Unsupervised Pathology Detection: A Deep Dive Into the State of the Art DOI
Ioannis Lagogiannis, Felix Meissen, Georgios Kaissis

et al.

IEEE Transactions on Medical Imaging, Journal Year: 2023, Volume and Issue: 43(1), P. 241 - 252

Published: July 28, 2023

Deep unsupervised approaches are gathering increased attention for applications such as pathology detection and segmentation in medical images since they promise to alleviate the need large labeled datasets more generalizable than their supervised counterparts detecting any kind of rare pathology. As Unsupervised Anomaly Detection (UAD) literature continuously grows new paradigms emerge, it is vital evaluate benchmark methods a common framework, order reassess state-of-the-art (SOTA) identify promising research directions. To this end, we diverse selection cutting-edge UAD on multiple datasets, comparing them against established SOTA brain MRI. Our experiments demonstrate that newly developed feature-modeling from industrial achieve performance compared previous work set variety modalities datasets. Additionally, show capable benefiting recently self-supervised pre-training algorithms, further increasing performance. Finally, perform series gain insights into some unique characteristics selected models code can be found under https://github.com/iolag/UPD_study/.

Language: Английский

A systematic review of trustworthy and explainable artificial intelligence in healthcare: Assessment of quality, bias risk, and data fusion DOI
A. S. Albahri, Ali M. Duhaim, Mohammed A. Fadhel

et al.

Information Fusion, Journal Year: 2023, Volume and Issue: 96, P. 156 - 191

Published: March 15, 2023

Language: Английский

Citations

364
The Current and Future State of AI Interpretation of Medical Images DOI
Pranav Rajpurkar, Matthew P. Lungren

New England Journal of Medicine, Journal Year: 2023, Volume and Issue: 388(21), P. 1981 - 1990

Published: May 24, 2023

The authors examine the advantages and limitations of current clinical radiologic AI systems, new workflows, potential effect generative large multimodal foundation models.

Language: Английский

Citations

222
Algorithmic fairness in artificial intelligence for medicine and healthcare DOI
Richard J. Chen, Judy J. Wang, Drew F. K. Williamson

et al.

Nature Biomedical Engineering, Journal Year: 2023, Volume and Issue: 7(6), P. 719 - 742

Published: June 28, 2023

Language: Английский

Citations

203
Heterogeneity and predictors of the effects of AI assistance on radiologists DOI Creative Commons
Feiyang Yu, Alex Moehring,

Oishi Banerjee

et al.

Nature Medicine, Journal Year: 2024, Volume and Issue: 30(3), P. 837 - 849

Published: March 1, 2024

Abstract The integration of artificial intelligence (AI) in medical image interpretation requires effective collaboration between clinicians and AI algorithms. Although previous studies demonstrated the potential assistance improving overall clinician performance, individual impact on remains unclear. This large-scale study examined heterogeneous effects 140 radiologists across 15 chest X-ray diagnostic tasks identified predictors these effects. Surprisingly, conventional experience-based factors, such as years experience, subspecialty familiarity with tools, fail to reliably predict assistance. Additionally, lower-performing do not consistently benefit more from assistance, challenging prevailing assumptions. Instead, we found that occurrence errors strongly influences treatment outcomes, inaccurate predictions adversely affecting radiologist performance aggregate all pathologies half investigated. Our findings highlight importance personalized approaches clinician–AI accurate models. By understanding factors shape effectiveness this provides valuable insights for targeted implementation AI, enabling maximum benefits clinical practice.

Language: Английский

Citations

47
Explainable AI in Healthcare Application DOI
Siva Raja Sindiramutty, Wee Jing Tee, Sumathi Balakrishnan

et al.

Advances in computational intelligence and robotics book series, Journal Year: 2024, Volume and Issue: unknown, P. 123 - 176

Published: Jan. 18, 2024

Given the inherent risks in medical decision-making, professionals carefully evaluate a patient's symptoms before arriving at plausible diagnosis. For AI to be widely accepted and useful technology, it must replicate human judgment interpretation abilities. XAI attempts describe data underlying black-box approach of deep learning (DL), machine (ML), natural language processing (NLP) that explain how judgments are made. This chapter provides survey most recent methods employed imaging related fields, categorizes lists types XAI, highlights used make topics more interpretable. Additionally, focuses on challenging issues applications guides development better deep-learning system explanations by applying principles analysis pictures text.

Language: Английский

Citations

19
Trustworthy deep learning framework for the detection of abnormalities in X-ray shoulder images DOI Creative Commons
Laith Alzubaidi, Asma Salhi, Mohammed A. Fadhel

et al.

PLoS ONE, Journal Year: 2024, Volume and Issue: 19(3), P. e0299545 - e0299545

Published: March 11, 2024

Musculoskeletal conditions affect an estimated 1.7 billion people worldwide, causing intense pain and disability. These lead to 30 million emergency room visits yearly, the numbers are only increasing. However, diagnosing musculoskeletal issues can be challenging, especially in emergencies where quick decisions necessary. Deep learning (DL) has shown promise various medical applications. previous methods had poor performance a lack of transparency detecting shoulder abnormalities on X-ray images due training data better representation features. This often resulted overfitting, generalisation, potential bias decision-making. To address these issues, new trustworthy DL framework been proposed detect (such as fractures, deformities, arthritis) using images. The consists two parts: same-domain transfer (TL) mitigate imageNet mismatch feature fusion reduce error rates improve trust final result. Same-domain TL involves pre-trained models large number labelled from body parts fine-tuning them target dataset Feature combines extracted features with seven train several ML classifiers. achieved excellent accuracy rate 99.2%, F1 Score Cohen’s kappa 98.5%. Furthermore, results was validated three visualisation tools, including gradient-based class activation heat map (Grad CAM), visualisation, locally interpretable model-independent explanations (LIME). outperformed orthopaedic surgeons invited classify test set, who obtained average 79.1%. proven effective robust, improving generalisation increasing results.

Language: Английский

Citations

17
Artificial intelligence-guided detection of under-recognised cardiomyopathies on point-of-care cardiac ultrasonography: a multicentre study DOI Creative Commons
Evangelos K. Oikonomou, Akhil Vaid, Gregory Holste

et al.

The Lancet Digital Health, Journal Year: 2025, Volume and Issue: 7(2), P. e113 - e123

Published: Jan. 29, 2025

Language: Английский

Citations

4
Deep learning generates synthetic cancer histology for explainability and education DOI Creative Commons
James M. Dolezal,

Rachelle Wolk,

Hanna M. Hieromnimon

et al.

npj Precision Oncology, Journal Year: 2023, Volume and Issue: 7(1)

Published: May 29, 2023

Artificial intelligence methods including deep neural networks (DNN) can provide rapid molecular classification of tumors from routine histology with accuracy that matches or exceeds human pathologists. Discerning how make their predictions remains a significant challenge, but explainability tools help insights into what models have learned when corresponding histologic features are poorly defined. Here, we present method for improving DNN using synthetic generated by conditional generative adversarial network (cGAN). We show cGANs generate high-quality images be leveraged explaining trained to classify molecularly-subtyped tumors, exposing associated state. Fine-tuning through class and layer blending illustrates nuanced morphologic differences between tumor subtypes. Finally, demonstrate the use augmenting pathologist-in-training education, showing these intuitive visualizations reinforce improve understanding manifestations biology.

Language: Английский

Citations

41
Artificial intelligence in liver cancers: Decoding the impact of machine learning models in clinical diagnosis of primary liver cancers and liver cancer metastases DOI Creative Commons
Anita K. Bakrania,

Narottam Joshi,

Xun Zhao

et al.

Pharmacological Research, Journal Year: 2023, Volume and Issue: 189, P. 106706 - 106706

Published: Feb. 20, 2023

Liver cancers are the fourth leading cause of cancer-related mortality worldwide. In past decade, breakthroughs in field artificial intelligence (AI) have inspired development algorithms cancer setting. A growing body recent studies evaluated machine learning (ML) and deep (DL) for pre-screening, diagnosis management liver patients through diagnostic image analysis, biomarker discovery predicting personalized clinical outcomes. Despite promise these early AI tools, there is a significant need to explain 'black box' work towards deployment enable ultimate translatability. Certain emerging fields such as RNA nanomedicine targeted therapy may also benefit from application AI, specifically nano-formulation research given that they still largely reliant on lengthy trial-and-error experiments. this paper, we put forward current landscape along with challenges management. Finally, discussed future perspectives how multidisciplinary approach using could accelerate transition medicine bench side clinic.

Language: Английский

Citations

40
Estimating age and gender from electrocardiogram signals: A comprehensive review of the past decade DOI Creative Commons
Mohammed Yusuf Ansari, Marwa Qaraqe, Fatmeh Charafeddine

et al.

Artificial Intelligence in Medicine, Journal Year: 2023, Volume and Issue: 146, P. 102690 - 102690

Published: Oct. 21, 2023

Twelve lead electrocardiogram signals capture unique fingerprints about the body's biological processes and electrical activity of heart muscles. Machine learning deep learning-based models can learn embedded patterns in to estimate complex metrics such as age gender that depend on multiple aspects human physiology. ECG estimated with respect chronological reflects overall well-being cardiovascular system, significant positive deviations indicating an aged system a higher likelihood mortality. Several conventional, machine learning, methods have been proposed from electronic health records, surveys, data. This manuscript comprehensively reviews methodologies for ECG-based estimation over last decade. Specifically, review highlights elevated is associated atherosclerotic disease, abnormal peripheral endothelial dysfunction, high mortality, among many other disorders. Furthermore, survey presents overarching observations insights across estimation. paper also several essential methodological improvements clinical applications ECG-estimated encourage further state-of-the-art methodologies.

Language: Английский

Citations

32