基于深度学习的胸部 X 光图像肺炎分类系统（三）

二分类胸片判断：

1. 数据加载时指定了两类标签

2. 损失函数用了二分类专用的

3. 输出层只有 1 个神经元，用了sigmoid激活函数

4. 预测时用 0.5 作为分类阈值

二分类胸片判断：

import numpy as np
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.metrics import f1_score, roc_auc_score, roc_curve, confusion_matrix, classification_report
from imblearn.over_sampling import RandomOverSampler
import tensorflow as tf
from keras import layers
from keras import models
# 或者更常用的是直接导入Sequential类
from keras.models import Sequential
from keras.preprocessing.image import ImageDataGenerator
import os
import zipfile
import requests
from tensorflow.python.keras.callbacks import EarlyStopping
#  这个代码执行 请切换环境到tf_env
plt.rcParams['font.sans-serif'] = ['SimHei']  # 使用 SimHei 字体
plt.rcParams['axes.unicode_minus'] = False    # 解决负号显示问题
plt.rcParams['font.size'] = 10  # 设置全局字体大小# 数据加载和预处理
def load_data(train_dir, test_dir, val_dir, img_size=(150, 150), batch_size=32):# 数据增强器 - 仅用于训练集train_datagen = ImageDataGenerator(rescale=1. / 255,rotation_range=10,width_shift_range=0.1,height_shift_range=0.1,shear_range=0.1,zoom_range=0.1,horizontal_flip=True)# 验证集和测试集只需要重新缩放val_test_datagen = ImageDataGenerator(rescale=1. / 255)# 加载训练数据train_generator = train_datagen.flow_from_directory(train_dir,target_size=img_size,batch_size=batch_size,class_mode='binary',classes=['NORMAL', 'PNEUMONIA'],shuffle=True)# 加载验证数据val_generator = val_test_datagen.flow_from_directory(val_dir,target_size=img_size,batch_size=batch_size,class_mode='binary',classes=['NORMAL', 'PNEUMONIA'],shuffle=False)# 加载测试数据test_generator = val_test_datagen.flow_from_directory(test_dir,target_size=img_size,batch_size=batch_size,class_mode='binary',classes=['NORMAL', 'PNEUMONIA'],shuffle=False)return train_generator, val_generator, test_generator# 处理样本不均衡（过采样）
def handle_imbalance(generator):# 提取特征和标签X, y = [], []num_batches = len(generator)# 重置生成器以确保从开始获取数据generator.reset()for i in range(num_batches):batch_x, batch_y = generator.next()X.append(batch_x)y.append(batch_y)X = np.concatenate(X)y = np.concatenate(y)# 打印原始分布print(f"原始样本分布: 正常={np.sum(y == 0)}, 肺炎={np.sum(y == 1)}")# 展平特征用于过采样X_flat = X.reshape(X.shape[0], -1)# 过采样少数类ros = RandomOverSampler(random_state=42)X_resampled, y_resampled = ros.fit_resample(X_flat, y)# 恢复图像形状X_resampled = X_resampled.reshape(-1, *X.shape[1:])print(f"过采样后分布: 正常={np.sum(y_resampled == 0)}, 肺炎={np.sum(y_resampled == 1)}")return X_resampled, y_resampled, y# 构建改进的CNN模型
def build_model(input_shape):model = models.Sequential([# 第一个卷积块layers.Conv2D(32, (3, 3), activation='relu', input_shape=input_shape),layers.BatchNormalization(),layers.MaxPooling2D((2, 2)),layers.Dropout(0.2),# 第二个卷积块layers.Conv2D(64, (3, 3), activation='relu'),layers.BatchNormalization(),layers.MaxPooling2D((2, 2)),layers.Dropout(0.3),# 第三个卷积块layers.Conv2D(128, (3, 3), activation='relu'),layers.BatchNormalization(),layers.MaxPooling2D((2, 2)),layers.Dropout(0.4),# 第四个卷积块layers.Conv2D(256, (3, 3), activation='relu'),layers.BatchNormalization(),layers.MaxPooling2D((2, 2)),layers.Dropout(0.5),# 分类器layers.Flatten(),layers.Dense(512, activation='relu'),layers.BatchNormalization(),layers.Dropout(0.5),layers.Dense(1, activation='sigmoid')])# 使用更稳定的优化器optimizer = tf.keras.optimizers.Adam(learning_rate=0.0001)model.compile(optimizer=optimizer,loss='binary_crossentropy',metrics=['accuracy',tf.keras.metrics.Precision(name='precision'),tf.keras.metrics.Recall(name='recall'),tf.keras.metrics.AUC(name='auc')])return model# 主函数
def main():# 假设数据集已经手动下载并解压train_dir = "chest_xray/train"test_dir = "chest_xray/test"val_dir = "chest_xray/val"# 加载数据img_size = (150, 150)batch_size = 32train_generator, val_generator, test_generator = load_data(train_dir, test_dir, val_dir, img_size, batch_size)# 处理样本不均衡X_train, y_train_resampled, y_train_original = handle_imbalance(train_generator)# 计算类别权重（基于原始分布）n_normal = np.sum(y_train_original == 0)n_pneumonia = np.sum(y_train_original == 1)total = n_normal + n_pneumoniaweight_for_normal = (1 / n_normal) * (total / 2.0)weight_for_pneumonia = (1 / n_pneumonia) * (total / 2.0)class_weights = {0: weight_for_normal, 1: weight_for_pneumonia}print(f"类别权重: 正常={weight_for_normal:.2f}, 肺炎={weight_for_pneumonia:.2f}")# 构建模型model = build_model((*img_size, 3))model.summary()# 提前停止回调early_stopping = EarlyStopping(monitor='val_loss',patience=5,restore_best_weights=True,verbose=1)# 训练模型history = model.fit(X_train, y_train_resampled,epochs=30,batch_size=32,validation_data=val_generator,class_weight=class_weights,callbacks=[early_stopping],verbose=1)# 评估模型 - 使用完整测试集test_generator.reset()test_steps = len(test_generator)test_results = model.evaluate(test_generator, steps=test_steps, verbose=1)print("\n测试集评估结果:")print(f"准确率: {test_results[1]:.4f}")print(f"精确率: {test_results[2]:.4f}")print(f"召回率: {test_results[3]:.4f}")print(f"AUC: {test_results[4]:.4f}")# 获取测试集所有预测结果test_generator.reset()y_true = []y_pred_prob = []for i in range(test_steps):batch_x, batch_y = test_generator.next()y_true.extend(batch_y)batch_pred = model.predict(batch_x, verbose=0).ravel()y_pred_prob.extend(batch_pred)y_true = np.array(y_true)y_pred_prob = np.array(y_pred_prob)y_pred = (y_pred_prob > 0.5).astype(int)# 计算额外指标f1 = f1_score(y_true, y_pred)auc = roc_auc_score(y_true, y_pred_prob)print(f"\nF1-score: {f1:.4f}")print(f"AUC-ROC: {auc:.4f}")# 分类报告print("\n分类报告:")print(classification_report(y_true, y_pred, target_names=['NORMAL', 'PNEUMONIA']))# 混淆矩阵cm = confusion_matrix(y_true, y_pred)print("混淆矩阵:")print(cm)# 绘制ROC曲线fpr, tpr, _ = roc_curve(y_true, y_pred_prob)plt.figure(figsize=(10, 6))plt.plot(fpr, tpr, color='darkorange', lw=2, label=f'ROC曲线 (AUC = {auc:.4f})')plt.plot([0, 1], [0, 1], color='navy', lw=2, linestyle='--')plt.xlim([0.0, 1.0])plt.ylim([0.0, 1.05])plt.xlabel('False Positive Rate')plt.ylabel('True Positive Rate')plt.title('接收者操作特征曲线(ROC)')plt.legend(loc="lower right")plt.savefig('roc_curve.png', dpi=300)plt.show()# 绘制训练历史plt.figure(figsize=(12, 8))plt.subplot(2, 2, 1)plt.plot(history.history['accuracy'], label='训练准确率')plt.plot(history.history['val_accuracy'], label='验证准确率')plt.title('准确率')plt.legend()plt.subplot(2, 2, 2)plt.plot(history.history['loss'], label='训练损失')plt.plot(history.history['val_loss'], label='验证损失')plt.title('损失')plt.legend()plt.subplot(2, 2, 3)plt.plot(history.history['precision'], label='训练精确率')plt.plot(history.history['val_precision'], label='验证精确率')plt.title('精确率')plt.legend()plt.subplot(2, 2, 4)plt.plot(history.history['recall'], label='训练召回率')plt.plot(history.history['val_recall'], label='验证召回率')plt.title('召回率')plt.legend()plt.tight_layout()plt.savefig('training_history.png', dpi=300)plt.show()if __name__ == "__main__":main()

这段代码里有很多地方明确体现了这是一个二分类任务（判断 “正常胸片” 和 “肺炎胸片” 两类），最关键的有这几个地方：