top of page
  • Writer's pictureMariam Heikal

Data and delusions: early detection of schizophrenia in MENA


Did you know that 1 in 300 people globally is affected by schizophrenia? More than 24 million individuals worldwide suffer from this disease [1].

Not only is schizophrenia one of the top 15 major causes of disability worldwide [2], but it also reduces a person's life expectancy by 10–20 years compared to the general population. In addition, only 15% of individuals diagnosed with schizophrenia can hold down a full-time job [3].

Although schizophrenia is a treatable mental disorder, nearly half of those diagnosed in the United States do not receive treatment. The situation is even worse in developing countries, where approximately 70% of the schizophrenic population is left untreated.

These are global statistics. Let's take a deeper look at the statistics of the MENA region.

According to a recent study [4], Arab countries have, on average, a greater burden of mental illnesses as assessed by disability-adjusted life years than the rest of the world. Despite this, our region lacks reliable statistics on schizophrenia. With no evidence reported that demonstrates a distinction between the prevalence of schizophrenia and other psychotic disorders in Egypt and other countries, it can be estimated that there are between 500,000 and 1.5 million cases of schizophrenia in Egypt [5]. In 2006, when the referenced research was conducted, it was expected that these numbers would double in 40 years.


What happens when schizophrenia is not diagnosed at an early stage? In a TED talk, Cecilia McGough, an astronomer and the founder and executive director of Students with Psychosis, shared her experience with schizophrenia, highlighting the risks associated with late diagnosis.

McGough was thought to have schizophrenia her entire life, but she particularly struggled with the disease during her junior year of high school. Her symptoms intensified in college. She described her life as a "waking nightmare" following this intensification. The harrowing experience of continuous hallucinations—"I’m just someone who cannot turn off my nightmares, even when I am awake," she explained—was a significant contributor to her suicide attempt during her junior year of college [6].

Cecilia explained that she learned to ignore and pretend not to see her hallucinations but that the experience led to triggers, such as seeing the colors red and white, which are associated with her hallucinations. This was the primary reason why, at the TED conference where she gave her speech, the lights were dimmed to conceal the colors of their logo, and the usual red carpet was replaced with a black one. Minor changes helped relieve her symptoms.

What would have been the outcome had Cecilia been able to benefit from early detection of schizophrenia? What is the impact of early detection on a patient’s quality of life? More importantly for the purpose of this article: can machine learning (ML) and artificial intelligence (AI) assist in early detection of the disease?

Use-case of AI in early detection of schizophrenia

Due to the lack of a clinical test or radiological X-ray test for schizophrenia, the usual diagnostic procedure is neither clear nor rapid, and in many cases, results in an incorrect diagnosis.

Typically, psychiatrists arrive at a diagnosis by gathering information from a clinical interview with the patient and from the patient's family and friends, in order to determine whether the patient genuinely exhibits indicators of schizophrenia. The information gathered is then compared to the Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria for schizophrenia symptoms [7].

Recent advancements in ML offer intriguing approaches that could assist clinicians in diagnosing a variety of mental disorders, including schizophrenia.

Recent research aimed to identify schizophrenia from T1-weighted magnetic resonance imaging (MRI) scans using a deep learning model. MRI equipment is used to generate high quality images of soft tissues and internal organs. The quality of the images depends on the imaging techniques utilized, specifically T1 and T2. T1 and T2 are the relaxation durations utilised during tissue scanning as an interval between pulse sequences. In addition, a gradient class activation map (Grad-CAM) was used to explain the model's decision-making process. Grad-CAM is a technique for increasing the transparency of convolutional neural network (CNN)-based models by visualizing the significant input regions for model predictions [8].

Using a two-step procedure, the built model can identify schizophrenic patients. Initially, features are extracted hierarchically from the input T1 MRI scan, beginning with down-sampling the features and passing them to multiple convolution blocks. The final prediction result is then generated by feeding the retrieved features to a classifier consisting of three dense layers with dropout regularization. On unseen MRI scans, the model was able to differentiate nearly flawlessly between individuals with schizophrenia and healthy controls [8].

Moreover, by visualizing the class activation maps through Grad-CAM, the model identifies the brain regions that contributed the most to the diagnosis of schizophrenia, making it transferrable and explicable. Utilizing such models expedites and simplifies the diagnostic procedure for psychiatrists, resulting in an early diagnosis of schizophrenia.

That said, early diagnosis is not just limited to MRIs. Other clinical data, such as electroencephalogram (EEG) recordings, blood biomarkers, and RNA sequences, have been used in conjunction with ML models to detect schizophrenia [9, 10].

Despite the certainty that ML can uncover minute similarities in patient data and improve the diagnosis process for psychiatrists, the question remains as to whether we have the necessary infrastructure to implement this technology in the MENA region.

Technical feasibility in the MENA region

To implement this technology in the MENA, mental health facilities must be equipped with digital healthcare-supporting equipment and systems.

Recent research was undertaken to determine psychiatrists' perspectives on the present situation of Egypt’s mental health care services and their current understanding of electronic mental health (EMH). Although research indicated that Egypt lacks adequate mental health resources, it also showed that Egyptian psychiatrists were interested in EMH and believe that web-based platforms can contribute to the improvement of Egypt's mental health care system.

In Libya, the mental health system is deemed ineffective and insufficient to meet the country's requirements, according to a new study on the country's mental health services [11]. Several hospitals and clinics have provided telepsychiatry, online psychological support, and psychiatric counselling services as a result of recent technological advancements in the mental health industry.

However, these technological advancements are insufficient to develop AI models that can assist psychiatrists with diagnosis. If you recall our discussion around the use of MRI scans to diagnose schizophrenia, for example: psychiatrists must have access to digital platforms with an easy-to-use interface that support the integration of such models and provide technology that allows the patient's scans to be fed into the model.

The first step towards implementing ML models in mental health systems would be to transition from a traditional notes-based system to web-based platforms that allow psychiatrists to record data about different patients and reuse it effectively. In addition to enhancing our mental health infrastructure, the MENA region must further invest in mental health research and specifically the use of ML for mental health. Indeed, studies show that Arab countries are responsible for approximately 1% of the world's peer-reviewed papers in mental health research – this is despite the fact that mental illness is the main cause of disability in the Arab region (where 5.54% of the world's population resides) [12].

Understandably, there are concerns regarding the non-technical implementation factors, such as social accessibility, the availability of suitable expertise in the region, a lack of AI trust, as well as others. I delve further into these issues in this article on AI-based models for diagnosing delirium. Give it a read and let me know what you think!


  1. World Health Organization. (2022, June 8). Mental disorders. World Health Organization. Retrieved from

  2. Global, regional, and national incidence, prevalence, and years lived with disability for 328 diseases and injuries for 195 countries, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet. 2017 Sep 16;390(10100):1211-1259. PMID: 28919117

  3. Facts and statistics for schizophrenia - how common is it? FHE Health – Addiction & Mental Health Care. (n.d.). Retrieved from

  4. Kamel, M. M., Westenberg, J. N., Choi, F., Tabi, K., Badawy, A., Ramy, H., ... & Krausz, M. (2020). Electronic mental health as an option for Egyptian psychiatry: cross-sectional study. JMIR mental health, 7(8), e19591.

  5. Zahran, N. S., Khalil, A. H., Okasha, T. A., & Adel Sadek, H. (2006). Systematic review of Egyptian studies on schizophrenia. Published Master's Dissertation), Faculty of Medicine. Ain Shams University.

  6. TEDx Talks. (2017, March 27). I Am No A Monster: Schizophrenia | Cecilia McGough | TEDxPSU [Video]. YouTube.

  7. How Schizophrenia Is Diagnosed | Schizophrenia. (2013, August 25). [Video]. YouTube.

  8. Zhang, J., Rao, V. M., Tian, Y., Yang, Y., Acosta, N., Wan, Z., ... & Guo, J. (2022). Detecting Schizophrenia with 3D Structural Brain MRI Using Deep Learning. arXiv preprint arXiv:2206.12980.

  9. Buettner, R., Beil, D., Scholtz, S., & Djemai, A. (2020). Development of a machine learning based algorithm to accurately detect schizophrenia based on one-minute EEG recordings.

  10. Li, Z., Li, X., Jin, M., Liu, Y., He, Y., Jia, N., ... & Yu, Q. (2022). Identification of potential blood biomarkers for early diagnosis of schizophrenia through RNA sequencing analysis. Journal of Psychiatric Research, 147, 39-49.

  11. Shoib, S., Baiou, A., Saleem, S. M., Chandradasa, M., & Gaffaz, R. (2022). Mental health services in conflict areas–An experience from Libya. Asian Journal of Psychiatry, 73, 103106.

  12. Maalouf, F. T., Alamiri, B., Atweh, S., Becker, A. E., Cheour, M., Darwish, H., ... & Akl, E. A. (2019). Mental health research in the Arab region: challenges and call for action. The Lancet Psychiatry, 6(11), 961-966.


bottom of page