Upload 7 files

src/__init__.py

@@ -0,0 +1,7 @@
+import logging
+import os
+from pathlib import Path 
+
+logging.basicConfig(level=logging.INFO,format = "[%(asctime)s]: %(message)s:")
+project_name = "medical_report_analysis"
+

src/document.py

@@ -0,0 +1,45 @@
+import fitz
+from PIL import Image
+from src.setting import LOCAL_TEST
+import numpy as np
+from src import logging
+
+
+class PDF_Processing:
+    def __init__(self):
+        pass
+
+    def pdf_to_image(file):
+        """
+        This function will take pdf file and convert first page into image and return it.
+
+        LOCAL_TEST : True / False
+        
+        """
+
+        try:
+            if LOCAL_TEST:
+                pdf_file = fitz.open(filename=file,filetype="pdf")
+            else:
+                pdf_file = fitz.open(stream=file.read(), filetype="pdf")
+
+            page = pdf_file.load_page(0)
+            pix = page.get_pixmap(matrix=fitz.Matrix(300/72, 300/72))  # 300 DPI
+            image_bytes = pix.samples
+            image = Image.frombytes("RGB", [pix.width, pix.height], image_bytes)
+            image_np = np.array(image)
+
+            return image_np
+        except Exception as e:
+            logging.info(f"Error {e} : pdf_to_image")
+
+    def load_image(file):
+        """
+        This function take image and return numpy array
+        """
+        try:
+            image = Image.open(file)
+            image_np = np.array(image)
+            return image_np
+        except Exception as e:
+            logging.info(f"Error {e} : load_image")

src/llms.py

@@ -0,0 +1,33 @@
+import os
+import google.generativeai as genai
+from src.setting import MODEL_NAME
+from dotenv import load_dotenv
+from src.prompts import Prompts
+from src import logging
+load_dotenv()
+
+
+class LLM:
+    def __init__(self):
+        self.API_KEY = os.getenv("GOOGLE_API_KEY")
+
+    def get_json(self,input_data:str,key:str = None):
+        """
+        Input_data : It is a string input. It can take json as well as raw text
+
+        key : Default None.
+            It can Json and None. If the input_data is json than key will json else None
+        
+        """
+        if key == "json":
+            prompts = Prompts.text_json_prompt().format(text = input_data)
+        else:
+            prompts = Prompts.final_prompt().format(json_data = input_data)
+
+        try:
+            genai.configure(api_key = self.API_KEY)
+            model = genai.GenerativeModel(model_name=MODEL_NAME)
+            response = model.generate_content(prompts)
+            return response.text
+        except Exception as e:
+            logging.info(f"Error :{e}  : LLM.get_json")

src/ocr_model.py

@@ -0,0 +1,25 @@
+from easyocr import Reader
+from src.setting import OCR_MODEL_LANGUAGE
+from src import logging
+
+class OCR:
+    def __init__(self):
+        pass
+
+    def extract_text(image):
+        """
+        image : image or numpy array.
+
+        This function is for OCR 
+
+        Return : raw text from ocr model
+        
+        """
+        try:
+            ocr = Reader(lang_list=[OCR_MODEL_LANGUAGE])
+            result = ocr.readtext(image)
+            text = [bbox[1] for bbox in result]
+            text = "\n".join(text)
+            return text
+        except Exception as e:
+            logging.info(f"Error : {e} : OCR.extract_text")

src/pipeline.py

@@ -0,0 +1,34 @@
+from src.document import PDF_Processing
+from src.ocr_model import OCR
+from src.llms import LLM
+from src import logging
+import os
+
+
+class Pipeline:
+    def __init__(self):
+        self.cwd = os.getcwd()
+
+    def process(file,type):
+        """
+        file : data it can be image or pdf
+        Type : format of PDF / Image (png, jpg)
+        
+        return : Clean Text.
+        """
+        try:
+            print("startd")
+            if type == "pdf":
+                image = PDF_Processing.pdf_to_image(file)
+            else:
+                image = PDF_Processing.load_image(file)
+            text = OCR.extract_text(image)
+            json_text = LLM().get_json(input_data=text,key = "json")
+            final = LLM().get_json(input_data=json_text)
+            return final
+        except Exception as e:
+            logging.info(f"Error :{e} :Pipeline.process")
+
+if __name__ == "__main__":
+    path = "test_docs/CBC-test-report-format-example-sample-template-Drlogy-lab-report.pdf"
+    result = Pipeline.process(path)

src/prompts.py

@@ -0,0 +1,43 @@
+class Prompts:
+
+    def __init__(self) -> None:
+        pass
+
+    def final_prompt():
+        return """
+        You are a doctor explaining a patient's medical report in simple, easy-to-understand language. Focus only on the areas of concern and explain the health issues, if any, and provide natural suggestions to improve the condition.
+
+        Here is the report:
+
+        Report: {json_data}
+
+        Based on the above report, explain only the report do not give patient details and provide natural remedies or ways to heal.
+        """
+
+
+    def text_json_prompt():
+        return """
+        You are an expert in text comprehension. Your task is straightforward: understand the provided text and return the output in the specified JSON format.
+
+        Text: {text}
+
+        Output: Please provide the output in the following JSON format:
+
+        {{
+        "patient_name": "<patient name>",
+        "lab_no": "<lab number>",
+        "lab_name": "<laboratory name>",
+        "collection_date_time": "<sample collection date and time>",
+        "reported_date_time": "<report date and time>",
+        "test_name": "<test name>",
+        "patient_age": "<patient age>",
+        "patient_gender": "<patient gender>",
+        "Report": {{
+            "<investigation name>": {{
+            "result": "<result value>",
+            "reference_value": "<reference range>",
+            "unit": "<unit of measurement>"
+            }}
+        }}
+        }}
+        """

src/setting.py

@@ -0,0 +1,3 @@
+LOCAL_TEST = False   # if local test = True than pass the pdf file path 
+MODEL_NAME = "gemini-pro" # LLM model name 
+OCR_MODEL_LANGUAGE = "en" # ocr model language.