app.py

import gradio as gr
import pandas as pd
import fitz  # PyMuPDF
import pytesseract
from pdf2image import convert_from_path
import os
import base64
from google import genai
from google.genai import types

google_api=os.getenv("google_api")




def read_pdf(pdf_path):
    text = ""
    doc = fitz.open(pdf_path)

    for page_num in range(len(doc)):
        page = doc.load_page(page_num)
        page_text = page.get_text("text").strip()  # Extract text from page

        # Extract Images for OCR
        images = page.get_images(full=True)  # Check if the page has images

        ocr_text = ""
        if images:  # If images exist, process them
            print(f"Page {page_num + 1} contains images, performing OCR...")
            img_pages = convert_from_path(pdf_path, first_page=page_num + 1, last_page=page_num + 1)
            
            for img in img_pages:
                ocr_text += pytesseract.image_to_string(img).strip() + "\n"

        # Combine both text extraction methods
        combined_text = f"{page_text}\n{ocr_text}".strip()
        
        if combined_text:
            text += combined_text + "\n\n"

    return text.strip()






def generate(extracted_text):
    client = genai.Client(
        api_key=google_api,
    )

    model = "gemini-2.0-flash"
    contents = [
        types.Content(
            role="user",
            parts=[
                types.Part.from_text(text="""The following text is extracted from a medical lab report using OCR.
There may be errors such as missing decimals, incorrect test names, and incorrect reference ranges.
Please correct the errors and extract both metadata and structured lab test data.
ALWAYS MAKE SURE THAT THE VALUE ALIGNS WITH THE REAL RANGE OF THE TEST
AND CLEARLY IDENTIFY REDS WITH LOW AND HIGH
Return the output in structured JSON format with all the information in lowercase to standardization.
And follow the JSON format provided and don't add any additional details in meta data or lab report other than that are specified


Extracted Text:
Dr. Onkar Test Sanjeevan Hospital\\n\\nMBBS, MD | Reg No: T123 12/4, Paud Road, Kothrud, Pune - 411023\\nPh: 0202526245, 8983390126, Timing: 09:15 AM -\\n02:30 PM, 05:30 PM - 09:30 PM, APPOINTMENTS\\nONLY | Closed: Monday,Friday\\n\\n \\n\\nPatient UID: 87 Report No: 00018\\n\\nName: AMAR SHAHA (Male} Rey, Date: 09-Jul-20\\n\\nAge 40 years Sample Collected At Hospital Lab\\n\\nAddress: MG Road, PUNE Sample Type/Quantity: Blood\\n\\nRef. By Doctor . Sample Collection D/T: 09-Jul-20, 9.50 AM\\nCr Test Result D/T: 09-Jul-20, 4:53 PM\\n\\n \\n \\n\\nDr. Amit Deshmukh\\n\\n     \\n\\nHEMOGRAM\\n\\nINVESTIGATION RESULT UNIT REF, RANGE\\nHAEMOGLOBIN : 14 gms/dl 12.0 - 17.0\\nRBC COUNT E 44 millfeumm 4.1-5.1\\nHAEMOTOCRIT (PCV) E 30 % 32.0 - 47.0\\nMCV $ 78 fl 760 - 100.0\\nMCH H 3246 Py 260-320\\nMCHC | : 328 n% 315-3465 ,\\nROW ; 13.9 % 11.6-150\\nMPV ; 11.2 fn 68- 12.6\\nWBC COUNT : 4567 /eamm 4000 - 11000\\nDIFFERENTIAL COUNT\\nNEUTROPHILS |» : 56 %y 40-70\\nLYMPHOCYTES ; 20 % 20.0- 45.0\\nEOSINOPHILS . 4 « % 0-6\\nMONOCYTES : 5 %

Expected JSON format:
{
    \"metadata\": {
        \"patient_name\": \"Prasahsst Pawar\",
        \"age\": \"20\",
        \"gender\": \"Male\",
        \"lab_name\": \"XYZ Diagnostics\",
        \"report_date\": \"05-03-2025\"
    },
    \"lab_tests\": [
        {
            \"test_name\": \"hemoglobin\",
            \"value\": \"14.2\",
            \"unit\": \"g/dL\",
            \"reference_range\": \"13.5 - 17.5 g/dL\"
        },
        {
      \"test_name\": \"rbc count\",
      \"value\": \"5.2\",
      \"unit\": \"million/cu mm\",
      \"reference_range\": \"4.1-5.1\"
    },
        {
            \"test_name\": \"glucose\",
            \"value\": \"65\",
            \"unit\": \"mg/dL\",
            \"reference_range\": \"70 - 110 mg/dL\"
        }
    ],
\"reds\":{
    \"low\":[\"glucose\"],
    \"high\":[\"rbc count\"]
}
}"""),
            ],
        ),
        types.Content(
            role="model",
            parts=[
                types.Part.from_text(text="""{
  \"lab_tests\": [
    {
      \"reference_range\": \"12.0 - 17.0\",
      \"test_name\": \"haemoglobin\",
      \"unit\": \"gms/dl\",
      \"value\": \"14\"
    },
    {
      \"reference_range\": \"4.1-5.1\",
      \"test_name\": \"rbc count\",
      \"unit\": \"mill/cu mm\",
      \"value\": \"4.4\"
    },
    {
      \"reference_range\": \"32.0 - 47.0\",
      \"test_name\": \"haematocrit (pcv)\",
      \"unit\": \"%\",
      \"value\": \"30\"
    },
    {
      \"reference_range\": \"76.0 - 100.0\",
      \"test_name\": \"mcv\",
      \"unit\": \"fl\",
      \"value\": \"78\"
    },
    {
      \"reference_range\": \"26.0-32.0\",
      \"test_name\": \"mch\",
      \"unit\": \"pg\",
      \"value\": \"32.46\"
    },
    {
      \"reference_range\": \"31.5-34.5\",
      \"test_name\": \"mchc\",
      \"unit\": \"%\",
      \"value\": \"32.8\"
    },
    {
      \"reference_range\": \"11.6-15.0\",
      \"test_name\": \"rdw\",
      \"unit\": \"%\",
      \"value\": \"13.9\"
    },
    {
      \"reference_range\": \"6.8- 12.6\",
      \"test_name\": \"mpv\",
      \"unit\": \"fn\",
      \"value\": \"11.2\"
    },
    {
      \"reference_range\": \"4000 - 11000\",
      \"test_name\": \"wbc count\",
      \"unit\": \"/cu mm\",
      \"value\": \"4567\"
    },
    {
      \"reference_range\": \"40-70\",
      \"test_name\": \"neutrophils\",
      \"unit\": \"%\",
      \"value\": \"56\"
    },
    {
      \"reference_range\": \"20.0- 45.0\",
      \"test_name\": \"lymphocytes\",
      \"unit\": \"%\",
      \"value\": \"20\"
    },
    {
      \"reference_range\": \"0-6\",
      \"test_name\": \"eosinophils\",
      \"unit\": \"%\",
      \"value\": \"4\"
    },
    {
      \"reference_range\": \"2-10\",
      \"test_name\": \"monocytes\",
      \"unit\": \"%\",
      \"value\": \"5\"
    }
  ],
  \"metadata\": {
    \"age\": \"40\",
    \"gender\": \"male\",
    \"lab_name\": \"sanjeevan hospital\",
    \"patient_name\": \"amar shaha\",
    \"report_date\": \"09-jul-20\"
  },
  \"reds\": {
    \"high\": [
      \"mch\"
    ],
    \"low\": [
      \"haematocrit (pcv)\"
    ]
  }
}"""),
            ],
        ),
        types.Content(
            role="user",
            parts=[
                types.Part.from_text(text=extracted_text),
            ],
        ),
    ]
    generate_content_config = types.GenerateContentConfig(
        temperature=1,
        top_p=0.95,
        top_k=40,
        max_output_tokens=8192,
        response_mime_type="application/json",
        response_schema=genai.types.Schema(
            type = genai.types.Type.OBJECT,
            enum = [],
            required = ["metadata", "lab_tests", "reds"],
            properties = {
                "metadata": genai.types.Schema(
                    type = genai.types.Type.OBJECT,
                    enum = [],
                    required = ["patient_name", "age", "gender", "lab_name", "report_date"],
                    properties = {
                        "patient_name": genai.types.Schema(
                            type = genai.types.Type.STRING,
                        ),
                        "age": genai.types.Schema(
                            type = genai.types.Type.STRING,
                        ),
                        "gender": genai.types.Schema(
                            type = genai.types.Type.STRING,
                        ),
                        "lab_name": genai.types.Schema(
                            type = genai.types.Type.STRING,
                        ),
                        "report_date": genai.types.Schema(
                            type = genai.types.Type.STRING,
                        ),
                    },
                ),
                "lab_tests": genai.types.Schema(
                    type = genai.types.Type.ARRAY,
                    items = genai.types.Schema(
                        type = genai.types.Type.OBJECT,
                        enum = [],
                        required = ["test_name", "value", "unit", "reference_range"],
                        properties = {
                            "test_name": genai.types.Schema(
                                type = genai.types.Type.STRING,
                            ),
                            "value": genai.types.Schema(
                                type = genai.types.Type.STRING,
                            ),
                            "unit": genai.types.Schema(
                                type = genai.types.Type.STRING,
                            ),
                            "reference_range": genai.types.Schema(
                                type = genai.types.Type.STRING,
                            ),
                        },
                    ),
                ),
                "reds": genai.types.Schema(
                    type = genai.types.Type.OBJECT,
                    enum = [],
                    required = ["low", "high"],
                    properties = {
                        "low": genai.types.Schema(
                            type = genai.types.Type.ARRAY,
                            items = genai.types.Schema(
                                type = genai.types.Type.STRING,
                            ),
                        ),
                        "high": genai.types.Schema(
                            type = genai.types.Type.ARRAY,
                            items = genai.types.Schema(
                                type = genai.types.Type.STRING,
                            ),
                        ),
                    },
                ),
            },
        ),
        system_instruction=[
            types.Part.from_text(text="""Always return the output as JSON only"""),
        ],
    )



    # for chunk in client.models.generate_content_stream(
    #     model=model,
    #     contents=contents,
    #     config=generate_content_config,
    # ):
    #     print(chunk.text, end="")


    response = client.models.generate_content(
        model=model,
        contents=contents,
        config=generate_content_config,
    )

    json_response = response.text  # The API should return JSON text
    parsed_json = json.loads(json_response)  # Convert JSON string to Python dictionary

    return parsed_json





# Gradio interface function
def process_pdf(pdf):
    text = read_pdf(pdf)  # Extract text from PDF
    output = generate(text)  # Generate structured JSON

    return output

def show_to_UI(pdf):
    output = process_pdf(pdf)  # Call process_pdf to get JSON
    # output= {"metadata": {"name": "John Doe", "age": 30}, "lab_tests": [{"test": "Blood Sugar", "value": "120 mg/dL"}]}

    # Extract metadata
    metadata = output["metadata"]
    labtests = pd.DataFrame(output["lab_tests"])
    reds = pd.DataFrame(output["reds"])

    metadata_str = f"**Patient Name:** {metadata['patient_name']}\n\n" \
                   f"**Age:** {metadata['age']}\n\n" \
                   f"**Gender:** {metadata['gender']}\n\n" \
                   f"**Lab Name:** {metadata['lab_name']}\n\n" \
                   f"**Report Date:** {metadata['report_date']}"

    return metadata_str, labtests, reds, output  

# Define Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("# Medical Lab Report Processor")

    with gr.Row():
        pdf_input = gr.File(label="Upload PDF Report")
        submit_btn = gr.Button("Process")

    metadata_output = gr.Markdown("**Patient Name: Prashasst...**")
    lab_test_output = gr.Dataframe(label="Lab Test Results")
    reds_output = gr.Dataframe(label="Reds")
    output_JSON = gr.JSON(label="Extracted Report")  # Show JSON output

    submit_btn.click(process_pdf, inputs=[pdf_input], outputs=[metadata_output, lab_test_output, reds_output,output_JSON])

demo.launch()