During the last few years, there has been a lot of noise surrounding both modern medicine and artificial intelligence (AI). Across social media platforms, you will see users sharing misguided opinions on both of these topics, which has driven a great deal of fear and misunderstanding across our greater society. Though in reality, both modern medicine and AI are our greatest chance at solving our world’s complex health challenges. 

Modern medicine has become a powerful tool against some of the most deadly diseases. With the world’s best and brightest minds, we have developed vaccines, and therapeutics that have greatly improved our quality of life. Modern-day efforts aim to incorporate AI into medicinal treatments to help connect challenging aspects of the bigger puzzle. Most recently, this approach has been tested in cancer treatment [1]. As cancer is a rapidly increasing disease across the world, there are large amounts of data that have been collected by physicians in an effort to find a cure. There is no one-size-fits-all treatment for cancer, though with AI, a tool called machine learning can sort through these massive datasets to help build a model that can ideally one day find the most suitable treatment for a particular patient [1].

Despite endless stores of knowledge and advances in technology, mental health treatment still remains a challenging puzzle even in the 21st century. The human brain is highly powerful and any disruption to its healthy balance of chemical signals results in largely misunderstood mental health conditions. In terms of medical treatment, patients are prescribed a cocktail of drugs by their psychiatrists in an effort to return homeostasis to the brain. Similar to unpredictable physical diseases like cancer, there is no one-size-fits-all treatment for mental health. Therefore, psychiatric medicine would largely benefit from an AI-driven approach to help solve more pieces in this challenging health puzzle. 

Our World’s Most Challenging Puzzle: The Human Brain

Our brain, though only about 2% of our body weight, is stacked with over 80 billion neurons [2]. It is the controller of all aspects of our personal experience with the world, from how we think, feel, and move. The brain is driven by neurotransmitters – a series of chemicals  that maintain its health, particularly serotonin and dopamine [3]. When the brain experiences disruption to its ecosystem, either too much or too little of a particular neurotransmitter can result in mental health conditions that change the way a person thinks, feels, and moves. The field of psychiatry aims to address these chemical imbalances with psychiatric drugs that aim to regulate the levels of neurotransmitters within a patient’s brain. Though some patients respond well, the process of identifying a suitable psychiatric drug for an individual can be lengthy and involve a great deal of trial and error [4]. Modern medicine aims to address individual differences to treatment using the biopsychosocial model [5]. This model acknowledges that each person experiences a unique set of external factors that can influence their response to a certain treatment, including pre-existing health conditions, and environmental and social factors [3]. Global clinicians and scientists have painstakingly attempted to collate this information and piece together therapies to solve these complex medical puzzles. But, at this current pace, do they stand the chance of beating the rapid increase in mental health conditions?

The Mystery of Mental Health Biomarkers

To identify diseases and disorders, there are certain tell-tale signs referred to as biomarkers that can help accelerate a diagnosis. A biomarker is defined as a specific biological property that serves as a sign of a condition or disease [6]. These biomarkers can be collected from a patient’s blood, saliva, or other tissue samples to identify and predict abnormalities that can confirm the presence of disease [6]. For example, higher levels of a biomarker referred to as cancer antigen 125 or CA 125 on a blood test may alert a physician to investigate a person for cancer [6]. However, unlike physical illnesses, there is limited knowledge surrounding how mental health conditions can be identified or treated. 

Alternatively, genetic biomarkers can help doctors to predict how likely their patient is to develop a psychiatric condition based on their family inheritance. Though in order to collect appropriate tissue for a diagnosis for brain disease, modern practices require the collection of cerebrospinal fluid from a painful spinal tap or the investigation of cadaver brains [7]. The use of cadaver brains present challenges for neuroscience, as deceased brain tissue does not provide the same physiological measurements as living brain tissue [8].

Since the invention of non-invasive imaging technology, clinicians are provided with unique insight into the structure and function of the brain in a living patient. Currently, the most advanced imaging technique is magnetic resonance imaging (MRI). Typically, doctors will ask a patient to seek an MRI scan of their brain if they are suspicious of a traumatic brain injury or tumor. Though more recently, MRI has also been suggested as a diagnostic tool for the field of psychiatry. Recent efforts have aimed to identify well-documented brain abnormalities within a particular mental health condition using MRI to compare and contrast with normal brain scans [9]. However, the size and complexity of MRI datasets have made it challenging for clinicians to identify specific biomarker fingerprints to determine mental health conditions. This leads us to wonder, does AI have the power to successfully manage and analyze these massive datasets to solve these mental health mysteries? 

Solving the Puzzle with AI

There has been rising interest in the use of AI as a tool to make meaningful comparisons between biomedical imaging datasets, psychological and social factors that drive a particular mental health condition. Advances in research place AI technology in the unique position to be highly successful at predicting and diagnosing mental illness. Therefore, there have been efforts to develop a machine learning AI model that can determine which data points will be most suitable and effective for each condition. 

A machine learning AI model is driven by sets of data and operates like a human brain by sorting through this information to find the most important and relevant information for a particular question. An AI model can be trained to flag the most relevant information to potentially identify a specific and effective drug. As research and clinical data datasets grow, AI models will become more accurate and increase the efficiency of diagnosis and treatment [10]. The use of an AI model can allow clinicians to view the most relevant information of each MRI dataset and build a more consistent framework of how a particular mental illness impacts the brain’s structures. 

These systems have already proved useful in developing more targeted psychiatric drugs for schizophrenia and depressive disorders [11]. Recently, a study from a team at the University of Texas investigated the diagnosis of bipolar disorder, finding that the traditional symptom-based approach based upon the Diagnosis and Statistical Manual for Mental Disorders was able to diagnose a patient with bipolar disorder at a success rate of 71% [10]. Though on the other hand, a machine learning model built from neuroimaging and neurocognitive data was able to diagnose a patient with bipolar disorder with up to 94% accuracy [11]. In addition, a team of computer scientists and neuroscientists were also able to successfully diagnose depression with 85% based on MRI imaging datasets [12].

 Ethical Dilemma

As our understanding of the brain and AI models become more powerful, the more opportunities we will have to optimize treatment for mental illnesses. Though this approach has the ability to be highly useful in the field of psychiatry, there are some valid ethical and safety concerns. The Health Insurance Portability and Accountability Act (HIPAA) was established in the United States to protect patient health data. There is concern that the use of clinical mental health data with AI models will breach HIPAA regulations, due to the highly sensitive nature of these data points and emerging data breaches [13]. Furthermore, there is concern that specific people working on these AI models may hold biases for certain mental health conditions which may impact the overall accuracy of the data it will collect [14]. This has encouraged the incorporation of AI fundamentals delivered through Continuing Medical Education programs to educate medical professionals on ethical and effective clinical practice [15]. Beyond educating medical professionals, there have been efforts within the US FDA to establish regulations for using AI in medical treatment [15]. Overall, with the appropriate training and oversight, the benefits of AI in psychiatric medicine will be revolutionary mental healthcare. 

References

1.     Liao J, Li X, Gan Y, Han S, Rong P, Wang W, et al. Artificial intelligence assists precision medicine in cancer treatment. Front Oncol [Internet]. 2023 [cited 2023 Mar 29];12. Available from: https://www.frontiersin.org/articles/10.3389/fonc.2022.998222
2.     Lisman J. The Challenge of Understanding the Brain: Where We Stand in 2015. Neuron. 2015 May 20;86(4):864–82.
3.     Koppe G, Meyer-Lindenberg A, Durstewitz D. Deep learning for small and big data in psychiatry. Neuropsychopharmacology. 2021 Jan;46(1):176–90.
4.     Middleton H, Moncrieff J. Critical psychiatry: a brief overview. BJPsych Adv. 2019 Jan;25(1):47–54.
5. 5. Xiao X, Song H, Sang T, Wu Z, Xie Y, Yang Q. Analysis of real-world implementation of the Biopsychosocial Approach to healthcare: Evidence from a combination of qualitative and quantitative methods [Internet]. Frontiers; 2021 [cited 2023 May 13]. Available from: https://www.frontiersin.org/articles/10.3389/fpsyt.2021.725596/full
6.     Henry NL, Hayes DF. Cancer biomarkers. Mol Oncol. 2012 Apr;6(2):140–6.
7.     García-Gutiérrez MS, Navarrete F, Sala F, Gasparyan A, Austrich-Olivares A, Manzanares J. Biomarkers in Psychiatry: Concept, Definition, Types and Relevance to the Clinical Reality. Front Psychiatry [Internet]. 2020 [cited 2023 Mar 29];11. Available from: https://www.frontiersin.org/articles/10.3389/fpsyt.2020.00432
8.     Opitz A, Falchier A, Linn GS, Milham MP, Schroeder CE. Limitations of ex vivo measurements for in vivo neuroscience. Proc Natl Acad Sci. 2017 May 16;114(20):5243–6.
9.     Falkai P, Schmitt A, Andreasen N. Forty years of structural brain imaging in mental disorders: is it clinically useful or not? Dialogues Clin Neurosci. 2018 Sep;20(3):179–86.
10.     Wu MJ, Passos IC, Bauer IE, Lavagnino L, Cao B, Zunta-Soares GB, et al. Individualized identification of euthymic bipolar disorder using the Cambridge Neuropsychological Test Automated Battery (CANTAB) and machine learning. J Affect Disord. 2016 Mar 1;192:219–25.
11.   Wu MJ, Mwangi B, Bauer IE, Passos IC, Sanches M, Zunta-Soares GB, et al. Identification and individualized prediction of clinical phenotypes in bipolar disorders using neurocognitive data, neuroimaging scans and machine learning. NeuroImage. 2017 Jan 15;145:254–64.
12.   Rosa MJ, Portugal L, Hahn T, Fallgatter AJ, Garrido MI, Shawe-Taylor J, et al. Sparse network-based models for patient classification using fMRI. Neuroimage. 2015 Jan 15;105:493–506.
13.   Journal H. HIPAA, Healthcare Data, and Artificial Intelligence [Internet]. HIPAA Journal. 2022 [cited 2023 Apr 23]. Available from: https://www.hipaajournal.com/hipaa-healthcare-data-and-artificial-intelligence/
14.   Naik N, Hameed BMZ, Shetty DK, Swain D, Shah M, Paul R, et al. Legal and Ethical Consideration in Artificial Intelligence in Healthcare: Who Takes Responsibility? Front Surg [Internet]. 2022 [cited 2023 Apr 1];9. Available from: https://www.frontiersin.org/articles/10.3389/fsurg.2022.862322
15.   Paranjape K, Schinkel M, Nannan Panday R, Car J, Nanayakkara P. Introducing Artificial Intelligence Training in Medical Education. JMIR Med Educ. 2019 Dec 3;5(2):e16048.