Evaluating the environmental sustainability of AI in radiology: a systematic review of current practice

Abstract

Purpose: Data- and energy-heavy artificial intelligence (AI) technologies are increasingly applied in radiology, often without consideration of potential environmental consequences. We aimed to assess current practice in evaluating environmental sustainability (ES) impacts of AI-enabled clinical pathways in radiology.
Methods: We searched MEDLINE and Embase on 5 November 2024 for clinical radiology studies that used AI to aid in radiological diagnosis or intervention and discussed ES impacts. We included peer-reviewed, English-language studies published from 2015 onwards. Our primary outcome was any quantitative reporting of ES impacts (including carbon emissions, energy/water/mineral usage, waste generation/disposal and impacts on material environments), and our secondary outcome was any within-text qualitative discussion of ES impacts. For quantitative outcomes we conducted synthesis without meta-analysis based on effect size, with our secondary outcome synthesised narratively. The study was pre-registered with PROSPERO: CRD42024601818.
Results: Of 4,365 citations screened, 16 met our inclusion criteria. 6 reported quantitative ES outcomes, and 13 included qualitative discussion of ES. When applied to the same tasks, algorithms designed to be ‘lightweight’ (meaning less computationally intensive) generated from 2.19 to 17.15 times less carbon emissions (median 7.81, 16 datapoints) and from 1.60 to 751.62 times less energy consumption (median 3.22, 16 datapoints) compared with state-of-the-art alternatives, while maintaining similar or improved clinical performance. No studies compared ES outcomes for an AI-enabled pathway versus standard of care, and 70% of studies reporting only on our secondary outcome included just one sentence on sustainability. All included studies were published within the past four years, with most (75%) from 2023/24.
Conclusion: Despite increasing concern about the climate impacts of AI, environmental outcomes are rarely measured within evaluations of AI-enabled clinical pathways in radiology. However, evidence suggests designing AI products with sustainability in mind can substantially reduce their carbon footprint. Environmental sustainability should be better integrated into AI evaluation and procurement to ensure costs and benefits for both climate and health are fully considered.