AI: The Algorithmic Doctor
Remember March 2020, that difficult time when you were quarantined and so was your mind? Remember how the time you got to spend with yourself somewhere or the other became the reason for your anxiety? Well, the same was the case with many other people. And in such a scenario, imagine technology coming to your service and saving you. Yes, right from simple teleconsultation model using video conferencing, telephone, online chatrooms, to amazingly advanced asynchronous and synchronous models using AI, all of these have contributed to the development of psychiatry, and today we have all these options for us. In the previous blog, we have seen the scope of using AI as a mental health doctor, its need, and its applications. Now let’s try to understand what exactly is so magical about AI that makes us call it the future of psychiatry.
AI in psychiatry encompasses the use of techniques such as automated language processing and machine learning algorithms to assess a patient’s mental state. Machine learning algorithms have also aided in reducing the false-negative and false-positive diagnoses by supplementing human clinical ratings. Talking particularly about algorithms, we realize that it is these smart algorithms that support clinicians with early detection and diagnostics, with the flagging of suicide risks. These algorithms are also designed to help patients manage their condition through counseling and constantly being there for them.
The functioning of AI algorithms is largely based on the data fed to them. The process of classification and clustering of data is the first and foremost step for the algorithms to work which is followed by finding patterns in the huge piles of datasets. The data may be digital such as images, written or spoken texts. As treatment for mental health issues is majorly based on language interaction, analyzing the use of speech data is a viable way forward. This digital data may be collected from social media posts, therapy sessions, or survey responses.
On the other hand, the physical forms of data like saliva and blood samples can also be added to the equation and be studied more closely by A.I. algorithms. Pearl Chiu at the Human Neuroimaging Laboratory believes that feeding machine learning algorithms with the physical aspects will help us treat mental disorders like any other disease and help overcome the stigma around them.
Some most commonly used algorithms include the Bayesian model, Logistic regression, Decision tree, Support vector machines, Deep learning.
Bayesian model — Also known as a classification algorithm. It is based on Bayes’ theorem and characteristic condition-independent hypothesis. Recently it is employed to treat psychiatric disorders.
The Strüngmann Forum on Computational Psychiatry used Bayesian inference to connect underlying causes (genetics and sociological phenomena), hypothesized theoretical constructs, and symptoms.
Furthermore, Grove et al. tried to explore the relationship between visual integration and general cognition. The Bayesian model was used for comparison of the disease categorization systems and has common psychopathological information from diagnostic groups.
Logistic regression — LR is an important statistical-based AI algorithm that often employs LR models to diagnose psychiatric disorders. The accuracy of LR models is so high that they are commonly applied in clinical practice. Shen et al. generated a risk stratification model to obtain the odds ratio (OR) of psychiatric comorbidities by a classification and regression tree method. Using the LR method, the OR of psychiatric comorbidities was calculated between subjects with and without borderline personality disorder.
Decision tree-It is a flowchart-like diagram that shows the various outcomes from a series of decisions, including chance event outcomes and utility. In AI, a decision tree is a predictive model that represents a mapping between object properties and object values. With a decision tree, Sattler et al. analyzed data from the Spence Children’s Anxiety Scale (SCAS) and worked out two screening algorithms to diagnose obsessive-compulsive disorder from a combined clinical and community sample of children and families.
Support vector machines-The SVM is a current supervised learning method, the decision boundary of which is the maximum margin hyperplane for solving learning samples models that have been commonly used for diagnosing psychiatric disorders. Peng et al. employed a multi-kernel SVM-based model to locate potential users who might suffer from depression by extracting three social methods (user microblog text, user profile, and user behaviors).
Deep learning-At present, deep learning outperforms the aforementioned AI models by a considerable margin. Deep learning refers to a set of algorithms on a multi-layer neural network that uses various machine learning algorithms to solve various problems such as images and text.
Heinsfeld et al. applied deep learning algorithms on a large brain imaging dataset to identify patients with autism spectrum disorder based solely on the patients’ brain activation patterns. The findings revealed that 70% accuracy was achieved in the dataset, and that deep learning methods can classify large datasets better than other methods. Furthermore, the results showed the promise of deep learning for clinical datasets and illustrated the future application of AI in the identification of mental disorders.
Discussing all these algorithms we see that the analytic performance of deep learning for diagnosing psychiatric disorders is better, however, we still need to overcome problems like higher requirements for computer configurations; and data quantity for accuracy, also the time being consumed by experiments. These problems are worthy of further study and discussion in the future.
In short, although AI has made great progress in diagnosing psychiatric disorders, there are still many research areas for the improvement of AI-based applications. Such as advancing classic shallow learning algorithms which will make it easy to share and use information among high-dimensional features, such as, incorporating unsupervised learning to perform automatic annotation for unlabelled psychiatric disorder imaging data. Since the current AI-based model can only process homologous datasets, its generalizability is insufficient. Therefore, migration learning, multi-view learning, and ensemble learning will be used to process big psychiatric disorder data in the distant future.
It is evident that AI can act as a boon in the diagnosis and treatment of mental health, and advancements in this field are certainly making it even better. It is astonishing but true that AI, a man’s creation, the result of mere algorithms, can solve one of the most pestering and complex predicaments faced by the man himself. Yet skeptics believe that AI’s risks are as large as its potential benefits. How can they be avoided? And why isn’t the most powerful technology being used more widely today to solve the world’s greatest “wicked” problems? The main purpose of AI is for humans to have a machine that thinks faster and more efficiently. But the question becomes, in the process, will machines take over our world? Will it help us reach our highest potential or destroy us in the process? The answers to these questions lie hidden under the gleam of an unforeseen future. We opened the window of technicalities for our readers. Now whether to cherish or fear the advancement we leave to them.