CT-Based Clinical-Radiomics for Differentiating Talaromyces marneffei Spondylitis From Tuberculous Spondylitis Before...

Talaromyces marneffei spondylitis (TMS) is an uncommon but clinically important form of talaromycosis in endemic regions. Where talaromycosis and tuberculosis overlap, TMS may be worked up as tuberculous spondylitis (TBS), delaying fungal investigations and antifungal treatment.

To assess whether pretreatment CT, routine clinical variables and radiomics features can help differentiate TMS from TBS at the initial assessment stage.

This retrospective multicentre study included 263 patients meeting the prespecified composite reference standard (83 TMS and 180 TBS). Centre A formed the training and internal-validation cohorts; Centres B and C formed the external cohort (n = 50; 16 TMS and 34 TBS). Six clinical variables and 14 CT radiomics features were used to develop clinical, radiomics and combined models.

TMS showed a stronger systemic inflammatory phenotype, whereas TBS generally showed more frequent disc-endplate involvement and osseous destruction. The clinical-radiomics model achieved the highest AUCs in the training, internal validation and external validation cohorts: 0.958 (95% CI, 0.924-0.992), 0.940 (95% CI, 0.871-1.000) and 0.927 (95% CI, 0.867-0.987), respectively. The external confidence interval was wide, reflecting the limited external sample size.

In regions where talaromycosis and tuberculosis overlap, a pretreatment CT-based clinical-radiomics approach may help flag patients in whom TMS should be considered early. This approach could support earlier fungal investigations while definitive microbiological, molecular or histopathological results are pending. It addresses only the TMS-versus-TBS differential and is not a general spinal-infection triage tool; the findings support further evaluation in endemic settings.