No-Code and Low-Code Artificial Intelligence in Healthcare Imaging: A Scoping Review.
Authors
Affiliations (5)
Affiliations (5)
- Department of Surgical and Diagnostic Sciences, Marquette University School of Dentistry, Milwaukee, Wisconsin, USA; Department of Oral and Craniofacial Sciences, School of Dentistry, Oregon Health and Science University, Portland, OR, USA. Electronic address: [email protected].
- Technological Innovation Center, Depatrment of General Dental Sciences, Marquette University School of Dentistry, Milwaukee, Wisconsin, USA.
- Department of General Dentistry, Rīga Stradiņš University, Rīga, Latvia; Institute of Stomatology, Rīga Stradiņš University, Rīga, Latvia; Department of Conservative Dentistry, Periodontology and Digital Dentistry, LMU University Hospital, LMU Munich, Germany.
- Department of Conservative Dentistry, Periodontology and Digital Dentistry, LMU University Hospital, LMU Munich, Germany.
- Dean's Office, Marquette University School of Dentistry, Milwaukee, Wisconsin, USA.
Abstract
The aim of this review was to evaluate applications, methodological characteristics, limitations and future directions of no-code, low-code and automated machine learning (LCNC/AutoML) approaches in healthcare imaging. PubMed, Web of Science, and Google Scholar were searched using terms related to no-code artificial intelligence (AI), low-code AI, and AutoML combined with medical and dental imaging keywords. Studies were included if they applied LCNC or AutoML approaches to healthcare imaging tasks. Data extraction focused on platform type, imaging modality, clinical domain, task type and reported performance metrics. One-hundred thirty-eight studies met the inclusion criteria, including 127 medical and 11 dental imaging studies. Dental applications were limited to no-code platforms and focused on classification, object detection, and segmentation tasks, primarily involving caries detection, restoration identification and developmental staging. Medical imaging studies demonstrated broader adoption across specialties and imaging modalities, frequently employing AutoML or low-code frameworks for diagnostic and prognostic classification predominantly. Overall, LCNC/AutoML models showed promising performance, though external validation was inconsistently reported. LCNC/AutoML platforms might enable clinician-driven AI development for imaging tasks in dentistry and screening-focused medical applications. While these approaches lower technical barriers and may mitigate distribution shift through localised model training and application, limitations related to generalisability, regulatory compliance and task complexity persist. LCNC/AutoML tools are best positioned as complementary, context-specific solutions rather than replacements for traditional AI development. Accessible AI development platforms may facilitate broader clinician participation in imaging research and support the development of locally tailored diagnostic tools, particularly in resource-limited or emerging research environments such as dental imaging.