“The most reliable way to predict the future is to create it.”
Abraham Lincoln
We continue with the second part of our list of 2023 predictions for AI developments in the domain of clinical medicine and healthcare:
- Increasing medical legal discussions to involve aspects of artificial intelligence in healthcare
The dual possibilities in this domain are equally likely to occur in the very near future: failure to deploy an artificial intelligence tool that is widely adopted by health systems and that leads to a patient’s poor outcome, or utilization of an artificial intelligence tool that some observers feel may have led to a poor outcome due to some technical issue with the deployment of this artificial intelligence tool (such as a prediction sepsis tool failing to accurately predict as expected).
- More deployment of multimodal artificial intelligence in digital health with different data types
This strategy to increase the sophistication of artificial intelligence use in clinical care, especially chronic disease management, is accompanied by use of data outside the hospital setting such as wearable technology and sensor utilization. The real world data and experience will become increasingly important as more sources of data will be coming from outside the clinic and hospital settings to become part of the portfolio of data sources for real-world, real-time analysis.
- Decreasing the chasm of artificial intelligence in healthcare with innovative tools and models
Just as desktop computing became popular when it no longer required programming, AI in healthcare will accelerate its applications when these tools are much more flexible and user-friendly. Innovations such as MLOps (that renders ML more usable) and foundation models (which uses self-supervised learning and can learn from large amounts of unlabeled data) decrease the chasm.
- Incorporating artificial intelligence into all versions of extended reality tools in healthcare
Even with simple algorithms, artificial intelligence can add an important learning dimension to all modalities of extended reality (virtual, augmented, and hybrid). The future of medical education and clinical training will be incorporating more such tools. Most of the medical education and clinical training can be very well covered with “intelligent” reality, potentially with Metaverse as a part of this education and training portfolio.
- Exploring appropriate use of text-to-image (and also image-to-text) AI tools in healthcare
In the near future, anyone in healthcare can state “moderate-sized patent ductus arteriosus and echocardiogram with color Doppler” and this description will generate the appropriate image. Conversely, a medical image from radiology or pathology or a video taken during a procedure will have automated written descriptions. This interesting capability will be partly from a convergence of search and generative AI to create a text-to-image continuum.
- Deep learning tools for protein structure determination making impact in drug development
DeepMind’s AlphaFold has been impressive in bridging the genomic sequence-to-protein structure gap in determination of protein structure with expediency not seen before in this domain. The new knowledge from all the protein structural determinations will most likely lead to a surge of new understanding of mechanisms of disease as well as novel drug discoveries to treat diseases that may not have had adequate therapeutic options.
Overall, most if not all of these predictions are likely to occur if not this coming year, almost certainly by the end of this decade. The expediency that we see these developments will largely depend on investment momentum in technology and artificial intelligence as well as adoption potential of these tools by the medical community.
