Newer generative AI tools in dentistry

With GAI, the possibilities for revolutionizing healthcare are endless, from dental care to drug discovery. Moreover, the only GAI that we are aware of are ChatGPT (1) and Google Gemini (2) which are transforming various industries.

GAI's capabilities are impressive, this article will discuss the GAI models available and their true potential which can be used in dentistry and healthcare.

1. StyleGAN:

Models like Generative Adversarial Networks (GAN) for treatment planning leverage specific data and disease progression to develop individualized treatment plans, minimizing side effects and maximizing efficacy. A study conducted at the University of Osaka, Japan suggests that StyleGAN2 (3) can generate and reconstruct panoramic radiographs and thus may be applied to the anonymization and data compression of medical images to craft personalized 3D models of teeth, allowing dentists to create custom-fit implants, bridges, and dentures that feel like an extension of oneself.

GAI's reach extends beyond mere aesthetics, as tools like DeepCare (4) analyze data to predict CBCT Segmentation which is a laborious task, but with GAI, this company has achieved CBCT segmentation for treatment planning in a mere fraction of seconds for diagnosing potential dental issues, empowering patients to take proactive measures to prevent future dental issues. Platforms like Smile Design AI (5) also generate personalized simulations of potential transformations, enabling informed decisions before the first cosmetic touch.

2. CycleGAN:

Data scarcity often obscures crucial details in medical images, hindering accurate diagnosis. GAI steps in as a skilled sculptor with tools like CycleGAN (6), which generates synthetic medical images, providing AI models with abundant training data, leading to more accurate diagnoses, especially in areas like cancer detection. GAI's Deep Denoising techniques allow for clearer, noise-free scans, enabling doctors to see finer details and make informed decisions. CycleGAN for sharper dental diagnostics: There is research that explores the potential of CycleGAN, a type of deep learning algorithm, to enhance the resolution of medical images, specifically chest X-rays. While the study focuses on lung pathologies (4), the underlying principles hold promise for significant advancements in dental diagnostics.

Here's how CycleGAN could transform dentistry:

a. Improved caries detection: By enhancing the clarity of dental X-rays, CycleGAN could enable dentists to identify early-stage cavities often missed by traditional methods, leading to earlier intervention and potentially preventing extensive procedures. Enhanced fracture visualization: Subtle fractures, particularly in complex root structures, can be challenging to diagnose on conventional X-rays. CycleGAN's ability to sharpen images could improve fracture detection, leading to more accurate diagnosis and treatment planning.

b. Reduced need for repeat imaging: With sharper images, dentists might require fewer repeat X-rays for confirmation or to monitor treatment progress, minimizing radiation exposure for patients. It's important to note that this is an early-stage exploration, and further research is needed to validate its efficacy in clinical settings. However, the potential benefits of CycleGAN for improved dental diagnostics and patient care are undeniable. This technology could pave the way for a future where precise diagnoses and minimally invasive treatments become the norm in dentistry. Personalized simulations are possible with GAI, such as realistic 3D models of organs incorporating unique data, enabling patient-specific surgical planning and treatment optimization – a significant leap forward in personalized medicine.

3. AlphaFold for drug discovery:

Drug discovery is a slow and arduous process, but GAI provides a portal to a faster, more efficient future. Tools like AlphaFold (7 ) design novel drug candidates with desired properties, potentially leading to the development of life-saving medications in record time. Models like Transformers analyze complex biological data, identifying potential drug targets, and offering new avenues for therapeutic development. Platforms like drug repurposing with generative models unlock new uses for existing drugs, optimizing resources and potentially expanding treatment options. AlphaFold is a new method for predicting protein structures with high accuracy even when no similar structure is known. This is a breakthrough in protein science, as knowing a protein's structure is crucial for understanding its function and developing drugs that target it.

AlphaFold in dentistry:

Many dental problems involve proteins, such as enzymes involved in tooth decay or proteins forming the building blocks of teeth and bones. By accurately predicting protein structures, AlphaFold could: Lead to the development of new drugs and therapies: For example, designing drugs that target specific enzymes involved in tooth decay or developing new materials for dental implants based on the structure of natural bone proteins. Improve diagnosis and treatment planning: By understanding the structure of proteins associated with certain dental diseases, dentists could diagnose them more accurately and personalize treatment plans. Aid in regenerative dentistry: By understanding how proteins interact to form teeth and bones, researchers could develop new methods for regenerating damaged tissues.

4. DeepCE:

GAI enables the delivery of personalized medicine like never before, with tools like DeepCE (8) using GAI for personalized genomics and disease risk prediction.

a. Predicting drug action:

DeepCE is an AI tool that uses a technique called "phenotype-based compound screening" and predicts how a drug might affect a cell based on its overall characteristics (phenotype) rather than just its chemical structure. By analyzing gene expression changes, DeepCE can identify existing drugs that could be repurposed for new uses, like treating COVID-19.

b. Drug discovery and repurposing:

DeepCE could be used to find existing drugs approved for other conditions that might be effective against oral diseases like periodontitis, caries, or oral cancer. This could save time and resources compared to developing entirely new drugs.

c. Personalized medicine:

By analyzing individual patient data, DeepCE could help predict which existing drugs might be most effective for their specific dental needs. Understanding disease mechanisms: DeepCE could be used to study how different drugs affect the gene expression of oral cells, providing insights into the mechanisms of dental diseases.

5. GAN-medEHR and Clinical trials:

Clinical trials are stepping stones to new drugs, but challenges like participant recruitment and lengthy processes often hinder progress. GAI offers a helping hand, with tools like GANs for synthetic Electronic Health Records (GAN-medEHR) (9) generating large-scale anonymized datasets for clinical trial simulations, potentially reducing the need for real-world participants, and accelerating the development process. Models like Bayesian Optimization for clinical trial design identify promising drug candidates and optimize trial protocols, leading to faster and more efficient development of effective therapies.

Conclusion:
The future: a balancing act of innovation and ethics

While GAI holds immense potential to transform healthcare, it is crucial to tread cautiously, ensuring responsible development and ethical implementation. Data privacy, bias, fairness, explainability and interpretability of models, and robust regulatory frameworks are crucial aspects that need careful consideration. By addressing these challenges responsibly, GAI can become the tool that unlocks a healthier future for all.

Remember, GAI is not a cure-all but a powerful tool that can revolutionize healthcare. Let us harness its technology responsibly, ensuring that this leads to a future where everyone has access to personalized, effective, and life-saving care.

References: 

1. T. Wu et al., "A Brief Overview of ChatGPT: The History, Status Quo and Potential Future Development," in IEEE/CAA Journal of Automatica Sinica, vol. 10, no. 5, pp. 1122-1136, May 2023, doi: 10.1109/JAS.2023.123618.
2. Saeidnia, Hamid Reza. (2023). Welcome to the Gemini era: Google DeepMind and the Information Industry. Library Hi Tech News. https://doi.org/10.1108/LHTN-12-2023-0214.
3. Kokomoto, K., Okawa, R., Nakano, K., & Nozaki, K. (2021). Intraoral image generation by progressive growing of generative adversarial network and evaluation of generated image quality by dentists. Scientific reports, 11(1), 18517. https://doi.org/10.1038/s41598-021-98043-3
4. DeepCare Diagnosis System. Offers second opinion for multi-modal dental image analysis. Dental AI 2.0: AI-Collaborative https://www.deepcare.com/solutions
5. Kurian, N., Sudharson, N. A., & Varghese, K. G. (2024). AI-driven smile designing. British dental journal, 236(3), 146. https://doi.org/10.1038/s41415-024-7087-3
6. Liang, Z., Huang, J. X., & Antani, S. (2022). Image Translation by Ad CycleGAN for COVID-19 X-Ray Images: A New Approach for Controllable GAN. Sensors (Basel, Switzerland), 22(24), 9628. https://doi.org/10.3390/s22249628
7. Jumper, J., Evans, R., Pritzel, A., Green, T., Et al (2021). Highly accurate protein structure prediction with AlphaFold. Nature, 596(7873), 583–589. https://doi.org/10.1038/s41586-021-03819-2
8. Novel ai tool takes a deepce dive into potential drugs for covid-19. Drug Target Review.
9. Baowaly, M. K., Lin, C. C., Liu, C. L., & Chen, K. T. (2019). Synthesizing electronic health records using improved generative adversarial networks. Journal of the American Medical Informatics Association : JAMIA, 26(3), 228–241. https://doi.org/10.1093/jamia/ocy142.

Tags: