With artificial intelligence, it is already possible to predict the risk of developing a disease even before it appears. Here are three cases in which good prediction results have been achieved:
- Fractures due to osteoporosis
- Cancer associated with gene mutations
- Alzheimer's disease
The researchers developed an algorithm, called Crystal Bone , using retrospective data from the electronic medical records of more than 1,000,000 patients. Crystal Bone applies machine learning techniques, in particular Natural Language Processing (NLP), and predicts whether a patient is likely to experience a fracture event, or whether the patient's course is similar to that of a typical patient who has suffered a future fracture caused by their condition.
The model accurately predicted fracture risk 1 to 2 years in advance for patients over 50 years of age.
The Bayes Mendel Lab at the Dana-Farber Cancer Institute and the Hughes Lab at the Massachusetts General Hospital have developed a tool to help clinicians and genetic counselors better interpret the results of cancer-related genetic studies, understand the risk of various cancers, and thus define a better strategy in preventive disease management .
In fact, there are more and more studies on specific genes, and it is not possible for clinicians and genetic counselors to always be aware of all these publications, nor to appreciate the relative accuracy and importance of each one. Thanks to artificial intelligence, again thanks to NLP algorithms, researchers are able to perform a broader review of the medical literature and identify reliable studies more easily and quickly. By doing so, they are able to obtain estimates of absolute risk, i.e. the probability of a health effect occurring in certain situations.
Researchers at IBM , also using NLP algorithms, were able to detect subtle differences in language in subjects who later developed Alzheimer's disease. In fact, they were more repetitive in their use of words even before developing the disease.
This artificial intelligence program predicted, with 75 % accuracy, who would have Alzheimer's disease.
The hope is to extend the work on Alzheimer's to find subtle changes in language use by people with no obvious symptoms, but who will go on to develop other neurological diseases, so that we can intervene early.
 Almog, Yasmeen Adar et al. “Deep Learning With Electronic Health Records for Short-Term Fracture Risk Identification: Crystal Bone Algorithm Development and Validation.” Journal of medical Internet research vol. 22,10 e22550. 16 Oct. 2020, doi:10.2196/22550
 Braun, D., Yang, J., Griffin, M. et al. “A Clinical Decision Support Tool to Predict Cancer Risk for Commonly Tested Cancer-Related Germline Mutations.” J Genet Counsel 27, 1187–1199 (2018). https://doi.org/10.1007/s10897-018-0238-4
 Elif Eyigoz,Sachin Mathur,Mar Santamaria,Guillermo Cecchi,Melissa Naylor et al. “Linguistic markers predict onset of Alzheimer's disease.” EClinicalMedicine, 22 October 2020, doi:https://doi.org/10.1016/j.eclinm.2020.100583
Laws and ethics labels do not solve the issue of trust, as a coping strategy in the face of risk and uncertainty. However, the anthropomorphic terms 'trust' and 'reliability' in the context of AI regulation, can be welcomed as they allow everyone to understand the main reasons for building and using these complex technological objects and their impact on society.
In the midst of a whirlwind of information about Atificial Intelligence (AI), it helps to know concrete examples of what is currently implemented to identify possible opportunities for your company. We presented some practical examples at the Ticino Chamber of Commerce.