Researchers developed a body composition fitness score using deep learning analysis of routine CT scans from 36,471 patients across 16 medical centers. The score predicted good exercise capacity (completing ≥7 minutes on Bruce protocol) with 77.1% accuracy, outperforming age-based predictions. Higher skeletal muscle density emerged as a key marker, and patients with high fitness scores showed 30% lower cardiovascular death risk. This breakthrough transforms routine medical imaging into actionable fitness assessment, particularly valuable for patients unable to perform traditional exercise stress tests due to mobility limitations, joint problems, or other medical conditions. The automated approach could revolutionize preventive cardiology by identifying at-risk patients during routine scans for other conditions. However, this preprint awaits peer review, and results may change during the editorial process. The study's multicenter design strengthens generalizability, though the observational nature cannot establish causation between muscle composition and fitness outcomes. If validated, this technology could enable early intervention strategies and personalized exercise prescriptions based on objective body composition data rather than subjective self-reported fitness levels.