Spaces:
Runtime error
Runtime error
File size: 9,506 Bytes
87110a5 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 |
import streamlit as st
import numpy as np
import scipy.stats as stats
from fpdf import FPDF
import base64
import os
from plots import test_profile
import matplotlib.pyplot as plt
from PIL import Image
# Function to calculate z-score
def calculate_z_score(test_score, mean, std_dev):
return (test_score - mean) / std_dev
def z_score_calculator(value, norm_mean, norm_sd):
z_value = (value - norm_mean) / norm_sd
stanine_value = round(1.25 * z_value + 5.5)
z_score = round(z_value, 2)
return z_score, stanine_value
def bnt_calculator(age, education, bnt):
if age <= 60 and education <= 12:
norm_mean = 54.5
norm_sd = 3.2
z_score, stanine_value = z_score_calculator(bnt, norm_mean, norm_sd)
return norm_mean, norm_sd, z_score, stanine_value
elif age <= 60 and education > 12:
norm_mean = 54.0
norm_sd = 4.4
z_score, stanine_value = z_score_calculator(bnt, norm_mean, norm_sd)
return norm_mean, norm_sd, z_score, stanine_value
elif age > 60 and education <= 12:
norm_mean = 54.8
norm_sd = 3.3
z_score, stanine_value = z_score_calculator(bnt, norm_mean, norm_sd)
return norm_mean, norm_sd, z_score, stanine_value
elif age > 60 and education > 12:
norm_mean = 56.2
norm_sd = 3.4
z_score, stanine_value = z_score_calculator(bnt, norm_mean, norm_sd)
return norm_mean, norm_sd, z_score, stanine_value
else:
print("missing value/ wrong format")
def fas_calculator(age, education, fas):
if age <= 60 and education <= 12:
norm_mean = 42.7
norm_sd = 13.7
z_score, stanine_value = z_score_calculator(fas, norm_mean, norm_sd)
return norm_mean, norm_sd, z_score, stanine_value
elif age <= 60 and education > 12:
norm_mean = 46.7
norm_sd = 13.7
z_score, stanine_value = z_score_calculator(fas, norm_mean, norm_sd)
return norm_mean, norm_sd, z_score, stanine_value
elif age > 60 and education <= 12:
norm_mean = 46.9
norm_sd = 10.4
z_score, stanine_value = z_score_calculator(fas, norm_mean, norm_sd)
return norm_mean, norm_sd, z_score, stanine_value
elif age > 60 and education > 12:
norm_mean = 51.6
norm_sd = 12.6
z_score, stanine_value = z_score_calculator(fas, norm_mean, norm_sd)
return norm_mean, norm_sd, z_score, stanine_value
else:
print("missing value/ wrong format")
def generate_graph(BNT_stanine, FAS_stanine):
# Create a plot
fig, ax = plt.subplots()
# Set axis labels and title
ax.set_xlabel('Stanine values')
ax.set_ylabel('Test')
# Set the y-axis to display the tests
ax.set_yticks([1, 2])
ax.set_yticklabels(['BNT', 'FAS'])
# Set the range of the x-axis
ax.set_xlim([0, 10])
# Add dots for BNT and FAS scores
ax.scatter(BNT_stanine, 1, s=100, label='BNT')
ax.scatter(FAS_stanine, 2, s=100, label='FAS')
# Add legend
ax.legend()
# Show the plot
# plt.show()
# Save the graph as a png file
fig.savefig('test_profile.png')
return 'test_profile.png'
def create_pdf(z_score, mean, std_dev, logo_path, plot_path):
pdf = FPDF()
pdf.add_page()
pdf.set_xy(0, 0)
pdf.set_font("Arial", size=12)
# Add logos
x_positions = [25, 85, 145]
for i, logo_path in enumerate(logo_paths):
pdf.image(logo_path, x=x_positions[i], y=8, w=40)
pdf.set_xy(10, 50)
# Add title and center it
title = "Z-Score Report"
pdf.set_font("Arial", style="B", size=16)
title_width = pdf.get_string_width(title) + 6
pdf.cell((210 - title_width) / 2)
pdf.cell(title_width, 10, title, 0, 1, "C")
# Add z-score and center it
pdf.set_font("Arial", size=12)
z_score_text = "Your z-score is: {:.2f}".format(z_score)
z_score_width = pdf.get_string_width(z_score_text) + 6
pdf.cell((210 - z_score_width) / 2)
pdf.cell(z_score_width, 10, z_score_text, 0, 1, "C")
# Add mean and standard deviation and center it
mean_std_text = "Mean: {}, Standard Deviation: {}".format(mean, std_dev)
mean_std_width = pdf.get_string_width(mean_std_text) + 6
pdf.cell((210 - mean_std_width) / 2)
pdf.cell(mean_std_width, 10, mean_std_text, 0, 1, "C")
# Add logo
pdf.image(plot_path, x=10, y=80, w=200)
# pdf.set_xy(10, 40)
# Add tool description and center it
pdf.set_xy(10, 230)
pdf.set_font("Arial", size=10)
description = "This PDF report was generated using the Z-Score Calculator Streamlit App."
pdf.multi_cell(0, 10, description, 0, "C")
# Add explanatory text about the collaboration between KI and KTH
pdf.set_xy(10, 255)
pdf.set_font("Arial", size=8)
collaboration_text = (
"Den här PDF:en är en del av ett samarbetsprojekt mellan Karolinska Institutet (KI) och "
"Kungliga Tekniska Högskolan (KTH) med målsättningen att använda artificiell intelligens (AI) och "
"teknik för att minska administration i sjukhusarbete. Projektet fokuserar på att utveckla och "
"implementera AI-baserade lösningar för att förbättra arbetsflöden, öka effektiviteten och "
"minska den administrativa bördan för sjukvårdspersonal. För frågor om formuläret kontakta Fredrik Sand fredrik.sand-aronsson@regionstockholm.se, för frågor om teknik kontakta Birger Moëll bmoell@kth.se."
)
line_width = 190
line_height = pdf.font_size_pt * 0.6
lines = collaboration_text.split(' ')
current_line = ''
for word in lines:
if pdf.get_string_width(current_line + word) < line_width:
current_line += word + ' '
else:
pdf.cell(line_width, line_height, current_line, 0, 1)
current_line = word + ' '
pdf.cell(line_width, line_height, current_line, 0, 1)
return pdf
def pdf_to_base64(pdf):
with open(pdf, "rb") as file:
return base64.b64encode(file.read()).decode('utf-8')
# Title and description
st.title("Z-Score Calculator")
st.write("Enter your test score, age, and education level to calculate the z-score.")
# Input fields
#test_score = st.number_input("Test Score", min_value=0, value=0, step=1)
age = st.number_input("Age", min_value=0, value=18, step=1)
education_level = st.number_input("Education Level in years", min_value=0, value=18, step=1)
isw = st.number_input("ISW", min_value=0, value=0, step=1)
bnt = st.number_input("BNT", min_value=0, value=0, step=1)
fas = st.number_input("FAS", min_value=0, value=0, step=1)
animal = st.number_input("Animal", min_value=0, value=0, step=1)
verb = st.number_input("Verb", min_value=0, value=0, step=1)
repetition = st.number_input("Repetition", min_value=0, value=0, step=1)
logicogrammatic = st.number_input("Logicogrammatic", min_value=0, value=0, step=1)
inference = st.number_input("Inference", min_value=0, value=0, step=1)
reading_speed = st.number_input("Reading Speed", min_value=0, value=0, step=1)
decoding_words = st.number_input("Decoding Words", min_value=0, value=0, step=1)
decoding_non_words = st.number_input("Decoding Non-Words", min_value=0, value=0, step=1)
months_backward = st.number_input("Months Backward", min_value=0, value=0, step=1)
pataka = st.number_input("Pataka", min_value=0, value=0, step=1)
# add all the tests
# Calculate mean and standard deviation based on age and education level
# For simplicity, we will use made-up values for mean and std_dev
mean = np.random.randint(50, 100)
std_dev = np.random.randint(10, 30)
# Calculate z-score and display result
if st.button("Calculate Z-Score"):
profile = test_profile(age, education_level, isw, bnt, fas)
# for each value in the profile, calculate the z-score
bnt_mean, bnt_std, z_bnt, stanine_bnt = bnt_calculator(age, education_level, bnt)
fas_mean, fas_std, z_fas, stanine_fas = fas_calculator(age, education_level, fas)
# z_score = calculate_z_score(test_score, mean, std_dev)
st.write(f"Your bnt z-score is: {z_bnt:.2f}")
st.write(f"Mean: {bnt_mean}, Standard Deviation: {bnt_std}")
st.write(f"Your fas z-score is: {z_fas:.2f}")
st.write(f"Mean: {fas_mean}, Standard Deviation: {fas_std}")
# Create PDF
# logo_path="logo.jpg"
logo_paths = ["logo.jpg", "logo2.jpg", "logo3.jpg"]
# create the plot from the dataframe
# check if education level is more than 12 years, if more than 12, set value to one, otherwise zero
education_level = 1 if education_level > 12 else 0
plot_path = generate_graph(stanine_bnt, stanine_fas)
# create an image from the plot and add to streamlit display
image = Image.open(plot_path)
st.image(image, caption='Stanine plot', use_column_width=True)
pdf_filename = "z_score_report.pdf"
pdf = create_pdf(z_bnt, bnt_mean, bnt_std, logo_paths, plot_path)
pdf.output(name=pdf_filename)
# Download PDF
with open(pdf_filename, "rb") as file:
base64_pdf = base64.b64encode(file.read()).decode('utf-8')
pdf_display = f'<a href="data:application/octet-stream;base64,{base64_pdf}" download="{pdf_filename}">Download PDF</a>'
st.markdown(pdf_display, unsafe_allow_html=True)
# Remove PDF file after download
if os.path.exists(pdf_filename):
os.remove(pdf_filename) |