About the Article
Article Files
[This article belongs to Volume - 71, Issue - 4]
Published on : 2026-04-17 19:06:17
Article Code: AMJ-17-04-2026-12375
Title : Deep Learning-Integrated Multi-Omic Liquid Biopsy for Rapid Diagnosis of EBV-Associated Burkitt Lymphoma in the Eastern Mediterranean: A Prospective Diagnostic Accuracy Study
Author(s) : Nikolaos Papadopoulos, Maria Eleftheriadou
Abstract :
Background: Burkitt lymphoma (BL) is a highly aggressive non-Hodgkin lymphoma endemic to Epstein-Barr virus (EBV) and malaria regions. Diagnostic delays exceeding 90 days remain prevalent in resource-limited settings due to scarce pathology infrastructure. We developed and validated an artificial intelligence (AI)-integrated, multi-omic liquid biopsy platform for rapid BL diagnosis across the Eastern Mediterranean region.
Methods: We conducted a prospective, multicenter diagnostic accuracy study across 12 hospitals in Greece, Cyprus, and Lebanon (2021–2025). A total of 1,247 children and young adults (age 3–25 years) with suspected lymphoma underwent concurrent tissue biopsy (gold standard) and liquid biopsy. The comprehensive assay integrated targeted next-generation sequencing (NGS) for MYC-IG translocations, single-cell RNA sequencing (scRNA-seq), EBV fragmentomics, and plasma proteomics. We trained a deep learning ensemble model (convolutional neural network + transformer architecture) and compared its performance against conventional histopathology in a head-to-head prospective validation cohort.
Results: The AI-integrated comprehensive model achieved superior diagnostic accuracy (area under the receiver operating characteristic curve [AUC] 0.994; 95% confidence interval [CI] 0.988–0.998) compared to clinical models alone (AUC 0.851; ). Sensitivity was 98.7% (95% CI 96.9–99.5%) and specificity 99.3% (95% CI 98.2–99.8%). Median diagnostic turnaround time was reduced from 58.4 days (IQR 31.2–124.6) for tissue-based diagnosis to 18 hours (IQR 12–28 hours) for liquid biopsy ( ) using nanopore sequencing technology. In 52.3% of cases, liquid biopsy provided the only diagnostic result available at the initial multidisciplinary tumor board, enabling immediate treatment initiation. Notably, the assay detected cryptic MYC rearrangements in 12 cases missed by conventional fluorescence in situ hybridization (FISH). Cost-effectiveness analysis demonstrated an incremental cost-effectiveness ratio (ICER) of €280 per quality-adjusted life year (QALY) gained.
Conclusions: AI-integrated multi-omic liquid biopsy enables near-perfect diagnostic accuracy for EBV-associated BL with same-day turnaround, offering a transformative solution for pediatric oncology care in resource-constrained environments.