FUTURE-AI: Guiding Principles and Consensus Recommendations for Trustworthy Artificial Intelligence in Medical Imaging
arXiv:2109.09658
Abstract
The recent advancements in artificial intelligence (AI) combined with the extensive amount of data generated by today's clinical systems, has led to the development of imaging AI solutions across the whole value chain of medical imaging, including image reconstruction, medical image segmentation, image-based diagnosis and treatment planning. Notwithstanding the successes and future potential of AI in medical imaging, many stakeholders are concerned of the potential risks and ethical implications of imaging AI solutions, which are perceived as complex, opaque, and difficult to comprehend, utilise, and trust in critical clinical applications. Addressing these concerns and risks, the FUTURE-AI framework has been proposed, which, sourced from a global multi-domain expert consensus, comprises guiding principles for increased trust, safety, and adoption for AI in healthcare. In this paper, we transform the general FUTURE-AI healthcare principles to a concise and specific AI implementation guide tailored to the needs of the medical imaging community. To this end, we carefully assess each building block of the FUTURE-AI framework consisting of (i) Fairness, (ii) Universality, (iii) Traceability, (iv) Usability, (v) Robustness and (vi) Explainability, and respectively define concrete best practices based on accumulated AI implementation experiences from five large European projects on AI in Health Imaging. We accompany our concrete step-by-step medical imaging development guide with a practical AI solution maturity checklist, thus enabling AI development teams to design, evaluate, maintain, and deploy technically, clinically and ethically trustworthy imaging AI solutions into clinical practice.
Please refer to arXiv:2309.12325 for the latest FUTURE-AI framework for healthcare
References in corpus (23)
- Artificial Intelligence: the global landscape of ethics guidelines
- Image biomarker standardisation initiative
- Towards A Rigorous Science of Interpretable Machine Learning
- Axiomatic Attribution for Deep Networks
- Striving for Simplicity: The All Convolutional Net
- Phase recovery and holographic image reconstruction using deep learning in neural networks
- This Looks Like That: Deep Learning for Interpretable Image Recognition
- Measuring the Quality of Explanations: The System Causability Scale (SCS). Comparing Human and Machine Explanations
- Embracing Imperfect Datasets: A Review of Deep Learning Solutions for Medical Image Segmentation
- Fairness in Machine Learning
- To Trust Or Not To Trust A Classifier
- Test-Time Adaptable Neural Networks for Robust Medical Image Segmentation
- Concept Bottleneck Models
- Deep Structural Causal Models for Tractable Counterfactual Inference
- Deep neural network or dermatologist?
- Concept drift detection and adaptation for federated and continual learning
- Integrating uncertainty in deep neural networks for MRI based stroke analysis
- The Doctor Just Won't Accept That!
- Interpretable Deep Models for Cardiac Resynchronisation Therapy Response Prediction
- This Looks Like That... Does it? Shortcomings of Latent Space Prototype Interpretability in Deep Networks
- Estimating and Improving Fairness with Adversarial Learning
- Unbox the Black-box for the Medical Explainable AI via Multi-modal and Multi-centre Data Fusion: A Mini-Review, Two Showcases and Beyond
- Adversarial Robust Training of Deep Learning MRI Reconstruction Models