Trustworthy clinical AI solutions: A unified review of uncertainty quantification in Deep Learning models for medical image analysis

Lambert, Benjamin; Forbes, Florence; Doyle, Senan; Dehaene, Harmonie; Dojat, Michel

doi:10.1016/j.artmed.2024.102830

Cited by 24 publications

(4 citation statements)

References 220 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…UQ attributes included application category, method type, evaluation metrics, self-described uncertainty type (i.e., aleatoric vs. epistemic), and use of quantitative or qualitative evaluation methods. UQ application categories and definitions were adapted from Kahl et al [22] and Lambert et al [34]. Additional specific considerations for each category in the data extraction process are described in detail in Appendix B .…”

Section: Methodsmentioning

confidence: 99%

“…While previous systematic and scoping reviews have covered the topics of UQ in healthcare generally [25,31] and in relation to medical imaging [32][33][34], these studies lacked any explicit focus on RT-related applications. Therefore, we conducted this scoping review to synthesize current trends for UQ in RT and provide an outlook for the future of this important research area for clinicians and researchers.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Artificial Intelligence Uncertainty Quantification in Radiotherapy Applications - A Scoping Review

Wahid,

Kaffey,

Farris

et al. 2024

Preprint

View full text Add to dashboard Cite

Background/purpose: The use of artificial intelligence (AI) in radiotherapy (RT) is expanding rapidly. However, there exists a notable lack of clinician trust in AI models, underscoring the need for effective uncertainty quantification (UQ) methods. The purpose of this study was to scope existing literature related to UQ in RT, identify areas of improvement, and determine future directions. Methods: We followed the PRISMA-ScR scoping review reporting guidelines. We utilized the population (human cancer patients), concept (utilization of AI UQ), context (radiotherapy applications) framework to structure our search and screening process. We conducted a systematic search spanning seven databases, supplemented by manual curation, up to January 2024. Our search yielded a total of 8980 articles for initial review. Manuscript screening and data extraction was performed in Covidence. Data extraction categories included general study characteristics, RT characteristics, AI characteristics, and UQ characteristics. Results: We identified 56 articles published from 2015-2024. 10 domains of RT applications were represented; most studies evaluated auto-contouring (50%), followed by image-synthesis (13%), and multiple applications simultaneously (11%). 12 disease sites were represented, with head and neck cancer being the most common disease site independent of application space (32%). Imaging data was used in 91% of studies, while only 13% incorporated RT dose information. Most studies focused on failure detection as the main application of UQ (60%), with Monte Carlo dropout being the most commonly implemented UQ method (32%) followed by ensembling (16%). 55% of studies did not share code or datasets. Conclusion: Our review revealed a lack of diversity in UQ for RT applications beyond auto-contouring. Moreover, there was a clear need to study additional UQ methods, such as conformal prediction. Our results may incentivize the development of guidelines for reporting and implementation of UQ in RT.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Artificial Intelligence Uncertainty Quantification in Radiotherapy Applications - A Scoping Review

Wahid,

Kaffey,

Farris

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…Examining AI's functionalities and practical implementation strategies is vital to enhance privacy measures and optimize resource allocation in healthcare settings. However, it is essential to acknowledge that while AI shows promising outcomes, its implementation may pose unforeseen challenges and adverse effects, impacting medical treatment and healthcare delivery within hospitals [1,[12][13][14][15]. Furthermore, a pressing need exists to delve deeper into the potential applications of AI in nursing and healthcare to enhance overall healthcare provision.…”

Section: Introductionmentioning

confidence: 99%

“…It is essential to address pertinent issues such as bias and algorithmic considerations, which are integral to evaluating AI interventions' e cacy and ethical implications. Ultimately, prudent use of AI technology is essential in yielding outcomes aligned with the highest ethical standards and practices in nursing and healthcare [1,[14][15][16][17].…”

Section: Introductionmentioning

confidence: 99%

Exploring the Deep Learning of Artificial Intelligence in Nursing: A Concept Analysis with Walker and Avant’s Approach

Wangpitipanit,

Lininger,

Anderson

2024

Preprint

View full text Add to dashboard Cite

Background: In recent years, increased attention has been given to the use ofartificial intelligence (AI) or artificial intelligence (AI) deep learning(DL) in healthcare to address nursing challenges. However, the adoption of new technologies in nursing needs to be improved, and AI in nursing is still in its early stages. However, the current literature needs more clarity, which affects clinical practice, research, and theory development. This study aimed to clarify the meaning of deep learning and identify the defining attributes of artificial intelligence within nursing. Methods: We conducted a concept analysis of the deep learning of AI in nursing care using Walker and Avant's 8-step approach. Our search strategy employed Boolean techniques across databases, including BMC, CINAHL, ClinicalKey for Nursing, Embase, Google Scholar, Ovid, Scopus, SpringerLink, ProQuest, PubMed, and Web of Science. By focusing on relevant keywords in titles and abstracts from articles published between 2018 and 2024, we initially found 574 sources. Results: Thirty-six articles that met the inclusion criteria were analyzed in this study. The attributes of evidence included four themes: focus and immersion, coding and understanding, arranging layers and algorithms, and implementing within the process of use cases to modify recommendations. Antecedents, unclear systems and communication, insufficient data management knowledge and support, and compound challenges can lead to suffering and risky caregiving tasks. The application of DL deep learning techniques enables nurses to simulate scenarios, predict outcomes, and plan care with greater precision. Embracing deep learning equipment allows nurses to makebetter decisions and empower them with enhanced knowledge, while ensuring adequate support and resources is essential for caregiver and patient well-being, and access to necessary equipment is vital for high-quality home healthcare. Conclusion: This study provides a clearer understanding of the use of deep learning in nursing and its implications for nursing practice. Future research should focus on exploring the impact of deep learning on healthcare operations management through quantitative and qualitative studies. Additionally, the development of a framework to guide the integration of deep learning into nursing practice is recommended to facilitate its adoption and implementation.

show abstract

Towards Aleatoric and Epistemic Uncertainty in Medical Image Classification

Löhr,

Ingrisch,

Hüllermeier

2024

Artificial Intelligence in Medicine

View full text Add to dashboard Cite

Trustworthy clinical AI solutions: A unified review of uncertainty quantification in Deep Learning models for medical image analysis

Cited by 24 publications

References 220 publications

Artificial Intelligence Uncertainty Quantification in Radiotherapy Applications - A Scoping Review

Artificial Intelligence Uncertainty Quantification in Radiotherapy Applications - A Scoping Review

Exploring the Deep Learning of Artificial Intelligence in Nursing: A Concept Analysis with Walker and Avant’s Approach

Towards Aleatoric and Epistemic Uncertainty in Medical Image Classification

Contact Info

Product

Resources

About