Browsing by Author "Basmaji, John"

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    Doppler identified venous congestion in septic shock: protocol for an international, multi-centre prospective cohort study (Andromeda-VEXUS)
    (2023) Prager, Ross; Argaiz, Eduardo; Pratte, Michael; Rola, Philippe; Arntfield, Robert; Beaubien-Souligny, William; Denault, Andre Y.; Haycock, Korbin; Aguiar, Francisco Miralles; Bakker, Jan; Ospina-Tascon, Gustavo; Orozco, Nicolas; Rochwerg, Bram; Lewis, Kimberley; Quazi, Ibrahim; Kattan, Eduardo; Hernandez, Glenn; Basmaji, John
    Introduction Venous congestion is a pathophysiological state where high venous pressures cause organ oedema and dysfunction. Venous congestion is associated with worse outcomes, particularly acute kidney injury (AKI), for critically ill patients. Venous congestion can be measured by Doppler ultrasound at the bedside through interrogation of the inferior vena cava (IVC), hepatic vein (HV), portal vein (PV) and intrarenal veins (IRV). The objective of this study is to quantify the association between Doppler identified venous congestion and the need for renal replacement therapy (RRT) or death for patients with septic shock.Methods and analysis This study is a prespecified substudy of the ANDROMEDA-SHOCK 2 (AS-2) randomised control trial (RCT) assessing haemodynamic resuscitation in septic shock and will enrol at least 350 patients across multiple sites. We will include adult patients within 4 hours of fulfilling septic shock definition according to Sepsis-3 consensus conference. Using Doppler ultrasound, physicians will interrogate the IVC, HV, PV and IRV 6-12 hours after randomisation. Study investigators will provide web-based educational sessions to ultrasound operators and adjudicate image acquisition and interpretation. The primary outcome will be RRT or death within 28 days of septic shock. We will assess the hazard of RRT or death as a function of venous congestion using a Cox proportional hazards model. Sub-distribution HRs will describe the hazard of RRT given the competing risk of death.Ethics and dissemination We obtained ethics approval for the AS-2 RCT, including this observational substudy, from local ethics boards at all participating sites. We will report the findings of this study through open-access publication, presentation at international conferences, a coordinated dissemination strategy by investigators through social media, and an open-access workshop series in multiple languages.Trial registration number NCT05057611.
  • Thumbnail Image
    Item
    Improving the Generalizability and Performance of an Ultrasound Deep Learning Model Using Limited Multicenter Data for Lung Sliding Artifact Identification
    (2024) Wu, Derek; Smith, Delaney; VanBerlo, Blake; Roshankar, Amir; Lee, Hoseok; Li, Brian; Ali, Faraz; Rahman, Marwan; Basmaji, John; Tschirhart, Jared; Ford, Alex; VanBerlo, Bennett; Durvasula, Ashritha; Vannelli, Claire; Dave, Chintan; Deglint, Jason; Ho, Jordan; Chaudhary, Rushil; Clausdorff Fiedler, Hans Jurgen; Prager, Ross; Millington, Scott; Shah, Samveg; Buchanan, Brian; Arntfield, Robert
    Deep learning (DL) models for medical image classification frequently struggle to generalize to data from outside institutions. Additional clinical data are also rarely collected to comprehensively assess and understand model performance amongst subgroups. Following the development of a single-center model to identify the lung sliding artifact on lung ultrasound (LUS), we pursued a validation strategy using external LUS data. As annotated LUS data are relatively scarce—compared to other medical imaging data—we adopted a novel technique to optimize the use of limited external data to improve model generalizability. Externally acquired LUS data from three tertiary care centers, totaling 641 clips from 238 patients, were used to assess the baseline generalizability of our lung sliding model. We then employed our novel Threshold-Aware Accumulative Fine-Tuning (TAAFT) method to fine-tune the baseline model and determine the minimum amount of data required to achieve predefined performance goals. A subgroup analysis was also performed and Grad-CAM++ explanations were examined. The final model was fine-tuned on one-third of the external dataset to achieve 0.917 sensitivity, 0.817 specificity, and 0.920 area under the receiver operator characteristic curve (AUC) on the external validation dataset, exceeding our predefined performance goals. Subgroup analyses identified LUS characteristics that most greatly challenged the model’s performance. Grad-CAM++ saliency maps highlighted clinically relevant regions on M-mode images. We report a multicenter study that exploits limited available external data to improve the generalizability and performance of our lung sliding model while identifying poorly performing subgroups to inform future iterative improvements. This approach may contribute to efficiencies for DL researchers working with smaller quantities of external validation data.