Current role and future perspectives of artificial intelligence in echocardiography
Bibliographic record
Abstract
The history of early biomedical computing, including the first AI approaches, can be seen as a series of attempts to investigate, understand, and build computational models of the scientific knowledge and problem solving heuristics used by biomedical scientists, while also developing and testing computa- SummaryBackground: The rise of biomedical expert heuristic knowledgebased approaches for computational modeling and problem solving, for scientific inquiry and medical decision-making, and for consultation in the 1970's led to a major change in the paradigm that affected all of artificial intelligence (AI) research.Since then, AI has evolved, surviving several "winters", as it has oscillated between relying on expensive and hard-to-validate knowledge-based approaches, and the alternative of using machine learning methods for inferring classification rules from labelled datasets.In the past couple of decades, we are seeing a gradual but progressive intertwining of the two.Objectives: To give an overview of early directions in AI in medicine and threads of some subsequent developments motivated by the very different goals of scientific inquiry for biomedical research, and for computational modeling of clinical reasoning and more general healthcare problem solving from the perspective of today's "AI-Deep Learning Boom".To show how, from the beginning, AI was central to Biomedical and Health Informatics (BMHI), as a field investigating how to understand intelligent thinking in dealing professionally with the practice for healthcare, developing mathematical models, technology, and software tools to aid human experts in biomedicine, despite many previous bouts of "exuberant optimism" about the methodologies deployed.Methods: An overview and commentary on some of the early research and publications in AI in biomedicine, emphasizing the different approaches to the modeling of problems involved in clinical practice in contrast to those of biomedical science.A concluding reflection of a few current challenges and pitfalls of AI in some biomedical applications.Conclusion: While biomedical knowledge-based systems played a critical role in influencing AI in its early days, 50 years later they have taken a back seat behind "Deep Learning" which promises to discover knowledge structures for inference and prediction, both in science and for clinical decision-support.Early work on AI for medical consultation turned out to be more useful for explanation and teaching than for clinical practice, as had been originally intended.Today, despite the many reported successes of deep learning, fundamental scientific challenges arise in drawing on models of brain science, cognition, and language, if AI is to augment and complement rather than replace human judgment and expertise in biomedicine while also incorporating these advances for translational medicine.Understanding clinical phenotypes and how they relate to precision and personalization of care requires not only scientific inquiry, but also humanistic models of treatment that respond to patient and practitioner narrative exchanges, since it is the stories and insights of human experts which encourage what Norbert Weiner termed the ethical "human use of human beings", so central to adherence to the Hippocratic Oath
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".