“When did Noah build the ark? Long before it began to rain.”
This is how longevity magnate Dr. Peter Attia introduces his audience to the value of preventive care in his much-lauded book, “Outlive: The Science and Art of Longevity”[1]. In the book, he presents his transformative vision of healthcare — Medicine 3.0. This ambitious approach transcends the inefficiencies of Medicine 2.0, shifting the paradigm from merely treating diseases to optimizing overall healthspan.
A core principle of this new era is preventive care, an area where AI is making significant strides. This evolution in healthcare, powered by AI, is not just changing the game; it’s setting a whole new playing field.
The AI Healthcare Takeover
AI’s exponential growth within healthcare is far from surprising. Like in most other industries, this AI takeover owes to two irresistible benefits: tremendous cost savings and disruptive use cases.
On the cost-cutting front, it’s been recently estimated that AI could save the US annually up to a staggering $360 billion in healthcare costs; 35% of which would be administrative [2].
Automation of such administrative tasks, which take physicians an average of 15.6 hours a week [3], remains one of AI’s most powerful applications in the industry. Additionally, AI has been innovatively used in auditing prescriptions to improve the safety and efficacy of treatment regimens, supporting clinical decision-making, and enhancing access to care and knowledge in underserved communities with tools like chatbots [4].
While all these innovations are groundbreaking in their own right, AI’s true potential arguably lies in pushing the frontiers of preventive care forward through disease prediction.
Disease Prediction: A Step Beyond Diagnosis
To achieve this vision of prolific preventive care, relying solely on diagnosis may not cut it anymore. Traditional diagnosis often involves waiting for symptoms to manifest before identifying a disease. However, this approach can lead to late diagnoses. According to the report “Improving Diagnosis in Health Care” by the Institute of Medicine, 12 million American adults experience late diagnosis every year [5]. Depending on the individual, the consequences of this one late diagnosis can be too costly.
That’s why we need disease prediction.
Disease prediction forecasts the likelihood of a disease through various modeling techniques before symptoms even surface [6]. This shift from reactive diagnosis to proactive disease prediction is a key aspect of modern preventive care, aiming to address minor health disturbances before they develop into worrisome issues as well as to personalize prevention strategies.
In this quest, AI emerges as a backbone of accurate, feasible prediction. Disease prediction relies on AI’s unmatched capabilities to scour vast databases and recognize actionably relevant patterns and relationships [7]. To put that into perspective, that means AI can take a patient’s entire medical records and translate them into a single score that determines an individual’s risk of developing a disease and guides physicians in deciding the most appropriate course of action.
The Data AI Feeds on to Predict Disease
The concept of disease prediction went mainstream last year when actor Chris Hemsworth announced a hiatus from acting due to a heightened risk of Alzheimer’s disease [8]. That risk is due to his inheritance of two copies of the APOE ε4 allele, which is associated with a higher probability of developing Alzheimer’s [9].
That’s an example of genetic disease prediction. AI can analyze genomic data, but that’s just the tip of the iceberg. AI also utilizes other ‘omic’ data, which includes proteomics, transcriptomics, metabolomics, and more. All these ‘omics’ represent different types of molecular data that provide a comprehensive view of our biological systems [10].
In addition to ‘omic’ data, AI leverages electronic health records (EHRs), which contain a wealth of information about a patient’s medical history [11]. This data is generated through various sources, including medical imaging, lab tests, and patient interactions.
AI Techniques for Disease Prediction
According to the type of data, several AI techniques can be used to predict disease, including Machine Learning (ML), Deep Learning (DL), and Natural Language Processing (NLP).
ML refers to models that build their knowledge from data and autonomously make predictions and form relationships without being directly programmed to do so. In the context of disease prediction, an ML model could be trained on a dataset of patient information where the presence or absence of the disease is known. The model would learn the relationship between the patient information and the disease status, allowing it to predict the disease status for new patients [12].
DL employs artificial neural networks with multiple layers. These layers enable the model to comprehend and abstract sophisticated features from data similar to the way human brains do. In disease prediction, DL can be particularly useful for tasks like image analysis. For example, a DL model could be trained to identify tumors in MRI scans with high accuracy and performance [13].
NLP facilitates computer-human language interaction, enabling systems to comprehend, interpret, and generate human language adeptly. In healthcare, NLP can analyze patient records, medical literature, and clinical guidelines to support disease prediction. For instance, NLP can extract symptoms from a patient’s speech or text input, aiding in predicting the most probable disease [14].
AI Disease Prediction Breakthroughs
With the breadth of data it can analyze and the versatile techniques it employs, AI has consistently led the way in breakthroughs for disease prediction. That includes forecasting age-related diseases such as heart disease, cancer, and neurodegenerative conditions, which are primary focuses in longevity research. Now, let’s delve into some success stories:
Cancer
AI has shown remarkable efficacy in predicting various types of cancer, including breast, brain, lung, liver, and prostate cancer, even demonstrating greater accuracy in predicting cancer than clinicians [15].
A notable example is a DL model that effectively detected individuals with the greatest likelihood of developing pancreatic cancer, as much as three years prior to their official diagnosis, relying solely on the medical records of patients. This breakthrough could potentially intercept one of the deadliest cancers in the world before it develops symptoms [16].
The researchers behind this study state that their model is at least on par with current genetic sequencing tests. Yet, AI prediction is way more accessible to broader populations, including in regions with limited access to advanced medical technologies.
Heart Disease
In a landmark clinical trial led at the Smidt Heart Institute and Cedars-Sinai Division of AI in Medicine, AI demonstrated superior performance in assessing cardiac function compared to human sonographers [17].
The trial was blinded and randomized, with cardiologists evaluating the accuracy of left ventricular ejection fraction measurements without knowing their origin. Notably, the cardiologists could not differentiate between AI and human assessments and were less likely to adjust the AI-generated measurements.
Alzheimer’s Disease
Typically, Alzheimer’s disease is typically identified only after symptoms have manifested, at which point considerable brain damage has already occurred. However, certain Alzheimer’s risk factors are known, such as lifestyle habits, certain medications, and health conditions like obesity, hypertension, and high cholesterol. Hiding in plain sight, these factors are consistently documented in EHRs.
Recognizing the potential of this data, a team of researchers from the University of Florida Health and College of Medicine embarked on a pioneering study. Utilizing AI, they were able to forecast the onset of Alzheimer’s up to five years prior to an official diagnosis based on the analysis of routinely collected EHR data [18].
Depression
In an exceptionally intriguing application of AI for disease prediction, researchers have utilized NLP to analyze language used on Facebook. The goal was to predict if and when an individual would receive a first diagnosis of depression, as recorded in their EHRs [19].
Impressively, the model achieved a level of accuracy that could be clinically useful. What makes this approach particularly exciting is its non-invasive nature. AI can analyze autobiographical texts of patients (with their consent), such as social media posts, without requiring any additional time or effort from doctors or patients. The study found that it might be possible to predict future depression status as early as three months before it is first documented in the medical record.
Conclusion
AI’s ability to analyze vast amounts of data and predict diseases before they manifest is revolutionizing preventive care. Moreover, AI is decentralizing research, breaking down barriers, and democratizing access to vital health information.
On that specific front, Rejuve.AI is leading the way!
We are building the first-ever AI-powered decentralized longevity research network. By inviting researchers and health enthusiasts from all over the world to contribute, we are harnessing the collective intelligence of the global community. This collaboration is amplified via multiple cutting-edge AI models to tackle the complex challenges of longevity.
References:
[1] Attia, Peter, and Bill Gifford. Outlive. Vermilion, 28 Mar. 2023.
[2] Cutler, David M. “What Artificial Intelligence Means for Health Care.” JAMA Health Forum, vol. 4, no. 7, 6 July 2023, p. e232652, jamanetwork.com/journals/jama-health-forum/fullarticle/2807176, https://doi.org/10.1001/jamahealthforum.2023.2652..
[3] Medscape Physician Compensation Report 2021: The Recovery Begins, 2021, www.medscape.com/slideshow/2021-compensation-overview-6013761#21
[4] Secinaro, Silvana, et al. “The Role of Artificial Intelligence in Healthcare: A Structured Literature Review.” BMC Medical Informatics and Decision Making, vol. 21, no. 1, 10 Apr. 2021, link.springer.com/article/10.1186/s12911–021–01488–9, https://doi.org/10.1186/s12911-021-01488-9.
[5] National Academies of Sciences, Engineering. Improving Diagnosis in Health Care. Nap.nationalacademies.org, 22 Sept. 2015, nap.nationalacademies.org/catalog/21794/improving-diagnosis-in-health-care
[6] Grampurohit, Sneha , and Chetan Sagarnal. “Disease Prediction Using Machine Learning Algorithms | IEEE Conference Publication | IEEE Xplore.” Ieeexplore.ieee.org, 3 Aug. 2020, ieeexplore.ieee.org/document/9154130/
[7] Bajwa, Junaid, et al. “Artificial Intelligence in Healthcare: Transforming the Practice of Medicine.” Future Healthc J, vol. 8, no. 2, 2021, pp. e188–e194, www.rcpjournals.org/content/futurehosp/8/2/e188, https://doi.org/10.7861/fhj.2021-0095
[8] Burke, Kelly. “Chris Hemsworth to Take “Time Off” from Acting after Discovering Alzheimer’s Risk.” The Guardian, 21 Nov. 2022, www.theguardian.com/film/2022/nov/21/chris-hemsworth-to-take-time-off-from-acting-after-discovering-alzheimers-risk. Accessed 13 Nov. 2023.
[9] “Alzheimer’s Disease Genetics Fact Sheet.” National Institute on Aging, www.nia.nih.gov/health/genetics-and-family-history/alzheimers-disease-genetics-fact-sheet.
[10] Cheng, Hao, et al. “Artificial Intelligence-Based Omics Data Analysis | Frontiers Research Topic.” Www.frontiersin.org, www.frontiersin.org/research-topics/34606/artificial-intelligence-based-omics-data-analysis.
[11] Wu, PY, et al. “–Omic and Electronic Health Record Big Data Analytics for Precision Medicine.” IEEE Transactions on Biomedical Engineering, vol. 64, no. 2, Feb. 2017, pp. 263–273, https://doi.org/10.1109/tbme.2016.2573285..
[12] Ghaffar Nia, Nafiseh, et al. “Evaluation of Artificial Intelligence Techniques in Disease Diagnosis and Prediction.” Discover Artificial Intelligence, vol. 3, no. 1, 30 Jan. 2023, https://doi.org/10.1007/s44163-023-00049-5.
[13] Yu, Zengchen, et al. “Popular Deep Learning Algorithms for Disease Prediction: A Review.” Cluster Computing, 13 Sept. 2022, https://doi.org/10.1007/s10586-022-03707-y.
[14] Kumar, Rahul, et al. “Disease Prediction from Speech Using Natural Language Processing and Deep Learning Method.” Advances in Intelligent Systems and Computing, 2021, pp. 407–413, https://doi.org/10.1007/978-981-33-6984-9_33.
[15] Zhang, Bo, et al. “Machine Learning and AI in Cancer Prognosis, Prediction, and Treatment Selection: A Critical Approach.” Machine Learning and AI in Cancer Prognosis, Prediction, and Treatment Selection: A Critical Approach, vol. Volume 16, 1 June 2023, pp. 1779–1791, https://doi.org/10.2147/jmdh.s410301.
[16] Placido, Davide, et al. “A Deep Learning Algorithm to Predict Risk of Pancreatic Cancer from Disease Trajectories.” Nature Medicine, 8 May 2023, pp. 1–10, www.nature.com/articles/s41591-023-02332-5, https://doi.org/10.1038/s41591-023-02332-5.
[17] He, Bryan, et al. “Blinded, Randomized Trial of Sonographer versus AI Cardiac Function Assessment.” Nature, 5 Apr. 2023, https://doi.org/10.1038/s41586-023-05947-3.
[18] Li, Qian, et al. “Early Prediction of Alzheimer’s Disease and Related Dementias Using Real‐World Electronic Health Records.” Alzheimer’s & Dementia, 23 Feb. 2023, https://doi.org/10.1002/alz.12967.
[19] Eichstaedt, Johannes C., et al. “Facebook Language Predicts Depression in Medical Records.” Proceedings of the National Academy of Sciences, vol. 115, no. 44, 15 Oct. 2018, pp. 11203–11208, www.pnas.org/content/115/44/11203, https://doi.org/10.1073/pnas.1802331115.