This page gives an overview of title, presenter and abstract of all the poster presentations in the Poster Session 2 of the 2021 edition of the Point of Care Ultrasound conference. During this session five posters will be presented covering a broad spectrum of Point of Care Ultrasound topics.
Carotid Screening: How Much Information Does a Single M-line Carry about Arterial Structural and Material Stiffness?
Afrah Malik, CARIM School for Cardiovascular Diseases, Maastricht University Medical Centre
Purpose: Common carotid artery structural and material stiffness can be evaluated with high accuracy and precision using multiple M-line ultrasound. This study aims to assess whether a single M-line approach would be sufficient for screening purposes in low resource settings and areas.
Methods: We used common carotid data of 500 subjects randomly selected from The Maastricht Study to compare accuracy and reproducibility of single against multiple M-line approach. The triplicate longitudinal carotid M-line recordings were composed of 17 simultaneous lines at a frame rate of 498 Hz covering an area of 16.32 mm of the artery, enabling quantification of diameter, distension, intima-media thickness (IMT), distensibility coefficient (DC), and Young’s elastic modulus (YEM). The Multi M-line measure was obtained by averaging over the 17 lines, whereas the middle line was used for the single M-line approach. Paired measurements were assessed by Wilcoxon test, Bland-Altman analysis, and reproducibility coefficient of variation (CV).
Results: Diameter and IMT were not significantly different for the single and multiple approaches (p>0.14). However, distension and DC were significantly larger and YEM significantly lower (p<0.05) for single M-line compared to multiple M-line. Bland-Altman analysis showed good agreement between the approaches with biases±1.96SD of 4±95µm for diameter (mean=7793µm), -5±62µm for IMT (mean=854µm), -4±20µm for distension (mean=388µm), -0.2±0.8MPa^(-1) for DC (mean=15MPa^(-1)) and 0.0005±0.1126MPa for YEM (mean=0.71MPa). Differences in reproducibility CVs were statistically significant (p<0.001), however, moderate in magnitude. Single vs multiple values for CV were (2.5% vs 2.2%) for diameter, (11.8% vs 7.5%) for IMT, (11.7% vs 10.8%) for distension, (12.5% vs 11.5%) for DC and (19.2% vs 16.1%) for YEM.
Conclusion: Single M-line ultrasound can be used to assess arterial structural and material stiffness with very acceptable precision and accuracy. Image free, single M-line tools could support screening and population studies in low resource settings and areas globally.
Infertility is emerging as a serious problem in developed countries, where up to 20% of women in reproductive age have difficulty to get pregnant. In-vitro fertilization (IVF) represents the most advanced treatment against infertility. Yet, its success rate remains below 30%. Dysfunction of the uterine activity is believed to affect the success rate of embryo implantation.
In this study, we investigate novel uterine motion features and the use of machine learning for probabilistic classification of the uterine activity as either favorable or adverse to embryo implantation.
Uterine activity was measured by transvaginal ultrasound (TVUS) in 16 patients undergoing a complete IVF cycle. In this study, we analyzed the measurements taken one hour before embryo transfer.
A set of features, such as contraction frequency (CF), amplitude, power, velocity, direction, and coordination were extracted from the motion signals derived by a dedicated speckle tracking algorithm.
Three classifiers, namely, support vector machine, K-nearest neighbors (KNN), and Gaussian mixture model were considered in this study to classify successful and unsuccessful embryo implantation using the extracted features. The proposed classifiers were tested and trained in a nested cross validation loop to avoid overfitting. A full-grid search was adopted for hyperparameter optimization. The classifiers were validated with a leave-one-out approach. Accuracy was used as performance metric.
The best classification accuracy (93.8%) was obtained by KNN. CF and coordination were best predictors for embryo implantation. A larger dataset is required in the future to improve the robustness and generalizability of the classifiers.
3D dynamic contrast-enhanced ultrasound (DCE-US) enables analyzing the vasculature of the full prostate gland following a single ultrasound-contrast-agent (UCA) bolus injection. As angiogenesis is a recognized hallmark of prostate cancer (PCa), we have proposed contrast-ultrasound dispersion imaging (CUDI) techniques to characterize the UCA kinetics, and further analyze the underlying (micro)vasculature. The feasibility of a multiparametric approach combining complementary CUDI parameters has been investigated for PCa localization.
3D DCE-US recordings were carried out on 43 patients. All participants underwent 12-core systematic biopsy. For each recording, 16 CUDI parameters were obtained by performing 1D convective-dispersion model-fitting analysis, similarity analysis, system-identification analysis, 3D convective-dispersion modeling, and fractal dimension analysis. For each prostate, 12 regions corresponding to the biopsy locations were determined. The mean value of each CUDI parameter in the 12 regions was then calculated. A multiparametric approach for PCa classification was then implemented by a Gaussian Mixture Model (GMM). A leave-one-prostate-out cross-validation procedure was employed to assess the classification accuracy.
The proposed GMM-based multiparametric approach produced an area under the ROC curve of 0.76 and 0.81 for all PCa and significant PCa, respectively, showing superior performance compared to the individual CUDI parameters in distinguishing malignant and benign disease.
Multiparametric CUDI based on 3D DCE-US acquisitions shows potential for PCa localization. We recommend an improved validation by inclusion of full-prostate histopathological references. Moreover, extension of the dataset is required to improve the generalizability of our results.
Over 200 million people worldwide are affected by peripheral vascular diseases (PVDs), including arterial disease, chronic venous insufficiency, and deep vein thrombosis. Vascular ultrasonography remains a gold-standard method to measure luminal narrowing in stenotic vessels and diagnose PVD. Currently, however, obtaining accurate lumen diameter measurements with ultrasound requires expertise, which limits utility outside of dedicated imaging labs. We describe a method to assist vascular ultrasound users in identifying vessels and assessing lumen diameter in a fast and reliable manner. A deep learning model was developed that automatically detects and segments vascular structures as they appear during live scanning. The model utilizes both B-mode and Color Doppler image signals, and incorporates spatial and time domain information, to identify vessels of interest. Focusing on lower leg arterial examinations, we trained and evaluated the model on ultrasound videos obtained from femoral and infrapopliteal arterial scans in healthy human volunteers and perfused cadaver legs (7 subjects, 13 legs, 22 video sequences, 30,470 total frames). We obtained ground-truth annotated masks of the arterial segments in 400 video frames from two expert vascular sonographers, and used these to generate annotations for all remaining frames via a semi-automated approach. Comparing model predictions to ground-truth, we saw that incorporation of Color Doppler and time domain information allowed model performance to approach the inter-rater agreement between experts (0.92 ± 0.023 median Dice score for the model based on k-fold cross validation, as compared to 0.95 median Dice score for inter-rater agreement). Future work will investigate the clinical performance of automated arterial measurements in a prospective setting, as a means to improve workflow during ultrasound-based diagnostic evaluation of PVD.
Detection of clinically significant prostate cancer in biopsy-naïve men: direct comparison of systematic biopsy, multiparametric MRI- and contrast-ultrasound-dispersion imaging-targeted biopsy
Auke Jager, Eindhoven University of Technology / Amsterdam University Medical Centers
In this prospective, single-centre, head-to-head, diagnostic accuracy trial the detection rate of clinically significant prostate cancer (csPCa) was evaluated for systematic biopsy (SBx), contrast-ultrasound-dispersion imaging (CUDI) – targeted biopsy (TBx) and multiparametric magnetic resonance imaging – TBx.
150 Biopsy-naïve men underwent pre-biopsy mpMRI and CUDI. SBx were taken from all study participants by a clinician blinded for imaging results. In case of lesions suspected for PCa on mpMRI and/or CUDI, additional MRI-TRUS fusion-TBx and/or cognitive CUDI-TBx was performed by a second operator. csPCa was defined as International Society of Urological Pathology Grade Group (GG) ≥ 2.
The interim analysis, performed after inclusion of 150 men, csPCa detection rates for mpMRI-TBx and CUDI-TBx were 29% (41/142) and 28% (41/142), respectively. SBx performed significantly (P<0.05) better than both targeted biopsy strategies, detecting 39% csPCa (56/142).
Conclusion and discussion
The current study resulted in similar csPCa detection rates for CUDI-TBx and mpMRI-TBx, thereby showing the potential of ultrasound imaging quantification algorithms as a diagnostic imaging modality. A major shortcoming of CUDI, as performed in this trial, was the 2D image acquisition.
In march of 2021, a multicentre study will commence in the Netherlands (NCT04605276). The goal of this study is to collect 3D greyscale and 4D Contrast Enhanced Ultrasound Images from approximately 715 patients before prostate biopsies and/or radical prostatectomy. Using full mount prostate histopathology as the reference standard, we aim to improve the ultrasound image analyse algorithm. Preliminary assessment of the algorithm will be performed to determine sensitivity and specificity.
Click here for the full conference program
This conference is organised by Jakajima.