Diagnostic accuracy of deep learning in medical imaging: a systematic review and meta-analysis

npj Digital Medicine, Год журнала: 2021, Номер 4(1)

Опубликована: Апрель 7, 2021

Deep learning (DL) has the potential to transform medical diagnostics. However, diagnostic accuracy of DL is uncertain. Our aim was evaluate algorithms identify pathology in imaging. Searches were conducted Medline and EMBASE up January 2020. We identified 11,921 studies, which 503 included systematic review. Eighty-two studies ophthalmology, 82 breast disease 115 respiratory for meta-analysis. Two hundred twenty-four other specialities qualitative Peer-reviewed that reported on using imaging included. Primary outcomes measures accuracy, study design reporting standards literature. Estimates pooled random-effects In AUC's ranged between 0.933 1 diagnosing diabetic retinopathy, age-related macular degeneration glaucoma retinal fundus photographs optical coherence tomography. imaging, 0.864 0.937 lung nodules or cancer chest X-ray CT scan. For 0.868 0.909 mammogram, ultrasound, MRI digital tomosynthesis. Heterogeneity high extensive variation methodology, terminology outcome noted. This can lead an overestimation There immediate need development artificial intelligence-specific EQUATOR guidelines, particularly STARD, order provide guidance around key issues this field.

Язык: Английский

---

Applications of Artificial Intelligence and Machine learning in smart cities

Computer Communications, Год журнала: 2020, Номер 154, С. 313 - 323

Опубликована: Март 1, 2020

Язык: Английский

---

AI applications to medical images: From machine learning to deep learning

Physica Medica, Год журнала: 2021, Номер 83, С. 9 - 24

Опубликована: Март 1, 2021

Язык: Английский

---

Predicting cancer outcomes with radiomics and artificial intelligence in radiology

Nature Reviews Clinical Oncology, Год журнала: 2021, Номер 19(2), С. 132 - 146

Опубликована: Окт. 18, 2021

Язык: Английский

---

Artificial Intelligence in Cancer Research and Precision Medicine

Cancer Discovery, Год журнала: 2021, Номер 11(4), С. 900 - 915

Опубликована: Апрель 1, 2021

Artificial intelligence (AI) is rapidly reshaping cancer research and personalized clinical care. Availability of high-dimensionality datasets coupled with advances in high-performance computing, as well innovative deep learning architectures, has led to an explosion AI use various aspects oncology research. These applications range from detection classification cancer, molecular characterization tumors their microenvironment, drug discovery repurposing, predicting treatment outcomes for patients. As these start penetrating the clinic, we foresee a shifting paradigm care becoming strongly driven by AI. SIGNIFICANCE: potential dramatically affect nearly all oncology-from enhancing diagnosis personalizing discovering novel anticancer drugs. Here, review recent enormous progress application oncology, highlight limitations pitfalls, chart path adoption clinic.

Язык: Английский

---

Deep learning in cancer pathology: a new generation of clinical biomarkers

British Journal of Cancer, Год журнала: 2020, Номер 124(4), С. 686 - 696

Опубликована: Ноя. 17, 2020

Abstract Clinical workflows in oncology rely on predictive and prognostic molecular biomarkers. However, the growing number of these complex biomarkers tends to increase cost time for decision-making routine daily practice; furthermore, often require tumour tissue top diagnostic material. Nevertheless, routinely available contains an abundance clinically relevant information that is currently not fully exploited. Advances deep learning (DL), artificial intelligence (AI) technology, have enabled extraction previously hidden directly from histology images cancer, providing potentially useful information. Here, we outline emerging concepts how DL can extract summarise studies basic advanced image analysis cancer histology. Basic tasks include detection, grading subtyping images; they are aimed at automating pathology consequently do immediately translate into clinical decisions. Exceeding such approaches, has also been used tasks, which potential affecting processes. These approaches inference features, prediction survival end-to-end therapy response. Predictions made by systems could simplify enrich decision-making, but rigorous external validation settings.

Язык: Английский

---

AI in Medical Imaging Informatics: Current Challenges and Future Directions

IEEE Journal of Biomedical and Health Informatics, Год журнала: 2020, Номер 24(7), С. 1837 - 1857

Опубликована: Май 29, 2020

This paper reviews state-of-the-art research solutions across the spectrum of medical imaging informatics, discusses clinical translation, and provides future directions for advancing practice. More specifically, it summarizes advances in acquisition technologies different modalities, highlighting necessity efficient data management strategies context AI big healthcare analytics. It then a synopsis contemporary emerging algorithmic methods disease classification organ/ tissue segmentation, focusing on deep learning architectures that have already become de facto approach. The benefits in-silico modelling linked with evolving 3D reconstruction visualization applications are further documented. Concluding, integrative analytics approaches driven by associate branches highlighted this study promise to revolutionize informatics as known today continuum both radiology digital pathology applications. latter, is projected enable informed, more accurate diagnosis, timely prognosis, effective treatment planning, underpinning precision medicine.

Язык: Английский

---

Big Self-Supervised Models Advance Medical Image Classification

2021 IEEE/CVF International Conference on Computer Vision (ICCV), Год журнала: 2021, Номер unknown, С. 3458 - 3468

Опубликована: Окт. 1, 2021

Self-supervised pretraining followed by supervised fine-tuning has seen success in image recognition, especially when labeled examples are scarce, but received limited attention medical analysis. This paper studies the effectiveness of self-supervised learning as a pre-training strategy for classification. We conduct experiments on two distinct tasks: dermatology condition classification from digital camera images and multi-label chest X-ray classification, demonstrate that ImageNet, additional unlabeled domain-specific significantly improves accuracy classifiers. introduce novel Multi-Instance Contrastive Learning (MICLe) method uses multiple underlying pathology per patient case, available, to construct more informative positive pairs learning. Combining our contributions, we achieve an improvement 6.7% top-1 1.1% mean AUC respectively, outperforming strong baselines pretrained ImageNet. In addition, show big models robust distribution shift can learn efficiently with small number images.

Язык: Английский

---

How Machine Learning Will Transform Biomedicine

Cell, Год журнала: 2020, Номер 181(1), С. 92 - 101

Опубликована: Апрель 1, 2020

Язык: Английский

---

Guidelines for clinical trial protocols for interventions involving artificial intelligence: the SPIRIT-AI extension

Nature Medicine, Год журнала: 2020, Номер 26(9), С. 1351 - 1363

Опубликована: Сен. 1, 2020

The SPIRIT 2013 statement aims to improve the completeness of clinical trial protocol reporting by providing evidence-based recommendations for minimum set items be addressed. This guidance has been instrumental in promoting transparent evaluation new interventions. More recently, there a growing recognition that interventions involving artificial intelligence (AI) need undergo rigorous, prospective demonstrate their impact on health outcomes. SPIRIT-AI (Standard Protocol Items: Recommendations Interventional Trials-Artificial Intelligence) extension is guideline protocols evaluating with an AI component. It was developed parallel its companion reports: CONSORT-AI (Consolidated Standards Reporting Intelligence). Both guidelines were through staged consensus process literature review and expert consultation generate 26 candidate items, which consulted upon international multi-stakeholder group two-stage Delphi survey (103 stakeholders), agreed meeting (31 stakeholders) refined checklist pilot (34 participants). includes 15 considered sufficiently important These should routinely reported addition core items. recommends investigators provide clear descriptions intervention, including instructions skills required use, setting intervention will integrated, considerations handling input output data, human-AI interaction analysis error cases. help promote transparency Its use assist editors peer reviewers, as well general readership, understand, interpret critically appraise design risk bias planned trial.

Язык: Английский

---