AI-Powered Innovation: A New Era in Psychiatric Diagnosis with BraiNP

NPİSTANBUL Hastanesi, in collaboration with Üsküdar University, has pioneered a groundbreaking innovation in the diagnosis and treatment of psychiatric diseases with its AI-powered BraiNP model. Developed at an international level, this model aims to accelerate clinical decision-making processes and increase patient satisfaction with individualized treatments. 

The AI-based “BraiNP” model, developed through the scientific collaboration of NPİSTANBUL Hastanesi and Üsküdar University, is revolutionizing the diagnosis and treatment of psychiatric diseases. This innovative model, for which an international patent application has been filed, provides preliminary diagnosis and treatment prediction with high accuracy rates, significantly contributing to both the healthcare system and individual treatment processes.

What is BraiNP?

“BraiNP” is an advanced model developed with psychiatric data accumulated by NP Model since 1998, where neuroimaging data such as EEG and fMRI are analyzed using artificial intelligence algorithms. Emerging from scientific studies conducted at Üsküdar University's application and research centers, BraiNP demonstrates high performance in the preliminary diagnostic classification and prognostic prediction of treatment response for many psychiatric diseases.

The main goal in developing the model is to integrate prediction algorithms, which have been limited to academic publications, into the healthcare system, thereby ensuring the efficient use of healthcare resources through early diagnosis and effective treatment planning.

For Which Diseases Is It Used?

The BraiNP model can process data related to a wide range of psychiatric disorders. 

These include:

- Alcohol Addiction

- Mixed Substance Addiction

- Nicotine Addiction

- Gambling Addiction

- Behavioral Addictions

- Psychosis

- Anxiety

- Post-Traumatic Stress Disorder (PTSD)

- Obsessive Compulsive Disorder (OCD)

- Depression

- Bipolar Disorder

- Anorexia

- Autism

Additionally, the system contributes to the creation of new classification models by instantly analyzing data diagnosed by physicians in the field.

How Does the BraiNP Artificial Intelligence Model Work?

While classical artificial intelligence algorithms largely rely on commonly used biological markers in the literature for diagnosis, BraiNP significantly improves data resolution and model performance using next-generation deep learning methods.

The operational stages of the model are as follows:

Data Collection and Pre-processing: EEG (high temporal resolution) and fMRI (high spatial resolution) data are collected from patients or healthy individuals and pre-processed to remove noise.

Feature Extraction: With the developed algorithm, data is processed on GPU-supported computers in a cloud environment to extract distinctive features.

Model Development: Classification models are created using these features.

Performance Measurement: Models are tested and evaluated using various statistical metrics.

Clinical Validity: The clinical usability of the model is proven through external tests conducted with double-blind data.

Despite their low explainability, next-generation models have been made suitable for clinical interpretation by being supported with visual tools such as heatmaps.

International Patent Application and Its Benefits 

The BraiNP model has achieved high accuracy rates in areas such as distinguishing between OCD and healthy individuals, classifying unipolar–bipolar disorder, and predicting response to transcranial magnetic stimulation (TMS) in depression.

The international patent application for the model has been completed, thereby ensuring the system's global recognition and protection. The patent registration not only certifies the model's originality and innovativeness but also allows for its daily use by NPİSTANBUL Hastanesi physicians.

7 Key Contributions to the Healthcare System

The main contributions of the BraiNP application to the healthcare system in the short and long term are:

- Early Intervention: Mental disorders can be detected early, allowing for effective and rapid intervention.

- Prevention of Complications: The development of comorbid conditions and risky behaviors can be prevented.

- Reduced Suffering and Increased Quality of Life: Timely diagnosis reduces an individual's suffering.

- Personalized Treatment Plans: Unique treatment approaches can be developed for each patient.

- Efficient Use of Resources: Emergency room burden is reduced, and unnecessary hospitalizations are prevented.

- Family Education and Psychosocial Support: Families are enlightened at an early stage.

- Improved Prognosis: Long-term recovery and low recurrence rates are achieved during the treatment process.

Scientific Breakthroughs from Üsküdar University in Brain-Computer Interfaces and Artificial Intelligence 

Üsküdar University trains specialists at the postgraduate level in BCI (Brain-Computer Interfaces), artificial intelligence, and neuroimaging, while also conducting active R&D projects.

BCI systems enable individuals who have lost neuromuscular functions (ALS, paralysis, spinal cord injuries, etc.) to interact with computers or mechanical devices. Devices such as robotic arms, prostheses, and wheelchairs can be controlled using signal sources like EEG, MEG, and fMRI.

Next-Generation Big Data Application with Facial Emotion Recognition (FER) Model

Another artificial intelligence project developed within the university, the “Facial Emotion Recognition” (FER) model, can analyze an individual's dominant emotion from facial images in real-time.

Used for detecting basic emotions such as happiness, sadness, anger, surprise, fear, and disgust, this system offers versatile usage opportunities in health, education, marketing, and security sectors.

Prof. Dr. Türker Tekin Ergüzel stated that despite challenges such as cultural differences, visual obstacles (glasses, beards, etc.), and ethical concerns, they are working on a multi-modal prediction architecture by integrating this model with voice data.