Tootooni, M. Samie

Background: Adequate medication dosing in the intensive care unit (ICU) remains challenging due to rapid physiological variability, narrow therapeutic windows, and evolving standards of care. Artificial intelligence (AI) offers promise for personalized dosing, but its accuracy and clinical utility remain uncertain. Methods: We conducted a systematic scoping review (2011-2025) of PubMed, Embase, and Cochrane, following the Joanna Briggs Institute (JBI) and preferred reporting Items for systematic reviews and meta-analyses extension for scoping reviews (PRISMA-ScR) guidelines (open science framework [OSF]-registered). Eligible studies applied AI methods ( e.g. , machine learning [ML], deep learning [DL], reinforcement learning [RL], and natural language processing [NLP]) to ICU medication dosing or monitoring. Results: Data extraction included study design, drug class, model inputs and outputs, algorithms, validation approach, and reported outcomes. Forty-seven studies met the inclusion criteria, which were mostly retrospective and focused on adult ICU populations over the past five years. Antibiotics and anticoagulants were the most studied classes, with vancomycin being the most common. Models primarily used labs, vitals, and demographics, with limited incorporation of time series, pharmacokinetic data, or clinical notes. Outputs spanned dose prediction, serum concentration, risk stratification, adverse events, and decision support. Algorithms evolved from classical ML and ensembles to RL and long short-term memory (LSTM)-based models, although a few advanced beyond retrospective validation. Discussion: AI systems showed strong predictive performance and some clinical benefits ( e.g. , heparin target attainment, sedative reduction, and adverse event detection), but the field remains fragmented and clinically immature. Newer architectures and systematic explainability are lacking, and generalizability is uncertain. AI-enabled dosing holds promise for precision pharmacotherapy in critical illness but requires prospective multicenter validation, inclusion of population pharmacokinetic (PK) parameters, diverse cohorts, explainable design, and clinical integration.

Abstract Background We aimed to determine if patient symptoms and computed tomography enterography (CTE) and magnetic resonance enterography (MRE) imaging findings can be used to predict near-term risk of surgery in patients with small bowel Crohn’s disease (CD). Methods CD patients with small bowel strictures undergoing serial CTE or MRE were retrospectively identified. Strictures were defined by luminal narrowing, bowel wall thickening, and unequivocal proximal small bowel dilation. Harvey-Bradshaw index (HBI) was recorded. Stricture observations and measurements were performed on baseline CTE or MRE and compared to with prior and subsequent scans. Patients were divided into those who underwent surgery within 2 years and those who did not. LASSO (least absolute shrinkage and selection operator) regression models were trained and validated using 5-fold cross-validation. Results Eighty-five patients (43.7 ± 15.3 years of age at baseline scan, majority male [57.6%]) had 137 small bowel strictures. Surgery was performed in 26 patients within 2 years from baseline CTE or MRE. In univariate analysis of patients with prior exams, development of stricture on the baseline exam was associated with near-term surgery (P = .006). A mathematical model using baseline features predicting surgery within 2 years included an HBI of 5 to 7 (odds ratio [OR], 1.7 × 105; P = .057), an HBI of 8 to 16 (OR, 3.1 × 105; P = .054), anastomotic stricture (OR, 0.002; P = .091), bowel wall thickness (OR, 4.7; P = .064), penetrating behavior (OR, 3.1 × 103; P = .096), and newly developed stricture (OR: 7.2 × 107; P = .062). This model demonstrated sensitivity of 67% and specificity of 73% (area under the curve, 0.62). Conclusions CTE or MRE imaging findings in combination with HBI can potentially predict which patients will require surgery within 2 years.

The goal of this work is in situ monitoring of the functional properties of aerosol jet-printed electronic devices. In pursuit of this goal, the objective is to develop a multiple-input, single-output (MISO) machine learning model to estimate the device functional properties in a near real-time fashion as a function of process parameters as well as 2D/3D features of line morphology. The aim is to use the MISO model for in situ estimation and thus, monitoring of line/device resistance in aerosol jet printing (AJP) process. To realize this objective, silver nanoparticle structures are printed by varying three process parameters: (i) sheath gas flow rate (ShGFR), (ii) exhaust gas flow rate (EGFR), and (iii) print speed (PS). Subsequently, line morphology is captured in situ using a high-resolution charge-coupled device (CCD) camera, mounted coaxial to the nozzle. Besides, utilizing 2D/3D quantifiers (introduced in the authors’ previous publications), the line morphology is further quantified, and the extracted features (e.g., line width, overspray, cross-sectional area, etc.) are fed as inputs to a novel sparse representation-based classification (SRC) model. The four-point probe method is used for measurement of resistance, and definition of a priori classification labels. The outcome of this research paves the way for future control of device functional properties in AJP process.

AbstractBackgroundNephrotoxins contribute to 20%–40% of acute kidney injury (AKI) cases in the intensive care unit (ICU). The combination of piperacillin-tazobactam (PTZ) and vancomycin (VAN) has been identified as nephrotoxic, but existing studies focus on extended durations of therapy rather than the brief empiric courses often used in the ICU. The current study was performed to compare the risk of AKI with a short course of PTZ/VAN to with the risk associated with other antipseudomonal β-lactam/VAN combinations.MethodsThe study included a retrospective cohort of 3299 ICU patients who received ≥24 but ≤72 hours of an antipseudomonal β-lactam/VAN combination: PTZ/VAN, cefepime (CEF)/VAN, or meropenem (MER)/VAN. The risk of developing stage 2 or 3 AKI was compared between antibiotic groups with multivariable logistic regression adjusted for relevant confounders. We also compared the risk of persistent kidney dysfunction, dialysis dependence, or death at 60 days between groups.ResultsThe overall incidence of stage 2 or 3 AKI was 9%. Brief exposure to PTZ/VAN did not confer a greater risk of stage 2 or 3 AKI after adjustment for relevant confounders (adjusted odds ratio [95% confidence interval] for PTZ/VAN vs CEF/VAN, 1.11 [.85–1.45]; PTZ/VAN vs MER/VAN, 1.04 [.71–1.42]). No significant differences were noted between groups at 60-day follow-up in the outcomes of persistent kidney dysfunction (P = .08), new dialysis dependence (P = .15), or death (P = .09).ConclusionShort courses of PTZ/VAN were not associated with a greater risk of short- or 60-day adverse renal outcomes than other empiric broad-spectrum combinations.