AI Psychiatrist

• 

  Camera, EEG and Graph which captures the data in realtime

Inspiration

During NGO help for Destiny Path Association of Kenya I determined highest stress level which could compare with professional Jet Pilot or even astronaut. Such events develops negative experience and pressure of our conciousness, leading post traumatic syndrome. People failing maintaining their daily lives, isolation, starting argue each other. Than they are seeking FirstAid, but usual situation more complex and deep. It may be caused by past Disorders, Neuraldivergency and so on. The project aims to set a bridge between the Doctor and patient, preventing misleading or making situation even worse.

What it does

It could used as like multipurpose layer Question-Answer system, which using Machine Learning architecture, determine CURRENT probability of the complex mental disorder, or detect the level of the progression of a kind of. Tied to the situation, doctor can examine the state of the patient by video, if possible, or start quick QnA, which using ICD, can measure the speed of type or voice tembre, such help the guidance to the diagnosis. More, doctor can use the analyzer, which will compare previous analysis with the current, using very complex translator, that could "decipher" the illness history

How we built it

Frontend: WindowsForms Microsoft's C# user interface, Backend: Dotnet10 C#, Prolog, Microsoft MLNet. AI: Prolog KnowledgeBase, Automata theory, OpenCV Face recognition. PsychiatricData: EEG Spectrogram Images for Schizophrenia Detection from the Kaggle, Credits: Ayush Aote REFERENCE datasource

Challenges we ran into

Misunderstanding, afraid of innovation technologies

Accomplishments that we're proud of

We're evade the using of heavy models like LSTM, CNN and other "classical" NeuralNetwork Machine Learning architectures due to their delay and need of constant re-education, and data-hungry, plus much of calculation unit requiring. Instead, we presented the Automata theory, where UI program is In-Data: FaceRecognition via OpenCV and short Answer List, Trigger is Knowledge Minimal quantity: one Matrix per second and filled psych list, the combinational circuit is the KnowledgeBase which was created entirely on Prolog. And out conclusion is the Live-captured Spectrogram with real-life pre-trained weights, which neglegt the classical plot graphs and will probably, determine the MODA dots with existing EEG data. Such architecture shares the functions of the Software Suite, which gives an Opportunity Quickly change the goal, making entire solution pretty universal and fast-maintanable

What we learned

Automata theory, Visual prolog 11, decision tree making, EEG, facts about complex psychiatric diseases, Machine learning algorithms, Software Architecture design

What's next for AI Psychiatrist

Fix the minor bugs, the expanding of functionality, capability to work with more data

Built With

• c#
• copilot
• dart
• gemini
• tensorflow