Automating Cell Annotation

Automated cell type annotation using a convolutional autoencoder.

This work was presented and published in 2022 44th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC).

(click image to open in new window)

Convolutional Autoencoder Structure



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

 
    nFeatures = x_train.shape[1]
    model = None
    input_ae = Input(shape=(nFeatures, 1))

    x = Conv1D(16, 3, activation="relu", padding="same")(input_ae)
    x = Conv1D(16, 3, activation="relu", padding="same")(x)
    x = MaxPooling1D()(x)
    x = Conv1D(32, 3, activation="relu", padding="same")(x)
    x = Conv1D(32, 3, activation="relu", padding="same")(x)
    x = Dropout(0.7)(x)
    encoded = MaxPooling1D()(x)

    x = Conv1D(32, 3, activation="relu", padding="same")(encoded)
    x = Conv1D(32, 3, activation="relu", padding="same")(x)
    x = UpSampling1D()(x)
    x = Conv1D(16, 3, activation="relu", padding="same")(x)
    x = Conv1D(16, 3, activation="relu", padding="same")(x)
    x = UpSampling1D()(x)
    decoded = Conv1D(1, 3, activation='sigmoid', padding='same', name='ae_out')(x)

    y = Flatten()(encoded)
    y = Dense(64, activation = 'relu')(y)
    y = Dense(24, activation = 'relu')(y)

    output_class = Dense(4, activation = 'softmax', name = "class_out")(y)