Duration:
2024-2026
Funder:
Principal investigators:
Collaborators:
Dr Vo Phuong Truc, Dr Phan Cong Chien in University Medical Center HCMC
Background:
Tuberculous meningitis (TBM) is the most severe form of tuberculosis, characterized by high mortality and long-term disability rates (1), particularly in individuals co-infected with HIV (2). TBM results from Mycobacterium tuberculosis infiltrating the brain and meninges, triggering an inflammatory response which associated with poor outcomes (3, 4). Understanding the physiopathology of TBM is crucial for improving patient outcomes.
While clinical assessment-based monitoring of TBM has been challenging due to the non-specific manifestations, magnetic resonance imaging (MRI), a non-invasive, radiation-free imaging technique, provides dimensionally precise imaging that reveals characteristic brain abnormalities and structural changes. MRI is crucial for diagnosing, monitoring, and prognosis of TBM, revealing TBM complications like hydrocephalus, tuberculomas, and infarctions, and offers insights into the disease’s pathophysiology beyond the conventional clinical monitoring alone (5, 6).
Objective:
Methods:
Our sample size consists of serial brain MRI scans from TBM patients in two trials, ACT and LAST ACT. Scans were performed at baseline (±7 days) and on days 60 (±14 days) and 365 (±30 days). The study team, including two experienced neuroradiologists from UMC, will develop an MRI-focused structured case report form (CRF).
Our analysis will rigorously delineate the pathological manifestations of TBM as depicted on MRI, including hydrocephalus, edema, infarcts, tuberculoma, enhancement, and abscesses, by assessing their presence, size, number, and distribution in the brain’s structures, meninges, and cerebral ventricles. The neuroradiologists will be blinded with clinical data and independently interpret all MRI scans using RadiAnt DICOM Viewer v2023.1 software. Data of clinical severity, CSF biomarkers (cell counts, cytokines and MMPs), neurological events, and mortality are included in the analysis.
For aim 1, we will use descriptive statistics to characterize patients, including clinical features, inflammatory biomarkers, and MRI findings. Next, we will apply unsupervised machine learning to cluster patients based on their MRI data to determine the key MRI features that classify abnormal MRI endotypes. Finally for aim 2 and 3, we will perform correlation analyses using advanced regression models to examine the association between these key MRI features with clinical parameters, CSF biomarkers and treatment outcomes and test for interactions between HIV and MRI features to assess if HIV infection acts as an effect modifier in these associations.
References:
