RESUMEN
OBJECTIVES: Achieving a consensus on a definition for different aspects of radiomics workflows to support their translation into clinical usage. Furthermore, to assess the perspective of experts on important challenges for a successful clinical workflow implementation. MATERIALS AND METHODS: The consensus was achieved by a multi-stage process. Stage 1 comprised a definition screening, a retrospective analysis with semantic mapping of terms found in 22 workflow definitions, and the compilation of an initial baseline definition. Stages 2 and 3 consisted of a Delphi process with over 45 experts hailing from sites participating in the German Research Foundation (DFG) Priority Program 2177. Stage 2 aimed to achieve a broad consensus for a definition proposal, while stage 3 identified the importance of translational challenges. RESULTS: Workflow definitions from 22 publications (published 2012-2020) were analyzed. Sixty-nine definition terms were extracted, mapped, and semantic ambiguities (e.g., homonymous and synonymous terms) were identified and resolved. The consensus definition was developed via a Delphi process. The final definition comprising seven phases and 37 aspects reached a high overall consensus (> 89% of experts "agree" or "strongly agree"). Two aspects reached no strong consensus. In addition, the Delphi process identified and characterized from the participating experts' perspective the ten most important challenges in radiomics workflows. CONCLUSION: To overcome semantic inconsistencies between existing definitions and offer a well-defined, broad, referenceable terminology, a consensus workflow definition for radiomics-based setups and a terms mapping to existing literature was compiled. Moreover, the most relevant challenges towards clinical application were characterized. CRITICAL RELEVANCE STATEMENT: Lack of standardization represents one major obstacle to successful clinical translation of radiomics. Here, we report a consensus workflow definition on different aspects of radiomics studies and highlight important challenges to advance the clinical adoption of radiomics. KEY POINTS: Published radiomics workflow terminologies are inconsistent, hindering standardization and translation. A consensus radiomics workflow definition proposal with high agreement was developed. Publicly available result resources for further exploitation by the scientific community.
RESUMEN
Diffusion-weighted magnetic resonance imaging (DW-MRI) has evolved to provide increasingly sophisticated investigations of the human brain's structural connectome in vivo. Restriction spectrum imaging (RSI) is a method that reconstructs the orientation distribution of diffusion within tissues over a range of length scales. In its original formulation, RSI represented the signal as consisting of a spectrum of Gaussian diffusion response functions. Recent technological advances have enabled the use of ultra-high b-values on human MRI scanners, providing higher sensitivity to intracellular water diffusion in the living human brain. To capture the complex diffusion time dependence of the signal within restricted water compartments, we expand upon the RSI approach to represent restricted water compartments with non-Gaussian response functions, in an extended analysis framework called linear multi-scale modeling (LMM). The LMM approach is designed to resolve length scale and orientation-specific information with greater specificity to tissue microstructure in the restricted and hindered compartments, while retaining the advantages of the RSI approach in its implementation as a linear inverse problem. Using multi-shell, multi-diffusion time DW-MRI data acquired with a state-of-the-art 3 T MRI scanner equipped with 300 mT/m gradients, we demonstrate the ability of the LMM approach to distinguish different anatomical structures in the human brain and the potential to advance mapping of the human connectome through joint estimation of the fiber orientation distributions and compartment size characteristics.
Asunto(s)
Conectoma , Imagen de Difusión por Resonancia Magnética , Humanos , Imagen de Difusión por Resonancia Magnética/métodos , Encéfalo/diagnóstico por imagen , Encéfalo/fisiología , Algoritmos , AguaRESUMEN
OBJECTIVE: Before implementing radiomics in routine clinical practice, comprehensive knowledge about the repeatability and reproducibility of radiomic features is required. The aim of this study was to systematically investigate the influence of image processing parameters on radiomic features from magnetic resonance imaging (MRI) in terms of feature values as well as test-retest repeatability. MATERIALS AND METHODS: Utilizing a phantom consisting of 4 onions, 4 limes, 4 kiwifruits, and 4 apples, we acquired a test-retest dataset featuring 3 of the most commonly used MRI sequences on a 3 T scanner, namely, a T1-weighted, a T2-weighted, and a fluid-attenuated inversion recovery sequence, each at high and low resolution. After semiautomatic image segmentation, image processing with systematic variation of image processing parameters was performed, including spatial resampling, intensity discretization, and intensity rescaling. For each respective image processing setting, a total of 45 radiomic features were extracted, corresponding to the following 7 matrices/feature classes: conventional indices, histogram matrix, shape matrix, gray-level zone length matrix, gray-level run length matrix, neighboring gray-level dependence matrix, and gray-level cooccurrence matrix. Systematic differences of individual features between different resampling steps were assessed using 1-way analysis of variance with Tukey-type post hoc comparisons to adjust for multiple testing. Test-retest repeatability of radiomic features was measured using the concordance correlation coefficient, dynamic range, and intraclass correlation coefficient. RESULTS: Image processing influenced radiological feature values. Regardless of the acquired sequence and feature class, significant differences ( P < 0.05) in feature values were found when the size of the resampled voxels was too large, that is, bigger than 3 mm. Almost all higher-order features depended strongly on intensity discretization. The effects of intensity rescaling were negligible except for some features derived from T1-weighted sequences. For all sequences, the percentage of repeatable features (concordance correlation coefficient and dynamic range ≥ 0.9) varied considerably depending on the image processing settings. The optimal image processing setting to achieve the highest percentage of stable features varied per sequence. Irrespective of image processing, the fluid-attenuated inversion recovery sequence in high-resolution overall yielded the highest number of stable features in comparison with the other sequences (89% vs 64%-78% for the respective optimal image processing settings). Across all sequences, the most repeatable features were generally obtained for a spatial resampling close to the originally acquired voxel size and an intensity discretization to at least 32 bins. CONCLUSION: Variation of image processing parameters has a significant impact on the values of radiomic features as well as their repeatability. Furthermore, the optimal image processing parameters differ for each MRI sequence. Therefore, it is recommended that these processing parameters be determined in corresponding test-retest scans before clinical application. Extensive repeatability, reproducibility, and validation studies as well as standardization are required before quantitative image analysis and radiomics can be reliably translated into routine clinical care.
Asunto(s)
Procesamiento de Imagen Asistido por Computador , Imagen por Resonancia Magnética , Reproducibilidad de los Resultados , Procesamiento de Imagen Asistido por Computador/métodos , Fantasmas de ImagenRESUMEN
OBJECTIVES: The aim of this study was to compare a new compressed sensing (CS) method for T2-weighted propeller acquisitions (T2 CS ) with conventional T2-weighted propeller sequences (T2 conv ) in terms of achieving a higher image quality, while reducing the acquisition time. MATERIALS AND METHODS: Male participants with a clinical suspicion of prostate cancer were prospectively enrolled and underwent prostate magnetic resonance imaging at 3 T. Axial and sagittal images of the T2 conv sequence and the T2 CS sequence were acquired. Sequences were qualitatively assessed by 2 blinded radiologists concerning artifacts, image-sharpness, lesion conspicuity, capsule delineation, and overall image quality using 5-point Likert items ranging from 1 (nondiagnostic) to 5 (excellent). The apparent signal-to-noise ratio and apparent contrast-to-noise ratio were evaluated. PI-RADS scores were assessed for both sequences. Statistical analysis was performed by using Wilcoxon signed rank test and paired samples t test. Intrarater and interrater reliability of qualitative image evaluation was assessed using intraclass correlation coefficient (ICC) estimates. RESULTS: A total of 29 male participants were included (mean age, 66 ± 8 years). The acquisition time of the T2 CS sequence was respectively 26% (axial plane) and 24% (sagittal plane) shorter compared with the T2 conv sequence (eg, axial: 171 vs 232 seconds; P < 0.001). In the axial plane, the T2 CS sequence had fewer artifacts (4 [4-4.5] vs 4 [3-4]; P < 0.001), better image-sharpness (4 [4-4.5] vs 3 [3-3.5]; P < 0.001), better capsule delineation (4 [3-4] vs 3 [3-3.5]; P < 0.001), and better overall image quality (4 [4-4] vs 4 [3-4]; P < 0.001) compared with the T2 conv sequence. The ratings of lesion conspicuity were similar (4 [4-4] vs 4 [3-4]; P = 0.166). In the sagittal plane, the T2 CS sequence outperformed the T2 conv sequence in the categories artifacts (4 [4-4] vs 3 [3-4]; P < 0.001), image sharpness (4 [4-5] vs 4 [3-4]; P < 0.001), lesion conspicuity (4 [4-4] vs 4 [3-4]; P = 0.002), and overall image quality (4 [4-4] vs 4 [3-4]; P = 0.002). Capsule delineation was similar between both sequences (3 [3-4] vs 3 [3-3]; P = 0.07). Intraobserver and interobserver reliability for qualitative scoring were good (ICC intra: 0.92; ICC inter: 0.86). Quantitative analysis revealed a higher apparent signal-to-noise ratio (eg, axial: 52.2 ± 9.7 vs 22.8 ± 3.6; P < 0.001) and a higher apparent contrast-to-noise ratio (eg, axial: 44.0 ± 9.6 vs 18.6 ± 3.7; P ≤ 0.001) of the T2 CS sequence. PI-RADS scores were the same for both sequences in all participants. CONCLUSIONS: CS-accelerated T2-weighted propeller acquisition had a superior image quality compared with conventional T2-weighted propeller sequences while significantly reducing the acquisition time.
Asunto(s)
Imagen por Resonancia Magnética , Neoplasias de la Próstata , Humanos , Masculino , Persona de Mediana Edad , Anciano , Imagen por Resonancia Magnética/métodos , Próstata/diagnóstico por imagen , Reproducibilidad de los Resultados , Neoplasias de la Próstata/diagnóstico por imagen , Relación Señal-Ruido , ArtefactosRESUMEN
This retrospective study aims to evaluate the generalizability of a promising state-of-the-art multitask deep learning (DL) model for predicting the response of locally advanced rectal cancer (LARC) to neoadjuvant chemoradiotherapy (nCRT) using a multicenter dataset. To this end, we retrained and validated a Siamese network with two U-Nets joined at multiple layers using pre- and post-therapeutic T2-weighted (T2w), diffusion-weighted (DW) images and apparent diffusion coefficient (ADC) maps of 83 LARC patients acquired under study conditions at four different medical centers. To assess the predictive performance of the model, the trained network was then applied to an external clinical routine dataset of 46 LARC patients imaged without study conditions. The training and test datasets differed significantly in terms of their composition, e.g., T-/N-staging, the time interval between initial staging/nCRT/re-staging and surgery, as well as with respect to acquisition parameters, such as resolution, echo/repetition time, flip angle and field strength. We found that even after dedicated data pre-processing, the predictive performance dropped significantly in this multicenter setting compared to a previously published single- or two-center setting. Testing the network on the external clinical routine dataset yielded an area under the receiver operating characteristic curve of 0.54 (95% confidence interval [CI]: 0.41, 0.65), when using only pre- and post-therapeutic T2w images as input, and 0.60 (95% CI: 0.48, 0.71), when using the combination of pre- and post-therapeutic T2w, DW images, and ADC maps as input. Our study highlights the importance of data quality and harmonization in clinical trials using machine learning. Only in a joint, cross-center effort, involving a multidisciplinary team can we generate large enough curated and annotated datasets and develop the necessary pre-processing pipelines for data harmonization to successfully apply DL models clinically.
RESUMEN
OBJECTIVES: To develop a pipeline for automated body composition analysis and skeletal muscle assessment with integrated quality control for large-scale application in opportunistic imaging. METHODS: First, a convolutional neural network for extraction of a single slice at the L3/L4 lumbar level was developed on CT scans of 240 patients applying the nnU-Net framework. Second, a 2D competitive dense fully convolutional U-Net for segmentation of visceral and subcutaneous adipose tissue (VAT, SAT), skeletal muscle (SM), and subsequent determination of fatty muscle fraction (FMF) was developed on single CT slices of 1143 patients. For both steps, automated quality control was integrated by a logistic regression model classifying the presence of L3/L4 and a linear regression model predicting the segmentation quality in terms of Dice score. To evaluate the performance of the entire pipeline end-to-end, body composition metrics, and FMF were compared to manual analyses including 364 patients from two centers. RESULTS: Excellent results were observed for slice extraction (z-deviation = 2.46 ± 6.20 mm) and segmentation (Dice score for SM = 0.95 ± 0.04, VAT = 0.98 ± 0.02, SAT = 0.97 ± 0.04) on the dual-center test set excluding cases with artifacts due to metallic implants. No data were excluded for end-to-end performance analyses. With a restrictive setting of the integrated segmentation quality control, 39 of 364 patients were excluded containing 8 cases with metallic implants. This setting ensured a high agreement between manual and fully automated analyses with mean relative area deviations of ΔSM = 3.3 ± 4.1%, ΔVAT = 3.0 ± 4.7%, ΔSAT = 2.7 ± 4.3%, and ΔFMF = 4.3 ± 4.4%. CONCLUSIONS: This study presents an end-to-end automated deep learning pipeline for large-scale opportunistic assessment of body composition metrics and sarcopenia biomarkers in clinical routine. KEY POINTS: ⢠Body composition metrics and skeletal muscle quality can be opportunistically determined from routine abdominal CT scans. ⢠A pipeline consisting of two convolutional neural networks allows an end-to-end automated analysis. ⢠Machine-learning-based quality control ensures high agreement between manual and automatic analysis.
Asunto(s)
Sarcopenia , Composición Corporal , Humanos , Músculo Esquelético/diagnóstico por imagen , Control de Calidad , Sarcopenia/diagnóstico por imagen , Tomografía Computarizada por Rayos X/métodosRESUMEN
BACKGROUND: Diffusion-weighted imaging (DWI) is an essential component of the multiparametric MRI exam for the diagnosis and assessment of prostate cancer (PCa). Over the last two decades, various models have been developed to quantitatively correlate the DWI signal with microstructural characteristics of prostate tissue. The simplest approach (ADC: apparent diffusion coefficient) - currently established as the clinical standard - describes monoexponential decay of the DWI signal. While numerous studies have shown an inverse correlation of ADC values with the Gleason score, the ADC model lacks specificity and is based on water diffusion dynamics that are not true in human tissue. This article aims to explain the biophysical limitations of the standard DWI model and to discuss the potential of more complex, advanced DWI models. METHODS: This article is a review based on a selective literature review. RESULTS: Four phenomenological DWI models are introduced: diffusion tensor imaging, intravoxel incoherent motion, biexponential model, and diffusion kurtosis imaging. Their parameters may potentially improve PCa diagnostics but show varying degrees of statistical significance with respect to the detection and characterization of PCa in current studies. Phenomenological model parameters lack specificity, which has motivated the development of more descriptive tissue models that directly relate microstructural features to the DWI signal. Finally, we present two of such structural models, i.âe. the VERDICT (Vascular, Extracellular, and Restricted Diffusion for Cytometry in Tumors) and RSI (Restriction Spectrum Imaging) model. Both have shown promising results in initial studies regarding the characterization and prognosis of PCa. CONCLUSION: Recent developments in DWI techniques promise increasing accuracy and more specific statements about microstructural changes of PCa. However, further studies are necessary to establish a standardized DWI protocol for the diagnosis of PCa. KEY POINTS: · DWI is paramount to the mpMRI exam for the diagnosis of PCa.. · Though of clinical value, the ADC model lacks specificity and oversimplifies tissue complexities.. · Advanced phenomenological and structural models have been developed to describe the DWI signal.. · Phenomenological models may improve diagnostics but show inconsistent results regarding PCa assessment.. · Structural models have demonstrated promising results in initial studies regarding PCa characterization.. CITATION FORMAT: · Wichtmann BD, Zöllner FG, Attenberger UI etâal. Multiparametric MRI in the Diagnosis of Prostate Cancer: Physical Foundations, Limitations, and Prospective Advances of Diffusion-Weighted MRI. Fortschr Röntgenstr 2021; 193: 399â-â409.
Asunto(s)
Imágenes de Resonancia Magnética Multiparamétrica , Neoplasias de la Próstata , Imagen de Difusión por Resonancia Magnética , Imagen de Difusión Tensora , Humanos , Masculino , Imágenes de Resonancia Magnética Multiparamétrica/normas , Estudios Prospectivos , Neoplasias de la Próstata/diagnóstico por imagenRESUMEN
OBJECTIVE: The aim of this study was to evaluate a contrast media (CM)-saline mixture administration with DualFlow (DF) to adapt injection protocols to low-kilovolt (kV) computed tomography angiography (CTA). MATERIALS AND METHODS: In both a circulation phantom and animal model (5 Goettingen minipigs), 3 injection protocols were compared in dynamic thoracic CTA: (a) DF injection protocol at 80 kV with a iodine delivery rate (IDR) of 0.9 gI/s, a flowrate of 5 mL/s injected with a 60%/40% ratio of iopromide (300 mgI/mL) and saline (dose contrast medium 180 mgI/kg body weight [BW]); (b) reference CTA was performed at 120 kV and a 40% higher iodine dose applied at higher IDR (1.5 gI/s, 5 mL/s iopromide [300 mgI/mL]; no simultaneously administered saline; 300 mgI/kg BW); and (c) conventional single-flow (SF) protocol with identical IDR as the DF protocol at 80 kV (0.9 gI/s, 3 mL/s iopromide [300 mgI/mL]; no simultaneously administered saline; 180 mgI/kg BW). All 3 injection protocols are followed by a saline chaser applied at the same flow rate as the corresponding CM injection. Time attenuation curves representing the vascular bolus shape were generated for pulmonary trunk and descending aorta. RESULTS: In the circulation phantom, pulmonary and aortic time attenuation curves for the 80 kV DF injection protocols do not significantly differ from the 80 kV SF and the 120 kV SF reference. In the animal model, the 80 kV DF protocol shows similar pulmonal and aortic peak enhancement when compared with the 120 kV SF and 80 kV SF protocols. Also, the bolus length above an attenuation level of 300 HU reveals no significant differences between injection protocols. However, the time to peak was significantly shorter for the 80 kV DF when compared with the 80 kV SF protocol (15.78 ± 1.9 seconds vs 18.24 ± 2.0 seconds; P = 0.008). CONCLUSION: DualFlow injection protocols can be tailored for low-kV CTA by reducing the IDR while overall flow rate remains unchanged. Although no differences in attenuation were found, DF injections offer a shorter time to peak closer to the reference 120 kV protocol.This allows the use of DF injection protocols to calibrate bolus density in low-kV CTA and yields the potential for a more individualized CM administration.
Asunto(s)
Angiografía por Tomografía Computarizada/instrumentación , Fantasmas de Imagen , Solución Salina/administración & dosificación , Animales , Medios de Contraste/administración & dosificación , Femenino , Humanos , Inyecciones , Masculino , Arteria Pulmonar/diagnóstico por imagen , Porcinos , Porcinos EnanosRESUMEN
Today's health care environment is shifting rapidly, driven by demographic change and high economic pressures on the system. Furthermore, modern precision medicine requires highly accurate and specific disease diagnostics in a short amount of time. Future imaging technology must adapt to these challenges.Demographic change necessitates scanner technologies tailored to the needs of an aging and increasingly multimorbid patient population. Accordingly, examination times have to be short enough that diagnostic images can be generated even for patients who can only lie in the scanner for a short time because of pain or with low breath-hold capacity.For economic reasons, the rate of nondiagnostic scans due to artifacts should be reduced as far as possible. As imaging plays an increasingly pivotal role in clinical-therapeutic decision making, magnetic resonance (MR) imaging facilities are confronted with an ever-growing number of patients, emphasizing the need for faster acquisitions while maintaining image quality.Lastly, modern precision medicine requires high and standardized image quality as well as quantifiable data in order to develop image-based biomarkers on which subsequent treatment management can rely.In recent decades, a variety of approaches have addressed the challenges of high throughput, demographic change, and precision medicine in MR imaging. These include field strength, gradient, coil and sequence development, as well as an increasing consideration of artificial intelligence. This article reviews state-of-the art MR technology and discusses future implementation from the perspective of what we know today.
Asunto(s)
Atención a la Salud , Imagen por Resonancia Magnética/métodos , Artefactos , Inteligencia Artificial , Contencion de la Respiración , Humanos , Procesamiento de Imagen Asistido por ComputadorRESUMEN
Axon diameter and density are important microstructural metrics that offer valuable insight into the structural organization of white matter throughout the human brain. We report the systematic acquisition and analysis of a comprehensive diffusion MRI data set acquired with 300 mT/m maximum gradient strength in a cohort of 20 healthy human subjects that yields distinct and consistent patterns of axon diameter index in white matter tracts of arbitrary orientation. We use a straightforward, previously validated approach to estimating indices of axon diameter and volume fraction that involves interpolating the diffusion signal perpendicular to the principal fiber orientation and fitting a three-compartment model of intra-axonal, extra-axonal and free water diffusion. The resultant maps confirm the presence of larger diameter indices in the body of corpus callosum compared to the genu and splenium, as previously reported, and show larger axon diameter index in the corticospinal tracts compared to adjacent white matter tracts such as the cingulum. An anterior-to-posterior gradient in axon diameter index is also observed, with smaller diameter indices in the frontal lobes and larger diameter indices in the parieto-occipital white matter. These observations are consistent with known trends from prior histologic studies in humans and non-human primates. Rather than serving as fully quantitative measures of axon diameter and density, our results may be considered as axon diameter- and volume fraction-weighted images that appear to be modulated by the underlying microstructure and may capture broad trends in axonal size and packing density, acknowledging that the precise origin of such modulation requires further investigation that will be facilitated by the availability of high gradient strengths for in vivo human imaging.
Asunto(s)
Axones , Encéfalo/citología , Imagen de Difusión por Resonancia Magnética , Sustancia Blanca/citología , Adulto , Femenino , Humanos , Procesamiento de Imagen Asistido por Computador/métodos , Masculino , Modelos NeurológicosRESUMEN
We provide a comprehensive diffusion MRI dataset acquired with a novel biomimetic phantom mimicking human white matter. The fiber substrates in the diffusion phantom were constructed from hollow textile axons ("taxons") with an inner diameter of 11.8±1.2⯵m and outer diameter of 33.5±2.3⯵m. Data were acquired on the 3â¯T CONNECTOM MRI scanner with multiple diffusion times and multiple q-values per diffusion time, which is a dedicated acquisition for validation of microstructural imaging methods, such as compartment size and volume fraction mapping. Minimal preprocessing was performed to correct for susceptibility and eddy current distortions. Data were deposited in the XNAT Central database (project ID: dMRI_Phant_MGH).
RESUMEN
Diffusion microstructural imaging techniques have attracted great interest in the last decade due to their ability to quantify axon diameter and volume fraction in healthy and diseased human white matter. The estimates of compartment size and volume fraction continue to be debated, in part due to the lack of a gold standard for validation and quality control. In this work, we validate diffusion MRI estimates of compartment size and volume fraction using a novel textile axon ("taxon") phantom constructed from hollow polypropylene yarns with distinct intra- and extra-taxonal compartments to mimic white matter in the brain. We acquired a comprehensive set of diffusion MRI measurements in the phantom using multiple gradient directions, diffusion times and gradient strengths on a human MRI scanner equipped with maximum gradient strength (Gmax) of 300â¯mT/m. We obtained estimates of compartment size and restricted volume fraction through a straightforward extension of the AxCaliber/ActiveAx frameworks that enables estimation of mean compartment size in fiber bundles of arbitrary orientation. The voxel-wise taxon diameter estimates of 12.2⯱â¯0.9⯵m were close to the manufactured inner diameter of 11.8⯱â¯1.2⯵m with Gmaxâ¯=â¯300â¯mT/m. The estimated restricted volume fraction demonstrated an expected decrease along the length of the fiber bundles in accordance with the known construction of the phantom. When Gmax was restricted to 80â¯mT/m, the taxon diameter was overestimated, and the estimates for taxon diameter and packing density showed greater uncertainty compared to data with Gmaxâ¯=â¯300â¯mT/m. In conclusion, the compartment size and volume fraction estimates resulting from diffusion measurements on a human scanner were validated against ground truth in a phantom mimicking human white matter, providing confidence that this method can yield accurate estimates of parameters in simplified but realistic microstructural environments. Our work also demonstrates the importance of a biologically analogous phantom that can be applied to validate a variety of diffusion microstructural imaging methods in human scanners and be used for standardization of diffusion MRI protocols for neuroimaging research.