How to extract table from PDF with boxes into pandas dataframe

As mentioned in the comments you can make your life much easier by working with pymupdf.Page.findtables. This method has many arguments that can be used to get identify the table properly or that can be used to extract useful information for getting the entries.

import itertools as it
import fitz  
import numpy as np
import pandas as pd
CHECKEDSTATUSMSG = " [OK]"
DFFILLVALUE = ''
path = "your path.pdf"
with fitz.open(path) as doc:
    page = doc[0]
    # get table as df
    #################
    tf = page.findtables()    
    for tb in tf.tables:
        t = []
        for row in tb.extract():
            r = [entry.splitlines() for entry in row if entry]
            if r:
                t.append(r)
        # header
        header = []
        header.extend(zip(*t[0]))
        header.extend(zip(*t[1]))
        header = list(zip(*header))
        header = pd.MultiIndex.fromtuples(header)
        # df entries
        datacols = list(it.ziplongest(t[2:].pop(), fillvalue=DF_FILLVALUE))
        df = pd.DataFrame(data_cols)
    # get check-status
    ################## 
    n_cols = set()
    n_rows = set()
    rs = []
    for t, r in page.getbboxlog(): # https://pymupdf.readthedocs.io/en/latest/functions.html#Page.getbboxlog
        if t == 'stroke-path':
            r = fitz.IRect(*r)
            a = f"{r.getarea():.1f}"
            # you need to do some pre-processing to get the "right" values
            if a in {"196.0", "210.0"}:
                w, h = r.width/3, r.height/3
                subr = fitz.Rect(r.x0+w, r.y0+h, r.x1-w, r.y1-h)
                pix = page.getpixmap(colorspace="gray", clip=subr, annots=False)
                isrchecked = pix.isunicolor is False
                ncols.add(r.y0)
                nrows.add(r.x0)
                rs.append(((r.y0, r.x0), isrchecked))
    rs.sort()
    ncols = sorted(ncols)
    nrows = sorted(nrows)
    rs = [((nrows.index(y), ncols.index(x)), v) for (x, y), v in rs]
    d = dict(rs)
    # final df
    ##########
    dfchecked = pd.Series(d).unstack(level=1).T
    mask = dfchecked.fillna(False).astype(bool)
    p = df + np.where(mask, CHECKEDSTATUSMSG, "")
    pdf = pd.DataFrame(p)
    pdf.columns=header
    print(pdf)

Output

             I           II           III
      Column 1     Column 2      Column 3
0  Item 1 [OK]  Item 4 [OK]  Item 10 [OK]
1       Item 2       Item 5       Item 11
2       Item 3       Item 6       Item 12
3               Item 7 [OK]  Item 13 [OK]
4                    Item 8       Item 14
5                    Item 9  Item 15 [OK]

If you don't want None for the missing entries you can specify a default value with

defaultvalue = ''
it.ziplongest(, fillvalue=default_value)