MRI and Alzheimers

아래 출처를 통해, Data 분석 및 XGBoost 진행하였음

아직 이해하지 못하거나, 개념적인 부분이 미흡함

개념 부분을 공부해서 정리할 예정

출처 : https://www.kaggle.com/code/shreyaspj/alzheimer-s-analysis-using-mri?scriptVersionId=74709903&cellId=2

## MRI and Alzheimers
# 데이터출처 : https://www.kaggle.com/datasets/jboysen/mri-and-alzheimers

In [2]:

import tensorflow.compat.v1 as tf
from sklearn.metrics import confusion_matrix
import numpy as np
from scipy.io import loadmat
import os
from pywt import wavedec
from functools import reduce
from scipy import signal
from scipy.stats import entropy
from scipy.fft import fft, ifft
import pandas as pd
from sklearn.model_selection import train_test_split, StratifiedKFold
from sklearn.model_selection import RandomizedSearchCV
from sklearn.preprocessing import StandardScaler
from tensorflow import keras as K
import matplotlib.pyplot as plt
import scipy
from sklearn import metrics
from sklearn.ensemble import RandomForestClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score
from sklearn.model_selection import KFold,cross_validate
from tensorflow.keras.layers import Dense, Activation, Flatten, concatenate, Input, Dropout, LSTM, Bidirectional,BatchNormalization,PReLU,ReLU,Reshape
from sklearn.metrics import classification_report
from tensorflow.keras.models import Sequential, Model, load_model
import matplotlib.pyplot as plt;
from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint
from sklearn.decomposition import PCA
from sklearn.model_selection import cross_val_score
from tensorflow.keras.layers import Conv1D,Conv2D,Add
from tensorflow.keras.layers import MaxPool1D, MaxPooling2D
import seaborn as sns

In [3]:

# os.environ을 이용하여 Kaggle API Username, Key 세팅하기
os.environ['KAGGLE_USERNAME'] = ''
os.environ['KAGGLE_KEY'] = ''

In [4]:

# Linux 명령어로 Kaggle API를 이용하여 데이터셋 다운로드하기 (!kaggle ~)
# Linux 명령어로 압축 해제하기
!kaggle datasets download -d jboysen/mri-and-alzheimers
!unzip '*.zip'

mri-and-alzheimers.zip: Skipping, found more recently modified local copy (use --force to force download)
Archive:  mri-and-alzheimers.zip
replace oasis_cross-sectional.csv? [y]es, [n]o, [A]ll, [N]one, [r]ename: 

In [13]:

!ls

mri-and-alzheimers.zip	   oasis_longitudinal.csv
oasis_cross-sectional.csv  sample_data

In [14]:

# pd.read_csv()로 csv파일 읽어들이기
data_cross = pd.read_csv("oasis_cross-sectional.csv")
data_long = pd.read_csv("oasis_longitudinal.csv")
print(data_cross.info())
print(data_long.info())

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 436 entries, 0 to 435
Data columns (total 12 columns):
 #   Column  Non-Null Count  Dtype  
---  ------  --------------  -----  
 0   ID      436 non-null    object 
 1   M/F     436 non-null    object 
 2   Hand    436 non-null    object 
 3   Age     436 non-null    int64  
 4   Educ    235 non-null    float64
 5   SES     216 non-null    float64
 6   MMSE    235 non-null    float64
 7   CDR     235 non-null    float64
 8   eTIV    436 non-null    int64  
 9   nWBV    436 non-null    float64
 10  ASF     436 non-null    float64
 11  Delay   20 non-null     float64
dtypes: float64(7), int64(2), object(3)
memory usage: 41.0+ KB
None
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 373 entries, 0 to 372
Data columns (total 15 columns):
 #   Column      Non-Null Count  Dtype  
---  ------      --------------  -----  
 0   Subject ID  373 non-null    object 
 1   MRI ID      373 non-null    object 
 2   Group       373 non-null    object 
 3   Visit       373 non-null    int64  
 4   MR Delay    373 non-null    int64  
 5   M/F         373 non-null    object 
 6   Hand        373 non-null    object 
 7   Age         373 non-null    int64  
 8   EDUC        373 non-null    int64  
 9   SES         354 non-null    float64
 10  MMSE        371 non-null    float64
 11  CDR         373 non-null    float64
 12  eTIV        373 non-null    int64  
 13  nWBV        373 non-null    float64
 14  ASF         373 non-null    float64
dtypes: float64(5), int64(5), object(5)
memory usage: 43.8+ KB
None

In [15]:

print(data_cross.isna().sum())
print("\n")
data_long.isna().sum()

ID         0
M/F        0
Hand       0
Age        0
Educ     201
SES      220
MMSE     201
CDR      201
eTIV       0
nWBV       0
ASF        0
Delay    416
dtype: int64

Out[15]:

Subject ID     0
MRI ID         0
Group          0
Visit          0
MR Delay       0
M/F            0
Hand           0
Age            0
EDUC           0
SES           19
MMSE           2
CDR            0
eTIV           0
nWBV           0
ASF            0
dtype: int64

In [16]:

data_cross.dropna(subset=['CDR'],inplace=True)

In [17]:

data_cross.drop(columns=['ID','Delay'],inplace=True)
data_long = data_long.rename(columns={'EDUC':'Educ'})
data_long.drop(columns=['Subject ID','MRI ID','Group','Visit','MR Delay'],inplace=True)

In [18]:

data = pd.concat([data_cross,data_long], ignore_index = True)
data.head()

Out[18]:

M/FHandAgeEducSESMMSECDReTIVnWBVASF01234

F	R	74	2.0	3.0	29.0	0.0	1344	0.743	1.306
F	R	55	4.0	1.0	29.0	0.0	1147	0.810	1.531
F	R	73	4.0	3.0	27.0	0.5	1454	0.708	1.207
M	R	74	5.0	2.0	30.0	0.0	1636	0.689	1.073
F	R	52	3.0	2.0	30.0	0.0	1321	0.827	1.329

In [19]:

data.describe()

Out[19]:

AgeEducSESMMSECDReTIVnWBVASFcountmeanstdmin25%50%75%max

608.000000	608.000000	570.00000	606.000000	608.000000	608.000000	608.00000	608.000000
75.208882	10.184211	2.47193	27.234323	0.288651	1477.062500	0.73713	1.203597
9.865026	6.058388	1.12805	3.687980	0.377697	170.653795	0.04267	0.135091
33.000000	1.000000	1.00000	4.000000	0.000000	1106.000000	0.64400	0.876000
70.000000	4.000000	2.00000	26.000000	0.000000	1352.500000	0.70400	1.118000
76.000000	12.000000	2.00000	29.000000	0.000000	1460.000000	0.73600	1.202000
82.000000	16.000000	3.00000	30.000000	0.500000	1569.000000	0.76625	1.297500
98.000000	23.000000	5.00000	30.000000	2.000000	2004.000000	0.84700	1.587000

In [20]:

cor = data.corr()
plt.figure(figsize=(12,9))
sns.heatmap(cor, xticklabels=cor.columns.values,yticklabels=cor.columns.values, annot=True)

Out[20]:

<matplotlib.axes._subplots.AxesSubplot at 0x7f6c54220710>

In [29]:

data.isna().sum()

Out[29]:

M/F      0
Hand     0
Age      0
Educ     0
SES     38
MMSE     2
CDR      0
eTIV     0
nWBV     0
ASF      0
dtype: int64

In [30]:

from sklearn.impute  import SimpleImputer
imputer = SimpleImputer ( missing_values = np.nan,strategy='most_frequent')

imputer.fit(data[['SES']])
data[['SES']] = imputer.fit_transform(data[['SES']])

# We perform it with the median
imputer = SimpleImputer ( missing_values = np.nan,strategy='median')

imputer.fit(data[['MMSE']])
data[['MMSE']] = imputer.fit_transform(data[['MMSE']])

In [31]:

data

Out[31]:

M/FHandAgeEducSESMMSECDReTIVnWBVASF01234...603604605606607

F	R	74	2.0	3.0	29.0	0.0	1344	0.743	1.306
F	R	55	4.0	1.0	29.0	0.0	1147	0.810	1.531
F	R	73	4.0	3.0	27.0	0.5	1454	0.708	1.207
M	R	74	5.0	2.0	30.0	0.0	1636	0.689	1.073
F	R	52	3.0	2.0	30.0	0.0	1321	0.827	1.329
...	...	...	...	...	...	...	...	...	...
M	R	82	16.0	1.0	28.0	0.5	1693	0.694	1.037
M	R	86	16.0	1.0	26.0	0.5	1688	0.675	1.040
F	R	61	13.0	2.0	30.0	0.0	1319	0.801	1.331
F	R	63	13.0	2.0	30.0	0.0	1327	0.796	1.323
F	R	65	13.0	2.0	30.0	0.0	1333	0.801	1.317

608 rows × 10 columns

In [33]:

from sklearn import preprocessing
le = preprocessing.LabelEncoder()
data['CDR'] = le.fit_transform(data['CDR'].values)

In [34]:

data = pd.get_dummies(data)

In [35]:

plt.bar(data['CDR'].unique(),data['CDR'].value_counts())

Out[35]:

<BarContainer object of 4 artists>

In [36]:

data = data.drop(data[data['CDR']==3].index)
data.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 603 entries, 0 to 607
Data columns (total 11 columns):
 #   Column  Non-Null Count  Dtype  
---  ------  --------------  -----  
 0   Age     603 non-null    int64  
 1   Educ    603 non-null    float64
 2   SES     603 non-null    float64
 3   MMSE    603 non-null    float64
 4   CDR     603 non-null    int64  
 5   eTIV    603 non-null    int64  
 6   nWBV    603 non-null    float64
 7   ASF     603 non-null    float64
 8   M/F_F   603 non-null    uint8  
 9   M/F_M   603 non-null    uint8  
 10  Hand_R  603 non-null    uint8  
dtypes: float64(5), int64(3), uint8(3)
memory usage: 44.2 KB

In [37]:

y = data.pop('CDR')
x = data

In [38]:

x = StandardScaler().fit_transform(x)

In [39]:

X_train, X_test, y_train, y_test = train_test_split(x, y, random_state = 42,test_size=0.3)

In [40]:

from xgboost import XGBClassifier
FOLDS = 5

parametros_xgb = {
    "learning_rate": [0.01, 0.025, 0.005,0.5, 0.075, 0.1, 0.15, 0.2,0.3,0.8,0.9],
    "max_depth":[3,5,8,10,15,20,25,30,40,50],
    "n_estimators":range(1,1000)
    }

model_xgb= XGBClassifier(eval_metric='mlogloss')


xgb_random = RandomizedSearchCV(estimator = model_xgb, param_distributions = parametros_xgb, n_iter = 100, cv = FOLDS, 
                               verbose=0, random_state=42,n_jobs = -1, scoring='accuracy')
xgb_random.fit(X_train, y_train)

xgb_random.best_params_

Out[40]:

{'learning_rate': 0.5, 'max_depth': 30, 'n_estimators': 103}

In [41]:

model_xgb = xgb_random.best_estimator_
model_xgb.fit(X_train,y_train)
model_xgb.score(X_test,y_test)

Out[41]:

0.7845303867403315

In [43]:

from sklearn.ensemble import GradientBoostingClassifier
FOLDS =10

parametros_gb = {
    "loss":["deviance"],
    "learning_rate": [0.01, 0.025, 0.005,0.5, 0.075, 0.1, 0.15, 0.2,0.3,0.8,0.9],
    "min_samples_split": [0.01, 0.025, 0.005,0.4,0.5, 0.075, 0.1, 0.15, 0.2,0.3,0.8,0.9],
    "min_samples_leaf": [1,2,3,5,8,10,15,20,40,50,55,60,65,70,80,85,90,100],
    "max_depth":[3,5,8,10,15,20,25,30,40,50],
    "max_features":["log2","sqrt"],
    "criterion": ["friedman_mse",  "mae"],
    "subsample":[0.5, 0.618, 0.8, 0.85, 0.9, 0.95, 1.0],
    "n_estimators":range(1,100)
    }

model_gb= GradientBoostingClassifier()


gb_random = RandomizedSearchCV(estimator = model_gb, param_distributions = parametros_gb, n_iter = 100, cv = FOLDS, 
                               verbose=0, random_state=42,n_jobs = -1, scoring='accuracy')
gb_random.fit(X_train, y_train)

gb_random.best_params_

 

Out[43]:

{'criterion': 'mae',
 'learning_rate': 0.8,
 'loss': 'deviance',
 'max_depth': 20,
 'max_features': 'log2',
 'min_samples_leaf': 1,
 'min_samples_split': 0.025,
 'n_estimators': 68,
 'subsample': 0.95}

In [44]:

model_gb = gb_random.best_estimator_
print(model_gb.score(X_test,y_test))

0.8232044198895028

In [45]:

cross_val_score(model_gb, x, y, cv=10, scoring='accuracy').mean()

 

Out[45]:

0.7856284153005465

In [46]:

y_pred = model_gb.predict(X_test)

cm = confusion_matrix(y_test, y_pred)
clr = classification_report(y_test, y_pred)

plt.figure(figsize=(8, 8))
sns.heatmap(cm, annot=True, vmin=0, fmt='g', cbar=False, cmap='Blues')
plt.xlabel("Predicted")
plt.ylabel("Actual")
plt.title("Confusion Matrix")
plt.show()

print("Classification Report:\n----------------------\n", clr)

Classification Report:
----------------------
               precision    recall  f1-score   support

           0       0.85      0.93      0.89        96
           1       0.77      0.73      0.75        63
           2       0.88      0.64      0.74        22

    accuracy                           0.82       181
   macro avg       0.83      0.76      0.79       181
weighted avg       0.82      0.82      0.82       181