maancham / ephi

EPHI Explaianbel Private Healthcare Infra (LLM)

Product Demo

What doctors will see:

What patients will see:

Pseudocode

initialize_chatbot()
set_chatbot_personality("doctor's assistant")
patient_data = get_patient_data()

// First Iteration
diagnosis_result = chatbot_identify_issue(patient_data)
iteration_results = [diagnosis_result]

// Additional Iterations
for i in range(MinNumberOfIterations - 1):
    masked_data = apply_mask(patient_data, i)
    new_diagnosis_result = chatbot_identify_issue(masked_data)
    iteration_results.append(new_diagnosis_result)

// Output Results
output_results(iteration_results)

The algorithm initializes the language model using the initialize_chatbot() function, and sets the personality prompt for the chatbot using set_chatbot_personality(personality_prompt). The get_patient_data() function retrieves the medical records of a patient.

For the first iteration, the chatbot_identify_issue(patient_data) function is called to instruct the chatbot to identify the main issue in the patient data and output a diagnosis with supporting arguments. The resulting diagnosis is stored in diagnosis_result.

In subsequent iterations, the algorithm applies a mask algorithm to the patient data using apply_mask(patient_data, i), which drops some data points to simulate a smaller dataset. The masked data is then passed to chatbot_identify_issue(masked_data) to obtain a new diagnosis result, which is stored in new_diagnosis_result.

Finally, the output_results(iteration_results) function is called to output all the gathered diagnosis results from each iteration of the algorithm.