Implements the UNet model with 3D Convolutions.

UNet3D.py

def unrolled_UNet3D(inputs):
    """
    Implements the UNet model with 3D Convolutions. Refer to the following
    source code: https://github.com/jocicmarko/ultrasound-nerve-segmentation.
    """
    
    conv1 = Conv3D( 32, (3, 3), activation='relu', padding='same')(inputs)
    conv1 = Conv3D( 32, (3, 3), activation='relu', padding='same')(conv1)
    pool1 = MaxPooling3D(pool_size=(2, 2))(conv1)

    conv2 = Conv3D( 64, (3, 3), activation='relu', padding='same')(pool1)
    conv2 = Conv3D( 64, (3, 3), activation='relu', padding='same')(conv2)
    pool2 = MaxPooling3D(pool_size=(2, 2))(conv2)

    conv3 = Conv3D(128, (3, 3), activation='relu', padding='same')(pool2)
    conv3 = Conv3D(128, (3, 3), activation='relu', padding='same')(conv3)
    pool3 = MaxPooling3D(pool_size=(2, 2))(conv3)

    conv4 = Conv3D(256, (3, 3), activation='relu', padding='same')(pool3)
    conv4 = Conv3D(256, (3, 3), activation='relu', padding='same')(conv4)
    pool4 = MaxPooling3D(pool_size=(2, 2))(conv4)

    conv5 = Conv3D(512, (3, 3), activation='relu', padding='same')(pool4)
    conv5 = Conv3D(512, (3, 3), activation='relu', padding='same')(conv5)

    # residual connection
    rsdc6 = concatenate([Conv3DTranspose(256, (2, 2), strides=(2, 2), padding='same')(conv5), conv4], axis=3)
    
    conv6 = Conv3D(256, (3, 3), activation='relu', padding='same')(rsdc6)
    conv6 = Conv3D(256, (3, 3), activation='relu', padding='same')(conv6)

    # residual connection
    rsdc7 = concatenate([Conv3DTranspose(128, (2, 2), strides=(2, 2), padding='same')(conv6), conv3], axis=3)
    
    conv7 = Conv3D(128, (3, 3), activation='relu', padding='same')(rsdc7)
    conv7 = Conv3D(128, (3, 3), activation='relu', padding='same')(conv7)

    # residual connection
    rsdc8 = concatenate([Conv3DTranspose( 64, (2, 2), strides=(2, 2), padding='same')(conv7), conv2], axis=3)
    
    conv8 = Conv3D(64, (3, 3), activation='relu', padding='same')(rsdc8)
    conv8 = Conv3D(64, (3, 3), activation='relu', padding='same')(conv8)

    # residual connection
    rsdc9 = concatenate([Conv3DTranspose( 32, (2, 2), strides=(2, 2), padding='same')(conv8), conv1], axis=3)
    
    conv9 = Conv3D(32, (3, 3), activation='relu', padding='same')(rsdc9)
    conv9 = Conv3D(32, (3, 3), activation='relu', padding='same')(conv9)

    # sigmoid activation
    outputs = Conv3D(1, (1, 1), activation='sigmoid')(conv9)

    model = Model(inputs=[inputs], outputs=[outputs])
    model.compile(optimizer=Adam(lr=1e-5), loss=dice_coef_loss, metrics=[dice_coef])

    return model