Alexis Nolin-Lapalme

Automated electrocardiogram (ECG) interpretation has advanced, yet most systems remain narrow classifiers that emit fixed labels rather than the narratives or endpoint-specific answers clinicians need. Generative approaches could instead produce rich narratives, but are constrained by the gap between continuous biosignals and discrete language tokens. Here we present DeepECG-Tok, which reframes ECG interpretation as a unified instruction-following problem. A residual vector-quantization tokenizer (QINCo) maps 12-lead waveforms to language-model-compatible tokens. Its frozen embeddings achieved a macro-averaged area under the receiver operating characteristic curve (AUROC) of 0.96 for 77-condition classification, outperforming supervised and self-supervised baselines. Aligned with a large language model, a single instruction-tuned model performed ECG interpretation, structured reporting and clinical endpoint prediction, including left ventricular ejection fraction, structural heart disease and atrial fibrillation risk, using 7.27 million question–answer pairs. Frozen-tokenizer diagnostic classification transferred without retraining to two external cohorts, retaining macro-averaged AUROCs of 0.88–0.90, and clinical-endpoint prediction transferred to two further cohorts (external LVEF ≤40% AUROC 0.74–0.76). Evaluated end to end using an ontology-grounded large language model as a judge, which achieved a mean agreement (Cohen’s κ) of 0.82 against two cardiologists, the unified instruction-tuned model scored 0.71 internally and 0.50–0.53 in the same external cohorts. In a blinded reader study, board-certified cardiologists and residents rated its free-text reports comparably to reference clinician reports, with a paired win–tie–loss distribution of 32:35:33 and a forced-choice preference of 0.53 among decided cases, with no significant difference between the model and reference reports in either comparison. These findings establish discrete ECG tokenization as a foundation for general-purpose models that generate clinically useful interpretations and answer diverse questions directly from cardiac waveforms.

Purpose of Review: Structural heart disease (SHD) encompasses diseases involving the heart valves, chambers, walls, and muscles. Current diagnostic methods have limited accessibility and predictive value. This review aims to present recent advances in artificial intelligence (AI)-guided tools in the screening of SHD and valvular heart disease (VHD), and to present challenges and opportunities for their use in clinical practice. Recent Findings: AI-guided models trained on ECGs, chest X-rays, and coronary artery calcium scans have a high accuracy in the diagnosis of SHD, heart failure, low left ventricular ejection fraction, and VHD. Some of these models can highlight the signals that influence their predictions, improving explainability. Summary: The use of AI in screening for SHD and VHD could lead to earlier diagnosis, enhanced accuracy, and better accessibility. However, outcome data on earlier diagnosis using these tools is required before broad deployment.

Background and Aims: The 12-lead electrocardiogram (ECG) remains a cornerstone of cardiac diagnostics, yet existing artificial intelligence (AI) solutions for automated interpretation often lack generalizability, remain closed source, and are primarily trained using supervised learning (SL), which requires extensive labelled datasets and may limit adaptability across diverse clinical settings. Self-supervised learning (SSL) can potentially overcome these limitations by learning robust representations from unlabelled data. To address these challenges, this study developed and compared two open-source foundational ECG models: DeepECG-SL, a supervised multilabel ECG model, and DeepECG-SSL, a self-supervised model. Methods: Both models were trained on over 1 million ECGs using a standardized preprocessing pipeline and automated free-text extraction from ECG reports to predict 77 cardiac conditions. DeepECG-SSL leveraged unlabelled data through self-supervised contrastive learning and masked lead modelling before fine-tuning for downstream tasks, while DeepECG-SL was trained directly on labelled diagnostic data in an end-to-end fashion. Performance was evaluated across seven private, multilingual healthcare systems and four public ECG repositories, with assessment of fairness by age and sex, and investigation of privacy vulnerabilities as well as memory and compute requirements. Results: DeepECG-SSL achieved micro-averaged area under the receiver operating characteristic curves (AUROCs) across all 77 cardiac conditions for ECG interpretation of 0.990 [95% confidence interval (CI): 0.990, 0.990] on the internal dataset (MHI-ds), 0.981 (95% CI: 0.981, 0.981) on external public datasets (UKB, CLSA, MIMIC-IV and PTB), and 0.983 (95% CI: 0.983, 0.983) on external private datasets (UW, UCSF, JGH, NYP, MGH, CSH and CHUM), while DeepECG-SL demonstrated AUROCs of 0.992 (95% CI: 0.992, 0.992), 0.980 (95% CI: 0.980, 0.980), and 0.983 (95% CI: 0.983, 0.984), respectively. Fairness analyses revealed minimal disparities (true-positive rate and false-positive rate difference <0.1) across age and sex groups for both models. DeepECG-SSL demonstrated superior performance on limited-data digital biomarker tasks, with the largest improvements in long QT syndrome (LQTS) genotype classification (AUROC 0.931 vs 0.850, P = .026, n = 127 ECGs) and 5 year atrial fibrillation risk prediction (AUROC 0.742 vs 0.734, P < 0.001, n = 132 050 ECGs), while achieving superior performance in left ventricular ejection fraction ≤40% classification (AUROC 0.926 vs 0.917, P < 0.001, n = 25 252 ECGs) and comparable performance in LQTS detection (AUROC 0.767 vs 0.735, P = 0.117, n = 934 ECGs). Conclusions: This study establishes SSL as a promising paradigm for ECG analysis, particularly in settings with limited annotated data, enhancing accessibility, generalizability, and fairness in AI-driven cardiac diagnostics. By releasing model weights, preprocessing tools, and validation code, this work aims to support robust, data-efficient AI diagnostics across diverse clinical environments and questions.

Several approaches to manipulate the gut microbiome for improving the activity of cancer immune-checkpoint inhibitors (ICI) are currently under evaluation. Here, we show that oral supplementation with the polyphenol-rich berry camu-camu (CC; Myrciaria dubia) in mice shifted gut microbial composition, which translated into antitumor activity and a stronger anti–PD-1 response. We identified castalagin, an ellagitannin, as the active compound in CC. Oral administration of castalagin enriched for bacteria associated with efficient immunotherapeutic responses (Ruminococcaceae and Alistipes) and improved the CD8+/FOXP3+CD4+ ratio within the tumor microenvironment. Moreover, castalagin induced metabolic changes, resulting in an increase in taurine-conjugated bile acids. Oral supplementation of castalagin following fecal microbiota transplantation from ICI-refractory patients into mice supported anti–PD-1 activity. Finally, we found that castalagin binds to Ruminococcus bromii and promoted an anticancer response. Altogether, our results identify castalagin as a polyphenol that acts as a prebiotic to circumvent anti–PD-1 resistance.Significance:The polyphenol castalagin isolated from a berry has an antitumor effect through direct interactions with commensal bacteria, thus reprogramming the tumor microenvironment. In addition, in preclinical ICI-resistant models, castalagin reestablishes the efficacy of anti–PD-1. Together, these results provide a strong biological rationale to test castalagin as part of a clinical trial.This article is highlighted in the In This Issue feature, p. 873

Supplementary Figure from A Natural Polyphenol Exerts Antitumor Activity and Circumvents Anti–PD-1 Resistance through Effects on the Gut Microbiota

Several approaches to manipulate the gut microbiome for improving the activity of cancer immune-checkpoint inhibitors (ICI) are currently under evaluation. Here, we show that oral supplementation with the polyphenol-rich berry camu-camu (CC; Myrciaria dubia) in mice shifted gut microbial composition, which translated into antitumor activity and a stronger anti–PD-1 response. We identified castalagin, an ellagitannin, as the active compound in CC. Oral administration of castalagin enriched for bacteria associated with efficient immunotherapeutic responses (Ruminococcaceae and Alistipes) and improved the CD8+/FOXP3+CD4+ ratio within the tumor microenvironment. Moreover, castalagin induced metabolic changes, resulting in an increase in taurine-conjugated bile acids. Oral supplementation of castalagin following fecal microbiota transplantation from ICI-refractory patients into mice supported anti–PD-1 activity. Finally, we found that castalagin binds to Ruminococcus bromii and promoted an anticancer response. Altogether, our results identify castalagin as a polyphenol that acts as a prebiotic to circumvent anti–PD-1 resistance.Significance:The polyphenol castalagin isolated from a berry has an antitumor effect through direct interactions with commensal bacteria, thus reprogramming the tumor microenvironment. In addition, in preclinical ICI-resistant models, castalagin reestablishes the efficacy of anti–PD-1. Together, these results provide a strong biological rationale to test castalagin as part of a clinical trial.This article is highlighted in the In This Issue feature, p. 873

Supplementary Figure from A Natural Polyphenol Exerts Antitumor Activity and Circumvents Anti–PD-1 Resistance through Effects on the Gut Microbiota

Objective: The aim of this study was to analyze variants of the gene glial cells missing-2 (GCM2), encoding a parathyroid cell-specific transcription factor, in familial hypoparathyroidism and in familial isolated hyperparathyroidism (FIHP) without and with parathyroid carcinoma. Design: We characterized 2 families with hypoparathyroidism and 19 with FIHP in which we examined the mechanism of action of GCM2 variants. Methods: Leukocyte DNA of hypoparathyroid individuals was Sanger sequenced for CASR, PTH, GNA11 and GCM2 mutations. DNA of hyperparathyroid individuals underwent MEN1, CDKN1B, CDC73, CASR, RET and GCM2 sequencing. The actions of identified GCM2 variants were evaluated by in vitro functional analyses. Results: A novel homozygous p.R67C GCM2 mutation which failed to stimulate transcriptional activity in a luciferase assay was identified in affected members of two hypoparathyroid families. Oligonucleotide pull-down assay and in silico structural modeling indicated that this mutant had lost the ability to bind the consensus GCM recognition sequence of DNA. Two novel (p.I383M and p.T386S) and one previously reported (p.Y394S) heterozygous GCM2 variants that lie within a C-terminal conserved inhibitory domain were identified in three affected individuals of the hyperparathyroid families. One family member, heterozygous for p.I138M, had parathyroid carcinoma (PC), and a heterozygous p.V382M variant was found in another patient affected by sporadic PC. These variants exerted significantly enhanced in vitrotranscriptional activity, including increased stimulation of the PTH promoter. Conclusions: We provide evidence that two novel GCM2 R67C inactivating mutations with an inability to bind DNA are causative of hypoparathyroidism. Additionally, we provide evidence that two novel GCM2 variants increased transactivation of the PTH promoter in vitro and are associated with FIHP. Furthermore, our studies suggest that activating GCM2 variants may contribute to facilitating more aggressive parathyroid disease.

Several approaches to manipulate the gut microbiome for improving the activity of cancer immune-checkpoint inhibitors (ICI) are currently under evaluation. Here, we show that oral supplementation with the polyphenol-rich berry camu-camu (CC; Myrciaria dubia) in mice shifted gut microbial composition, which translated into antitumor activity and a stronger anti–PD-1 response. We identified castalagin, an ellagitannin, as the active compound in CC. Oral administration of castalagin enriched for bacteria associated with efficient immunotherapeutic responses (Ruminococcaceae and Alistipes) and improved the CD8+/FOXP3+CD4+ ratio within the tumor microenvironment. Moreover, castalagin induced metabolic changes, resulting in an increase in taurine-conjugated bile acids. Oral supplementation of castalagin following fecal microbiota transplantation from ICI-refractory patients into mice supported anti–PD-1 activity. Finally, we found that castalagin binds to Ruminococcus bromii and promoted an anticancer response. Altogether, our results identify castalagin as a polyphenol that acts as a prebiotic to circumvent anti–PD-1 resistance. Significance: The polyphenol castalagin isolated from a berry has an antitumor effect through direct interactions with commensal bacteria, thus reprogramming the tumor microenvironment. In addition, in preclinical ICI-resistant models, castalagin reestablishes the efficacy of anti–PD-1. Together, these results provide a strong biological rationale to test castalagin as part of a clinical trial. This article is highlighted in the In This Issue feature, p. 873