Interaction between clinicians and artificial intelligence to detect fetal atrioventricular septal defects on ultrasound: how can we optimize collaborative performance?

OBJECTIVES: Artificial intelligence (AI) has shown promise in improving the performance of fetal ultrasound screening in detecting congenital heart disease (CHD). The effect of giving AI advice to human operators has not been studied in this context. Giving additional information about AI model workings, such as confidence scores for AI predictions, may be a way of further improving performance. Our aims were to investigate whether AI advice improved overall diagnostic accuracy (using a single CHD lesion as an exemplar), and to determine what, if any, additional information given to clinicians optimized the overall performance of the clinician-AI team. METHODS: An AI model was trained to classify a single fetal CHD lesion (atrioventricular septal defect (AVSD)), using a retrospective cohort of 121 130 cardiac four-chamber images extracted from 173 ultrasound scan videos (98 with normal hearts, 75 with AVSD); a ResNet50 model architecture was used. Temperature scaling of model prediction probability was performed on a validation set, and gradient-weighted class activation maps (grad-CAMs) produced. Ten clinicians (two consultant fetal cardiologists, three trainees in pediatric cardiology and five fetal cardiac sonographers) were recruited from a center of fetal cardiology to participate. Each participant was shown 2000 fetal four-chamber images in a random order (1000 normal and 1000 AVSD). The dataset comprised 500 images, each shown in four conditions: (1) image alone without AI output; (2) image with binary AI classification; (3) image with AI model confidence; and (4) image with grad-CAM image overlays. The clinicians were asked to classify each image as normal or AVSD. RESULTS: A total of 20 000 image classifications were recorded from 10 clinicians. The AI model alone achieved an accuracy of 0.798 (95% CI, 0.760-0.832), a sensitivity of 0.868 (95% CI, 0.834-0.902) and a specificity of 0.728 (95% CI, 0.702-0.754), and the clinicians without AI achieved an accuracy of 0.844 (95% CI, 0.834-0.854), a sensitivity of 0.827 (95% CI, 0.795-0.858) and a specificity of 0.861 (95% CI, 0.828-0.895). Showing a binary (normal or AVSD) AI model output resulted in significant improvement in accuracy to 0.865 (P < 0.001). This effect was seen in both experienced and less-experienced participants. Giving incorrect AI advice resulted in a significant deterioration in overall accuracy, from 0.761 to 0.693 (P < 0.001), which was driven by an increase in both Type-I and Type-II errors by the clinicians. This effect was worsened by showing model confidence (accuracy, 0.649; P < 0.001) or grad-CAM (accuracy, 0.644; P < 0.001). CONCLUSIONS: AI has the potential to improve performance when used in collaboration with clinicians, even if the model performance does not reach expert level. Giving additional information about model workings such as model confidence and class activation map image overlays did not improve overall performance, and actually worsened performance for images for which the AI model was incorrect. © 2024 The Authors. Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.

Synergy in the heart: RV systolic function plays a key role in optimizing LV performance during exercise.

The right ventricle (RV) is often overlooked in the evaluation of cardiac performance and treatment of left ventricular (LV) heart diseases. However, recent evidence suggests the RV may play an important role in maintaining systemic cardiac function and delivering stroke volume (SV). We used exercise cardiac magnetic resonance and biomechanical modeling to investigate the role of the RV in LV stroke volume regulation. We studied SV augmentation during exercise by pharmacologically inducing negative chronotropy (sHRi) in healthy volunteers and investigating training-induced SV augmentation in endurance athletes. SV augmentation during exercise after sHRi is achieved differently in the two ventricles. In the RV, the larger SV is driven by increasing contraction down to lower end-systolic volume (ESV; P < 0.001). In the LV, SV augmentation is achieved through an increase in end-diastolic volume (EDV; P < 0.001), avoiding contraction to a lower ESV. The same mechanism underlies the enhanced SV response observed in athletes. Changes in atrial area during SV augmentation suggest that the improved LV EDV response is sustained by the larger RV contractions. Using our biomechanical model, we explain this behavior by showing that the RV systolic function-driven regulation of LV SV optimizes the energetic cost of LV contraction and leads to minimization of the total costs of biventricular contraction. In conclusion, this work provides mechanistic understanding of the pivotal role of the RV in optimizing LV SV during exercise. It demonstrates why optimizing RV function needs to become a key part of therapeutic strategies in patients and training for athletes.NEW & NOTEWORTHY The right ventricle appears to have an important impact on maintaining systemic cardiac function and delivering stroke volume. However, its exact role in supporting left ventricular function has so far been unclear. This study demonstrates a new mechanism of ventricular interaction that provides mechanistic understanding of the key importance of the right ventricle in driving cardiac performance.

An observational study to compare the utilisation of computed tomography colonography with optical colonoscopy as the first diagnostic imaging tool in patients with suspected colorectal cancer.

AIM: To evaluate the clinical and cost implications of using computed tomography colonography (CTC) compared to optical colonoscopy (OC) as the initial colonic investigation in patients with low-to-intermediate risk of colorectal cancer (CRC). MATERIALS AND METHODS: A non-randomised, prospective single-centre study recruited 180 participants to compare the cost implications of two clinical pathways used in the diagnosis of low-to-intermediate risk of CRC that differ in the initial diagnostic test, either CTC or OC. Costs were compared using generalised linear models (GLM) and combined with quality-adjusted life years (QALYs, based on the EQ-5D-5L) to estimate cost-effectiveness at 6 months post-recruitment. Secondary outcomes assessed access to care and patient satisfaction. RESULTS: Mean (SD, n) cost at 6 months post-recruitment per participant was £991 (£316, n=105) for the OC group and £645 (£607, n=68) for the CTC group, leading to an estimated cost difference of -£370 (95% CI: -£554, -£185, p<0.001). Assuming a £20,000 willingness-to-pay per QALY threshold, there was a 91.4% probability of CTC being cost-effective at month 6. The utilisation of CTC led to improved access to care, with a shorter mean time from referral from primary care to results (6.3 days difference, p=0.005). No differences in patient satisfaction were detected between both groups. CONCLUSION: The utilisation of CTC as the first-line investigation for patients with low-to-intermediate risk of CRC has the potential to release OC capacity, of pivotal importance for patients more likely to benefit from an invasive diagnostic approach.

SiRNA/DOX lodeded chitosan based nanoparticles: Development, Characterization and in vitro evaluation on A549 lung cancer cell line.

High-mobility group AT-hook2 (HMGA2), involved in epithelial mesenchymal transition (EMT) process, has a pivotal role in lung cancer metastasis. Lung cancer therapy with HMGA2 suppressing small interfering RNA (siRNA) has been introduced recently while doxorubicin (DOX) has been used as a frequent cancer chemotherapy agent. Both reagents have been faced with obstacles in clinic which make them ineffective. NanoParticles (NPs) provided a platform for efficient co delivery of the anticancer drugs. The aim of this study was production and in vitro characterization of different pharmacological groups (siRNA, DOX or siRNA-DOX) of carboxymethyl dextran thrimethyl chitosan nanoparticles (CMDTMChiNPs) on cytotoxicity, gene expression, apoptosis and migration of metastatic lung cancer cell line (A-549). CMDTMChiNPs were synthesized and encapsulated with siRNA, DOX or siRNA-DOX. Then the effects of HMGA2 siRNA and DOX co delivery was assessed in A549 viability and target genes (HMGA2, Ecadherin, vimentin and MMP9) by MTT and real time PCR, respectively. In addition capability of apoptosis induction and anti-migratory features of formulated NPs were analyzed by flowcytometry and wound healing assays. SiRNA-DOX-CMDTM ChiNPs approximate size were 207±5 with poly dispersity index (PDI) and zeta potential of 0.4 and 16.3±0.3, respectively. NPs loaded with DOX and siRNA were the most efficient drug formulations in A549 cell cytotoxicity, altering of EMT markers, apoptosis induction and migration inhibition. Generally our results showed that co delivery of HMGA2 siRNA and DOX by novel designed CMDTMChiNPs is a new therapeutic approach with great potential efficiency for lung cancer treatment.

Multidetector computed tomography sizing of bioprosthetic valves: guidelines for measurement and implications for valve-in-valve therapies.

AIM: To describe a technique for bioprosthetic multidetector computed tomography (MDCT) sizing and to compare MDCT-derived values against manufacturer-provided sizing. MATERIALS AND METHODS: Fourteen bioprosthetic stented valves commonly used in the aortic valve position were evaluated using a Philips 256 MDCT system. All valves were scanned using a dedicated cardiac CT protocol with a four-channel electrocardiography (ECG) simulator. Measurements were made of major and minor axes and the area and perimeter of the internal stent using varying reconstruction kernels and window settings. Measurements derived from MDCT (MDCT ID) were compared against the stent internal diameter (Stent ID) as provided by the valve manufacturer and the True ID (Stent ID + insertion of leaflets). All data were collected and analysed using SPSS for Mac (version 21). RESULTS: The mean difference between the MDCT ID and Stent ID was 0.6±1.9 mm (r=0.649, p=0.012) and between MDCT ID and True ID 2.1±2 mm (r=0.71, p=0.005). There was no difference in the major (p=0.90), minor (p=0.87), area (p=0.92), or perimeter (p=0.92) measurements when sharp, standard, and detailed stent kernels were used. Similarly, the measurements remained consistent across differing windowing levels. CONCLUSION: Bioprosthetic stented valves may be reliably sized using MDCT in patients requiring valve-in-valve (VIV) interventions where the valve type and size are unknown. In these cases, clinicians should be aware that MDCT has a tendency to overestimate the True ID size.

Detection of intracoronary thrombus by magnetic resonance imaging in patients with acute myocardial infarction.

BACKGROUND: Persistent intracoronary thrombus after plaque rupture is associated with an increased risk of subsequent myocardial infarction and mortality. Coronary thrombus is usually visualized invasively by x-ray coronary angiography. Non-contrast-enhanced T1-weighted magnetic resonance (MR) imaging has been useful for direct imaging of carotid thrombus and intraplaque hemorrhage by taking advantage of the short T1 of methemoglobin present in acute thrombus and intraplaque hemorrhage. The aim of this study was to investigate the use of non-contrast-enhanced MR for direct thrombus imaging (MRDTI) in patients with acute myocardial infarction. METHODS AND RESULTS: Eighteen patients (14 men; age, 61±9 years) underwent MRDTI within 24 to 72 hours of presenting with an acute coronary syndrome before invasive x-ray coronary angiography; MRDTI was performed with a T1-weighted, 3-dimensional, inversion-recovery black-blood gradient-echo sequence without contrast administration. Ten patients were found to have intracoronary thrombus on x-ray coronary angiography (left anterior descending, 4; left circumflex, 2; right coronary artery, 4; and right coronary artery-posterior descending artery, 1), and 8 had no visible thrombus. We found that MRDTI correctly identified thrombus in 9 of 10 patients (sensitivity, 91%; posterior descending artery thrombus not detected) and correctly classified the control group in 7 of 8 patients without thrombus formation (specificity, 88%). The contrast-to-noise ratio was significantly greater in coronary segments containing thrombus (n=10) compared with those without visible thrombus (n=131; mean contrast-to-noise ratio, 15.9 versus 2.6; P<0.001). CONCLUSION: Use of MRDTI allows selective visualization of coronary thrombus in a patient population with a high probability of intracoronary thrombosis.

Biophysical modeling to simulate the response to multisite left ventricular stimulation using a quadripolar pacing lead.

BACKGROUND: Response to cardiac resynchronization therapy (CRT) is reduced in patients with posterolateral scar. Multipolar pacing leads offer the ability to select desirable pacing sites and/or stimulate from multiple pacing sites concurrently using a single lead position. Despite this potential, the clinical evaluation and identification of metrics for optimization of multisite CRT (MCRT) has not been performed. METHODS: The efficacy of MCRT via a quadripolar lead with two left ventricular (LV) pacing sites in conjunction with right ventricular pacing was compared with single-site LV pacing using a coupled electromechanical biophysical model of the human heart with no, mild, or severe scar in the LV posterolateral wall. RESULT: The maximum dP/dt(max) improvement from baseline was 21%, 23%, and 21% for standard CRT versus 22%, 24%, and 25% for MCRT for no, mild, and severe scar, respectively. In the presence of severe scar, there was an incremental benefit of multisite versus standard CRT (25% vs 21%, 19% relative improvement in response). Minimizing total activation time (analogous to QRS duration) or minimizing the activation time of short-axis slices of the heart did not correlate with CRT response. The peak electrical activation wave area in the LV corresponded with CRT response with an R(2) value between 0.42 and 0.75. CONCLUSION: Biophysical modeling predicts that in the presence of posterolateral scar MCRT offers an improved response over conventional CRT. Maximizing the activation wave area in the LV had the most consistent correlation with CRT response, independent of pacing protocol, scar size, or lead location.

Wideband RCS reduction due to plasma generated by radioactive nuclei for cylindrical object.

Radar cross section reduction has been one of the most important research topics in recent years. Plasma-based stealth is a method of reducing the radar cross section, which dampens the electromagnetic waves and reduces the amount of return waves. In this paper, a coating of the radioactive nucleus [Formula: see text] on the surface of the cylinder with a radius of 10 cm is considered and the range of the emitted alpha particles and the electron density generated in the air are obtained using the Geant4 code under standard temperature and pressure conditions. By finite element method solution, the radar cross section of the conductive cylindrical object has been simulated and extracted in the presence and absence of plasma created by alpha-particles. The obtained results show a reduction of 5-8 dB [Formula: see text] in the radar cross section in the frequency range of 2-12 GHz for specific activity source of 1 Ci/[Formula: see text].

Toward true 3D visualization of active catheters using compressed sensing.

A crucial requirement in MR-guided interventions is the visualization of catheter devices in real time. However, true 3D visualization of the full length of catheters has hitherto been impossible given scan time constraints. Compressed sensing (CS) has recently been proposed as a method to accelerate MR imaging of sparse objects. Images acquired with active interventional devices exhibit a high CNR and are inherently sparse, therefore rendering CS ideally suited for accelerating data acquisition. A framework for true visualization of active catheters in 3D is proposed employing CS to gain high undersampling factors making real-time applications feasible. Constraints are introduced taking into account prior knowledge of catheter geometry and catheter motion over time to improve and accelerate image reconstruction. The potential of the method is demonstrated using computer simulations and phantom experiments and in vivo feasibility is demonstrated in a pig experiment.

Whole-heart imaging using undersampled radial phase encoding (RPE) and iterative sensitivity encoding (SENSE) reconstruction.

Whole-heart isotropic nonangulated cardiac magnetic resonance (CMR) is becoming an important protocol in simplifying MRI, since it reduces the need of cumbersome planning of angulations. However the acquisition times of whole-heart MRI are prohibitive due to the large fields of view (FOVs) and the high spatial resolution required for depicting small structures and vessels. To address this problem, we propose a three-dimensional (3D) acquisition scheme that combines Cartesian sampling in the readout direction with an undersampled radial scheme in the phase-encoding plane. Different undersampling patterns were investigated in combination with an iterative sensitivity encoding (SENSE) reconstruction and a 32-channel cardiac coil. Noise amplification maps were calculated to compare the performance of the different patterns using iterative SENSE reconstruction. The radial phase-encoding (RPE) scheme was implemented on a clinical MR scanner and tested on phantoms and healthy volunteers. The proposed method exhibits better image quality even for high acceleration factors (up to 12) in comparison to Cartesian acquisitions.

Cardiac function estimation from MRI using a heart model and data assimilation: advances and difficulties.

In this paper, we present a framework to estimate local ventricular myocardium contractility using clinical MRI, a heart model and data assimilation. First, we build a generic anatomical model of the ventricles including muscle fibre orientations and anatomical subdivisions. Then, this model is deformed to fit a clinical MRI, using a semi-automatic fuzzy segmentation, an affine registration method and a local deformable biomechanical model. An electromechanical model of the heart is then presented and simulated. Finally, a data assimilation procedure is described, and applied to this model. Data assimilation makes it possible to estimate local contractility from given displacements. Presented results on fitting to patient-specific anatomy and assimilation with simulated data are very promising. Current work on model calibration and estimation of patient parameters opens up possibilities to apply this framework in a clinical environment.

The expression of inducible cAMP early repressor (ICER) is altered in prostate cancer cells and reverses the transformed phenotype of the LNCaP prostate tumor cell line.

Inducible cAMP early repressor (ICER) has been shown to be an important mediator of cAMP antiproliferative activity. In this report, it was found that cAMP retards LNCaP cell growth; in contrast, cAMP inhibits the growth of PC-3 and DU-145 cells. ICER protein levels were markedly reduced in prostate cancer epithelial cells and undetectable and uninducible by cAMP in LNCaP and DU 145 cells. Forced expression of ICER in LNCaP cells caused inhibition of cell growth and thymidine incorporation and halted cells at the G(1) phase of the cell cycle. These ICER-bearing LNCaP cells were rendered unable to grow in soft agar and unable to form tumors in nude mice. These results suggest that deregulation of ICER expression may be related to carcinogenesis of the prostate gland.

A novel methodology for personalized simulations of ventricular hemodynamics from noninvasive imaging data.

Current state-of-the-art imaging techniques can provide quantitative information to characterize ventricular function within the limits of the spatiotemporal resolution achievable in a realistic acquisition time. These imaging data can be used to personalize computer models, which in turn can help treatment planning by quantifying biomarkers that cannot be directly imaged, such as flow energy, shear stress and pressure gradients. To date, computer models have typically relied on invasive pressure measurements to be made patient-specific. When these data are not available, the scope and validity of the models are limited. To address this problem, we propose a new methodology for modeling patient-specific hemodynamics based exclusively on noninvasive velocity and anatomical data from 3D+t echocardiography or Magnetic Resonance Imaging (MRI). Numerical simulations of the cardiac cycle are driven by the image-derived velocities prescribed at the model boundaries using a penalty method that recovers a physical solution by minimizing the energy imparted to the system. This numerical approach circumvents the mathematical challenges due to the poor conditioning that arises from the imposition of boundary conditions on velocity only. We demonstrate that through this technique we are able to reconstruct given flow fields using Dirichlet only conditions. We also perform a sensitivity analysis to investigate the accuracy of this approach for different images with varying spatiotemporal resolution. Finally, we examine the influence of noise on the computed result, showing robustness to unbiased noise with an average error in the simulated velocity approximately 7% for a typical voxel size of 2mm(3) and temporal resolution of 30ms. The methodology is eventually applied to a patient case to highlight the potential for a direct clinical translation.

ICER-IIgamma is a tumor suppressor that mediates the antiproliferative activity of cAMP.

The second messenger cAMP inhibits the proliferation of most cell types. The nuclear response of cAMP is mediated by transcription factors like the cAMP-Responsive Element Modulator (CREM) gene. One of the products of the CREM gene, the transcriptional repressor Inducible cAMP Early Repressor-IIgamma (ICER-IIgamma), is induced by cAMP. ICER-IIgamma blocks cells at the G2/M boundary of the cell cycle. Here we show that ICER-IIgamma dramatically inhibits the growth and DNA synthesis of mouse pituitary tumor cells and human choriocarcinoma cells. This alteration in cell growth is coupled with reduced ability of these cells to grow in an anchorage-independent manner and to form tumors in mice. These data demonstrate that ICER-IIgamma is a tumor suppressor gene product mediating the antiproliferative activity of cAMP.

Simulation of cardiac pathologies using an electromechanical biventricular model and XMR interventional imaging.

Simulating cardiac electromechanical activity is of great interest for a better understanding of pathologies and for therapy planning. Design and validation of such models is difficult due to the lack of clinical data. XMR systems are a new type of interventional facility in which patients can be rapidly transferred between X-ray and MR systems. Our goal is to design and validate an electromechanical model of the myocardium using XMR imaging. The proposed model is computationally fast and uses clinically observable parameters. We present the integration of anatomy, electrophysiology, and motion from patient data. Pathologies are introduced in the model and simulations are compared to measured data. Initial qualitative comparison on the two clinical cases presented is encouraging. Once fully validated, these models will make it possible to simulate different interventional strategies.

Radiation dose perturbation at tissue-titanium dental interfaces in head and neck cancer patients.

PURPOSE: To determine the dose perturbation effects at the tissue-metal implant interfaces in head and neck cancer patients treated with 6 MV and 10 MV photon beams. METHODS AND MATERIALS: Phantom measurements were performed to investigate the magnitude of dose perturbation to the tissue adjacent to the titanium alloy implants with (100 mu and 500 mu thick) and without hydroxylapatite (HA) coating. Radiographic and radiochromic films were placed at the upper (and lower) surface of circular metal discs (diameter x thickness: 15 x 3.2, 48 x 3.2, 48 x 3.8 mm2) in a solid water phantom and were exposed perpendicular to radiation beams. The dosimeters were scanned with automatic film scanners. Using a thin-window parallel-plate ion chamber, dose perturbation were measured for a 48 x 3.2 mm2 disc. RESULTS: At the upper surface of the tissue-dental implant interface, the radiographic data indicate that for 15 x 3.2 mm2 uncoated, as well as 100 mu coated discs, dose perturbation is about +22.5% and +20.0% using 6 MV and 10 MV photon beams, respectively. For 48 x 3.2 mm2 discs, these values basically remain the same. However, for 48 x 3.8 mm2 discs, these values increase slightly to about +23.0% and +20.5% for 6 MV and 10 MV beams, respectively. For 48 x 3.2 mm2 discs with 500 mu coating, dose enhancement is slightly lower than that obtained for uncoated and 100 mu coated discs for each beam energy studied. At the lower interface for 15 x 3.2 mm2 and 48 x 3.2 mm2 uncoated and 100 mu coated discs, dose reduction is similar and is about -13.5% and -9.5% for 6 MV and 10 MV beams, respectively. For 48 x 3.8 mm2 discs, dose reduction is about -14.5% and -10.0% for 6 MV and 10 MV beams, respectively. For 48 x 3.2 mm2 discs with 500 mu coating, the dose reduction were slightly higher than those for uncoated and 100 mu coated discs. CONCLUSIONS: For the beam energies studied, dose enhancement is slightly larger for the lower energy beam. The results of dose perturbation were similar for 100 mu coated and uncoated discs. These results were slightly lower for the 500 mu coated discs but are not clinically significant. The dosimetry results obtained from radiochromic films were similar to the ones obtained from radiographic film. The dose enhancement results obtained from ion chamber dosimetry are higher than those obtained from film dosimetry. The ion chamber data represent the data at "true" tissue-titanium interface, whereas the ones obtained from film dosimetry represent the data at film-titanium interface.

Hemodynamic response to continuous infusion of dobutamine in Alagille's syndrome.

BACKGROUND: Alagille's syndrome is a rare condition that is characterized by paucity of interlobular bile ducts and peripheral pulmonary artery stenosis. Liver transplantation in the setting of peripheral pulmonary stenosis and right ventricular hypertension seems to be associated with a higher mortality, which raises the concern that these patients are unable to increase their cardiac output in the immediate posttransplantation period to meet the demands of reperfusion and early graft dysfunction and cope with further increases in pulmonary vascular resistance. METHOD: Cardiac catheterization was performed in 15 children with Alagille's syndrome and peripheral pulmonary artery stenosis to measure the cardiac output response to dobutamine infusion. The cardiac output was measured before and during each increment of infusion of dobutamine at 10 microg/kg/min and 20 microg/kg/min by using a thermodilution catheter placed in the pulmonary artery. RESULTS: There was a significant change in the baseline cardiac index (P<0.001) with an infusion of 20 microg/kg/min of dobutamine with an increase from 4.4(1.0) L/min/m2 to 6.4(1.5) L/min/m2. There was, however, a wide variation between individuals in the increase in cardiac index, ranging from 7.3-95%. There was no correlation between the baseline systolic right ventricular/aortic pressure ratios and the increase in cardiac index (r=0.1086). CONCLUSION: The increase in cardiac index, in response to dobutamine in patients with Alagille's syndrome, is independent of the more conventional measurement of the right ventricular pressure. This method of producing a hemodynamic response is closer to the response that results after liver transplantation, and thus, it may be a better way of predicting the outcome of liver transplantation in these patients.