Explicit and automatic ejection fraction assessment on 2D cardiac ultrasound with a deep learning-based approach

et al. 2022

Preprint

Background: Variation of inferior vena cava (IVC) is used to predict fluid-responsiveness, but the IVC visualization with standard sagittal approach (SC, subcostal) cannot be always achieved. In such cases, coronal trans-hepatic (TH) window may offer an alternative, but the interchangeability of IVC measurements in SC and TH is not fully established. Further, artificial intelligence (AI) with automated border detection may be of clinical value but it needs validation. Methods: Prospective observational study in mechanically ventilated patients with pressure-controlled mode. Primary outcome was the IVC distensibility (IVCd) in SC and TH imaging, with measurements taken both in M-Mode or with AI software. We calculated mean bias, limits of agreement (LoA), and intra-class correlation (ICC) coefficient. Results: Thirty-three patients were included. Feasibility rate was 87.9% and 81.8% for SC and TH visualization, respectively. Comparing imaging from the same anatomical site acquired with different modalities (M-Mode vs AI), we found the following IVCd differences: 1)SC: mean bias -3.1%, LoA [-20.1;13.9], ICC=0.65; 2)TH: mean bias -2.0%, LoA [-19.3;15.4], ICC=0.65. When comparing the results obtained from the same modality but from different sites (SC vs TH), IVCd differences were: 3)M-Mode: mean bias 1.1%, LoA [-6.9;9.1], ICC=0.54; 4)AI: mean bias 2.0%, LoA [-25.7;29.7], ICC=0.32. Conclusions: In patients mechanically ventilated, AI software shows good accuracy (modest overestimation) and moderate correlation as compared to M-mode assessment of IVCd, both for SC and TH windows. However, precision seems suboptimal with wide LoA. The comparison of M-Mode or AI between different sites yields similar results but with weaker correlation.

Section: Comments On Results From Different Acquisition Modalitymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Inferior vena cava distensibility from subcostal and trans-hepatic imaging using both M- mode or artificial intelligence: a prospective study on mechanically ventilated patients. Short Title: Subcostal vs transhepatic IVC evaluation

Sanfilippo¹,

Via²,

et al. 2022

Preprint

“…The use of AI has been applied to the whole echocardiography setting (i.e. left ventricular systolic 23,24 and diastolic [25][26][27] function, to right ventricular function 28 , assessment of heart valve diseases 29 , diagnosis of congenital heart diseases 30 ) and also to predict of FR, with encouraging results. For instance, Bataille et al 31 showed that machine learning models predicted FR with comparable accuracy to the hemodynamic response to passive leg raising, and evaluation of the IVC was among the key variables identified by the model, together with other Doppler derived parameters.…”

Section: Discussionmentioning

confidence: 99%

“…Among these, also echocardiography is experiencing a significant expansion of AI applications that might help daily practice. Indeed, AI has been used for the assessment of left ventricular systolic 23,24 and diastolic [25][26][27] function, right ventricular function 28 , but also for the evaluation of heart valve 29 and congenital heart diseases 30 . Moreover, machine learning has been developed for predicting FR at patient's bedside 31 with preliminary data on the implementation of AI for IVC assessment 32 .…”

Section: Introductionmentioning

confidence: 99%

Inferior vena cava distensibility from subcostal and trans-hepatic imaging using both M-mode or artificial intelligence: a prospective study on mechanically ventilated patients.

Via

Sanfilippo

et al. 2023

Preprint

Background: Variation of inferior vena cava (IVC) is used to predict fluid-responsiveness, but the IVC visualization with standard sagittal approach (SC, subcostal) cannot be always achieved. In such cases, coronal trans-hepatic (TH) window may offer an alternative, but the interchangeability of IVC measurements in SC and TH is not fully established. Further, artificial intelligence (AI) with automated border detection may be of clinical value but it needs validation. Methods: Prospective observational validation study in mechanically ventilated patients with pressure-controlled mode. Primary outcome was the IVC distensibility (IVC-DI) in SC and TH imaging, with measurements taken both in M-Mode or with AI software. We calculated mean bias, limits of agreement (LoA), and intra-class correlation (ICC) coefficient. Results: Thirty-three patients were included. Feasibility rate was 87.9% and 81.8% for SC and TH visualization, respectively. Comparing imaging from the same anatomical site acquired with different modalities (M-Mode vs AI), we found the following IVC-DI differences: 1)SC: mean bias -3.1%, LoA [-20.1;13.9], ICC=0.65; 2)TH: mean bias -2.0%, LoA [-19.3;15.4], ICC=0.65. When comparing the results obtained from the same modality but from different sites (SC vs TH), IVC-DI differences were: 3)M-Mode: mean bias 1.1%, LoA [-6.9;9.1], ICC=0.54; 4)AI: mean bias 2.0%, LoA [-25.7;29.7], ICC=0.32. Conclusions: In patients mechanically ventilated, AI software shows good accuracy (modest overestimation) and moderate correlation as compared to M-mode assessment of IVC-DI, both for SC and TH windows. However, precision seems suboptimal with wide LoA. The comparison of M-Mode or AI between different sites yields similar results but with weaker correlation.

“…Among these, also echocardiography is experiencing a signi cant implementation of AI functions that may aid and/or simplify clinicians' work. For instance, AI has been applied to the estimation of left ventricular systolic [24,25] and diastolic[26-28] function, to right ventricular function [29], but it has also been adopted for the assessment of heart valve diseases [30,31] and for diagnosis of congenital heart diseases [32]. Further, machine learning methods have been developed for the improvement of bedside prediction of FR [33], and preliminary experiences with AI in the assessment of IVCc have been reported [34].…”

Section: Introductionmentioning

confidence: 99%

Assessment of the inferior vena cava collapsibility from subcostal and trans-hepatic imaging using both M-mode or artificial intelligence: a prospective study on healthy volunteers.

Sanfilippo

Via

et al. 2022

Preprint

Background Assessment of the inferior vena cava (IVC) respiratory variation may be clinically useful but imaging from subcostal (SC, sagittal) region is not always feasible. It is unclear if coronal trans-hepatic (TH) IVC imaging provides interchangeable results as compared to SC imaging. In this context, the use of artificial intelligence (AI) with software for automated border detection may be clinically helpful but needs validation. Methods We conducted a prospective observational study in healthy volunteers assessing the IVC collapsibility (IVCc) in both SC and TH imaging; IVCc measures were taken in M-mode and with the AI method. We calculated the mean bias and the limits of agreement (LoA), as well as the intra-class correlation (ICC) coefficient with 95% confidence intervals. Results Sixty volunteers were included; IVC was not visualized in five of them (n = 2, both for SC and TH windows, 3.3%; n = 3 IVC not seen in TH approach, 5%). Compared with standard M-mode, the AI assessment showed good accuracy both for SC (IVCc: bias − 0.7%, LoA [-24.9;23.6]) and TH approach (IVCc: bias 3.7%, LoA [-14.9;22.3]); moreover, the ICC coefficients showed moderate reliability: 0.57 [0.36;0.73] in SC, and 0.72 [0.55;0.83] in TH. Comparing results obtained at different anatomical sites (SC vs TH), the results produced by M-mode assessment were not interchangeable (IVCc: bias 13.9%, LoA [-18.1;45.8]). When this evaluation was performed with the AI software, such difference became smaller: IVCc bias 7.7%, LoA [-19.2;34.6]. The correlation between SC and TH assessments was poor for M-mode (ICC = 0.08 [-0.18;0.34]), whilst moderate for AI (ICC = 0.69 [0.52;0.81]). Conclusions The AI software shows good accuracy when compared with the traditional M-mode IVCc assessment conducted at the same anatomical site (both at SC or TH level). There are clinically relevant differences when comparing results obtained at different sites (SC vs TH). Although AI may reduce differences between sagittal and coronal IVCc measurements, the results from different sites do not seem clinically interchangeable.