How to identify input fields in unstructured PDFs using Python

I am trying to parse an unstructured PDF file and extract information about some input fields like radiobuttons, but I am not sure how to do that.

I tried using get_fields from PyPDF2, it does not return anything because of the nature of the PDF. When I use extract_text() it just gives "YES" or "NO" for the radiobutton components. Also when i print the full extracted text, there's often random spaces in between the words.Is there any method of accurately parsing unstructured PDF files?

I tried the way @jaco-kemp, but its not working. However, here one solution and works good for the pdf which you shared with me. But I am not sure is it going to work with others as it is little dirty :)) :

import fitz
import cv2
import numpy as np

def extract_radio_button_info(pdf_path):
    doc = fitz.open(pdf_path)
    for page_num in range(len(doc)):
        print('-' * 100)
        print('Page number:', page_num)
        page = doc[page_num]
        text = page.get_text("text")
        lines = text.splitlines()
        for i, line in enumerate(lines):
            if "SOUR SERVICE (PER NACE MR-01-75)" in line:
                text_instances = page.search_for("SOUR SERVICE (PER NACE MR-01-75)")
                if text_instances:
                    print(i, line)
                    for ii, text_rect in enumerate(text_instances):
                        print('Y coordinate:', text_rect.y1)
                        # Convert PDF page to image
                        pix = page.get_pixmap(alpha=False)
                        img = np.frombuffer(pix.samples, dtype=np.uint8).reshape(pix.height, pix.width, pix.n)
                        # Find radio button coordinates
                        radio_button_x = text_rect.x1 + 50  # Adjust the offset as needed
                        radio_button_y = text_rect.y1
                        # Extract the radio button regions from the image
                        yes_button_region = img[int(radio_button_y - 7):int(radio_button_y + 1),
                                             int(radio_button_x + 20):int(radio_button_x + 30)]
                        no_button_region = img[int(radio_button_y - 7):int(radio_button_y + 1),
                                            int(radio_button_x + 75):int(radio_button_x + 85)]
                        # Save the radio button regions for visualization (optional)
                        cv2.imwrite(f'yes_button_region.png', yes_button_region)
                        cv2.imwrite(f'no_button_region.png', no_button_region)
                        # Calculate the mean pixel intensity for each region
                        yes_mean_intensity = cv2.mean(yes_button_region)[0]
                        no_mean_intensity = cv2.mean(no_button_region)[0]
                        # Determine the selected option based on the darker region
                        if yes_mean_intensity < no_mean_intensity:
                            selected_option = "YES"
                        else:
                            selected_option = "NO"
                        print(f"Selected option: {selected_option}")



if __name__ == "__main__":
    pdf_path = "data/sample.pdf"
    extract_radio_button_info(pdf_path)