Machine learning for predictive analytics in medicine: real opportunity or overblown hype?
Bibliographic record
Abstract
This editorial refers to ‘Predicting deterioration of ventricular function in patients with repaired tetralogy of Fallot using machine learning’ by M.D. Samad et al., pp. 730--738. Machine learning, a branch of computer sciences and an application of artificial intelligence (AI), is not new. As early as 1959, Samuel1 published in the IBM Journal the results of an experiment in which machine learning algorithms using rule-based systems to successfully learn the rules and strategy of checkers were developed that outperformed average players. However, despite early success and a relatively short existence, the field of AI has already experienced major cycles of disillusion, commonly known as AI winters. These winters have two things in common: astronomical expectations and subsequent failure to deliver anticipated results. Since the mid-1990s, AI has experienced renewed interest fuelled mainly by the increasing availability of computing power and the significant technological development and successful implementation of AI-based algorithms for decision support in many industries including banking, insurance, advertising, and transportation. The medical field shares this zeal, where interest in AI has skyrocketed following high-profile publications that showcase projects which have garnered significant media attention, and where the number of publications of AI-based projects in medical journals is steadily increasing. However, with this increased interest comes the risk of a new wave of overhyped expectations and a potential backlash should the technology fail to deliver; or deliver less than, or slower, than anticipated. And this risk is very real, in the 2017 Gartner Hype Cycle for Emerging Technologies deep learning was at the peak of inflated expectations,2 replacing the 2016 ‘winner’: machine learning.3 The risk of such backlash is particularly high in medicine given several factors. At the outset, there are astronomical, likely overinflated, expectations in regards to the current capabilities of AI algorithms. Additionally, there is a general lack of recognition of the technological and infrastructure gap between what AI algorithms need for successful implementation and the information technology infrastructure available in the majority of medical settings. Finally, and most importantly, there is an incomplete understanding on how to actually integrate the information generated by AI algorithms in the evidence-based paradigm of medicine. Recently, the high-profile failure of large AI implementation projects in various medical systems, demonstrate many of these hallmarks. And while some disillusion in regards to the potential of AI to transform medicine is probably inevitable; we must realign expectations, work diligently to bridge these gaps and most importantly rapidly integrate viable projects in health systems to show real-world feasibility and value. All of this will be necessary in order to prevent the same backlash that has previously vanquished so many promising technologies in the medical system.
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 | 0.017 | 0.082 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.002 | 0.009 |
| Scholarly communication | 0.007 | 0.013 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.032 | 0.051 |
| Insufficient payload (model declined to judge) | 0.011 | 0.007 |
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".