Optimised body CT protocols using deep learning-based image reconstruction are associated with lower electricity use, reduced iodinated contrast media consumption and lower carbon dioxide equivalent emissions in a high-volume imaging setting. The retrospective single-centre experience, published in Insights into Imaging, compared four CT scanners at a tertiary referral hospital in Italy over an 18-month period. Two scanners used hybrid iterative reconstruction, while two newer scanners incorporated deep learning-based image reconstruction. The comparison focused on routine body CT examinations and assessed electricity consumption, carbon emissions and contrast media use at scanner level and per examination.
Lower Voltage Protocols Support CT Optimisation
The comparison included 42,300 body CT examinations acquired between January 2024 and June 2025. Hybrid iterative reconstruction scanners performed 23,096 examinations, while deep learning-based reconstruction scanners performed 19,204 examinations. Included protocols covered pulmonary embolism CT angiography, aortic CT angiography, coronary CT angiography, contrast-enhanced abdomino-pelvic CT, high-resolution chest CT, oncologic whole-body CT, trauma whole-body CT and paediatric CT. Patients with a body mass index of 30 or above were excluded to limit variability in acquisition parameters. Neuroimaging, skeletal segment imaging and extremity CT examinations were also excluded.
The deep learning-based protocols were derived from corresponding hybrid iterative reconstruction protocols, with acquisition parameters selectively optimised while scanner-dependent settings remained unchanged. Tube voltage was reduced to 80 or 100 kV according to examination type. Contrast-enhanced protocols also used lower iodinated contrast media doses, based on the improved contrast-to-noise ratio achievable with lower voltage imaging. Injection flow rate and timing were preserved across protocols. The same contrast media was used for both groups. The approach therefore combines newer reconstruction technology with protocol changes rather than testing reconstruction alone.
Must Read: Sustainable Breast Imaging Faces Carbon Trade-Offs
Electricity Use and Carbon Emissions Fall
Electricity consumption was lower for scanners equipped with deep learning-based reconstruction over the 18-month period. Hybrid iterative reconstruction scanners used 123,000 kWh, corresponding to 30.75 tonnes of carbon dioxide equivalent emissions from electricity. Deep learning-based reconstruction scanners used 66,927 kWh, corresponding to 16.73 tonnes of carbon dioxide equivalent emissions. At scanner level, this represented a reduction of 28,037 kWh and 7.01 tonnes of carbon dioxide equivalent emissions per scanner over the study period.
When normalised per examination, hybrid iterative reconstruction scanners consumed 5.33 kWh per CT examination, compared with 3.49 kWh for deep learning-based reconstruction scanners. Carbon emissions linked to electricity were 1.33 kg of carbon dioxide equivalent per examination in the hybrid iterative reconstruction group and 0.87 kg per examination in the deep learning-based reconstruction group. The lower energy demand is consistent with the reduced tube voltage used in the optimised protocols and with the near-quadratic relationship between tube voltage and X-ray tube power demand. The results are presented as environmental outcomes derived from aggregated scanner-level data, with protocol-level attribution used for normalisation.
Contrast Media Reduction Adds Environmental Benefit
Iodinated contrast media use was also lower in the deep learning-based reconstruction group. Total contrast media administration was 1,832 litres for hybrid iterative reconstruction scanners and 1,201 litres for deep learning-based reconstruction scanners. These volumes corresponded to 18.87 and 12.37 tonnes of carbon dioxide equivalent emissions respectively. Because total contrast use can reflect differences in examination case mix as well as protocol design, a reference-equivalent analysis applied both dosing approaches to the same deep learning-based reconstruction case mix.
Using that approach, lower nominal contrast doses on deep learning-based reconstruction scanners produced a reference-equivalent saving of 434 litres over 18 months or 289 litres per year. The reduced contrast media use corresponded to 4.47 tonnes of avoided carbon dioxide equivalent emissions over the study period and 60,730 litres of water preserved. Combined emissions from electricity and contrast media totalled 49.62 tonnes for hybrid iterative reconstruction scanners and 29.10 tonnes for deep learning-based reconstruction scanners. The environmental benefit therefore comes from both lower energy consumption and reduced contrast media use, with electricity accounting for the larger share.
Deep learning-based reconstruction supports routine body CT protocols with lower tube voltage and reduced iodinated contrast media dosing in a high-volume clinical setting. The comparison shows lower electricity consumption, lower contrast media use and reduced combined carbon dioxide equivalent emissions for scanners using the optimised protocols. The results should be interpreted with caution because scanner generation, protocol changes and reconstruction technology cannot be fully separated. Image quality was not directly compared, and environmental estimates for contrast media were based on life-cycle data.
Source: Insights into Imaging
Image Credit: iStock
References:
Franco PN, Maino C, Gandola D et al. (2026) Reduced environmental impact in body CT imaging with deep learning reconstruction: experience of a high-volume tertiary referral center. Insights Imaging; 17, 170.
