Saturday, October 22, 2011

This is a small script to visualize Medtronic / Minimed CareLink blood glucose measurement data that has been exported to a CSV file to produce a nice PDF graph with gnuplot, like this: Plot
#!/usr/bin/env python

import sys
import csv
import datetime
import subprocess

# number of days to plot, going back from today
num_days = 7

if len(sys.argv) != 2:
    print 'syntax: %s input.csv' % sys.argv[0]

start_date = - datetime.timedelta(num_days)

out_files = {}
averages = {}

def filename(y, m, d):
    return '%02d-%02d-%02d.dat' % (y, m, d)

with open(sys.argv[1]) as f:
    # skip 11 header lines
    for i in range(11):

    for row in csv.DictReader(f):
        (d, m, y) = (int(x) for x in row['Date'].split('/'))
        date = + 2000, m, d)
        if date >= start_date:
            # Taking the raw value is more precise
            #sg = row['Sensor Glucose (mmol/L)']
            if row['Raw-Type'] == 'GlucoseSensorData':
                raw = row['Raw-Values']
                assert raw.startswith('AMOUNT=')
                sg = int(raw[raw.find('=') + 1:raw.find(',')]) / 18.

                if (y, m, d) in out_files:
                    out = out_files[(y, m, d)]
                    out = out_files[(y, m, d)] = open(filename(y, m, d), 'wt')

                time = row['Time']
                (hours, minutes) = int(time[:2]), int(time[3:5])
                minutes -= minutes % 5 # to make windows overlap
                if (hours, minutes) in averages:
                    averages[(hours, minutes)][0] += sg
                    averages[(hours, minutes)][1] += 1
                    averages[(hours, minutes)] = [sg, 1]

                print >> out, row['Time'], sg

# close output files
for f in out_files.values():

# write averages
average_file = 'average.dat'
avg, avg_n = 0., 0
with open(average_file, 'wt') as f:
    for ((hours, minutes), (sigma, n)) in sorted(averages.items()):
        val = float(sigma) / n
        avg += val
        avg_n += 1
        print >> f, "%02d:%02d:00 %f" % (hours, minutes, val)
avg /= avg_n

# create gnuplot script
plot_file = 'plot.gnuplot'
eps_file = 'plot.eps'
with open(plot_file, 'wt') as f:
    plots = ', '.join('"%s" using 1:2 linewidth 2 title "%02d/%02d/%02d"' % (filename(y, m, d), d, m, y) for (y, m, d) in sorted(out_files.keys()))
    print >> f, '''
set terminal postscript eps color solid
set output "%s"
set style data lines
set title "Sensor glucose (mmol/L)"
set xdata time
set timefmt "%%H:%%M:%%S"
set format x "%%H:%%M"
set yrange [2:12]
set ytics 1
plot 4 notitle linewidth 3 linecolor rgb "black", 9 notitle linewidth 3 linecolor rgb "black", %f notitle linewidth 5 linecolor rgb "red", "%s" using 1:2 title "average" linewidth 5 linecolor rgb "red",''' % (eps_file, avg, average_file), plots

# Run gnuplot, and convert eps to pdf
subprocess.check_call('gnuplot ' + plot_file, shell=True)'epstopdf ' + eps_file, shell=True)


berkes said...

Is the export made on a linux machine? Can you actually access the contents on this CareLink USB under Linux? If so, any hints?

leo said...

berkes, unfortunately the data was read from the device on Windows. However, let me know in case you're interested in joining a development effort to get the reading working under Linux. I know a few guys who are trying to get this running, but unfortunately it's far from trivial and lots of work still needs to be done.

Kamil Szot said...

leo, I want to ask you for a favour but I don't see any way to contact you directly.

Could you contact me at ?