PET/CT Imaging Features and Clusters for Fever of Unknown Origin

A retrospective cohort of 289 adult patients with fever of unknown origin (FUO) or inflammation of unknown origin (IUO) who underwent 18F-FDG PET/CT imaging. The dataset contains extracted imaging features, including organ system involvement, maximum SUVmax values, and uptake patterns, which were used to derive four distinct patient clusters via consensus unsupervised learning. The clusters represent PET/CT phenotypes that showed low concordance with final clinical diagnoses, indicating they capture imaging patterns rather than specific disease etiologies.

Use Cases

• Validating unsupervised clustering algorithms on medical imaging features from PET/CT scans.
• Analyzing correlations between imaging phenotypes (e.g., reticuloendothelial involvement) and final diagnostic categories like infection or rheumatologic disease.
• Developing feature-based models for patient stratification in fever of unknown origin studies.

Strengths

• Includes data from 289 clinically characterized FUO/IUO patients with detailed demographic information (mean age 59 ± 17 years, 44.3% women).
• Imaging features are quantitatively extracted, including SUVmax values and categorized uptake patterns (focal, diffuse, mixed).
• Derived clusters are the result of a consensus approach using four distinct unsupervised machine learning algorithms.

Limitations

• Column names and the exact data structure are not provided in the source metadata.
• The dataset size is reported inconsistently across entries (15079 and 14761), creating ambiguity about the exact file size.
• As a retrospective study, it may have inherent selection biases and lacks prospective validation.

Provenance

Source

Retrospective cohort study by Sule Ceylan.

Collection Method

Data extracted from 18F-FDG PET/CT scans of adult FUO patients, with features on organ system involvement and radiotracer uptake.

Freshness

Last updated 2026-05-29.