Learning with less: A survey of deep learning in medical imaging under varying supervision levels
Clinical Snapshot
PICO Framework
| P — Population | Medical imaging datasets and deep learning models in healthcare applications |
| I — Intervention | Deep learning methods with varying supervision levels (no supervision, inexact, incomplete, inaccurate, limited supervision) |
| C — Comparator | Fully supervised deep learning approaches |
| O — Outcomes | Model performance, annotation efficiency, clinical applicability of weakly supervised methods |
Bottom Line
This systematic review addresses the critical challenge of developing effective deep learning models for medical imaging when comprehensive annotations are unavailable. The authors propose a useful categorization framework spanning five supervision levels, from unsupervised to limited supervision approaches. While the framework appears comprehensive and clinically relevant, the abstract provides insufficient detail about methodology, quality assessment, and quantitative findings. The review's value lies in organizing the rapidly evolving field of weakly supervised medical AI, potentially guiding clinicians and researchers toward appropriate methods for their annotation constraints. However, without access to the full methodology and results, it's difficult to assess the review's reliability or provide specific clinical recommendations. The work addresses an important barrier to AI implementation in healthcare settings with limited annotation resources.
Key Findings
Effect Size: Not applicable - narrative review
Primary Outcome: Categorization of deep learning methods by supervision level
Nnt Or Sensitivity: Framework identifies five main supervision categories: no supervision, inexact, incomplete, inaccurate, and limited supervision
Confidence Interval: Not reported
Clinical Application
High relevance for resource-constrained settings where full annotation is impractical Relevant to TGA software as medical device regulations and potential PBS reimbursement considerations for AI-assisted diagnostic tools Healthcare institutions implementing AI-assisted medical imaging with limited annotation resources
Abstract
The rapid evolution of deep learning has significantly advanced the field of medical image analysis. However, despite these achievements, further progress is hindered by the scarcity of large, well-annotated datasets. To overcome this limitation, recent years have seen a growing focus on developing methods that leverage varying levels of supervision as alternatives to fully supervised learning. Instead of relying solely on complete annotations, these approaches employ partial annotations or other supervision strategies to train deep learning models. This paper presents a systematic review of deep learning methods for medical image analysis under different levels of supervision. To this end, we categorize these methods based on the level of supervision they rely on, encompassing categories such as no supervision, inexact supervision, incomplete supervision, inaccurate supervision, and only limited supervision. We further divide these categories into finer subcategories. For example, we categorize inexact supervision into multiple instance learning and learning with weak annotations. Similarly, we categorize incomplete supervision into semi-supervised learning, active learning, and domain-adaptive learning and so on. Furthermore, we systematically summarize commonly used datasets for different types of supervision in deep learning for medical image analysis and explore future research directions to conclude this survey.
References
- 1.Kumari, S., & Singh, P. (2026). Learning with less: A survey of deep learning in medical imaging under varying supervision levels. Artificial Intelligence in Medicine, 103375. https://doi.org/10.1016/j.artmed.2026.103375
Related Research
Journal of medical Internet research
Magnetic Resonance Imaging-Based Artificial Intelligence in Predicting Prostate Cancer Biochemical Recurrence: Systematic Review and Meta-Analysis
8 July 2026
Korean journal of radiology
Agentic Artificial Intelligence for the Automated Generation of Accurate Summary Podcasts of Radiology Research Papers
3 July 2026
Medicine
Artificial intelligence-guided clavicle fracture detection on plain radiographs: A retrospective diagnostic accuracy study.
28 June 2026
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service