Skip to main content

Table 1 Features extracted during our analysis of images and corresponding descriptions

From: CausalCervixNet: convolutional neural networks with causal insight (CICNN) in cervical cancer cell classification—leveraging deep learning models for enhanced diagnostic accuracy

Feature Type

Feature Name

Description

Shape

Nucleus/Cytoplasm

Area

This quantifies the number of pixels that make up the segmented image of the nucleus and cytoplasm

Nucleus/Cytoplasm

Ratio

This ratio provides insights into the relative sizes of the nucleus and cytoplasm, calculated as Nucleus Area divided by the sum of Nucleus Area and Cytoplasm Area

Nucleus/Cytoplasmic

Perimeter

It measures the length of the perimeter around the object, offering information about its shape and boundaries

Nucleus/Cytoplasm

Roundness

This metric gauge the roundness of the object by comparing its actual area to the area inside a circle defined by its longest diameter

Nucleus/Cytoplasm

Shortest Diameter

This represents the largest diameter a circle can have while being completely inscribed within the object

Nucleus/Cytoplasm

Longest Diameter

This signifies the shortest diameter a circle can have while fully circumscribing the object

Nucleus/Cytoplasm

Elongation

Elongation is determined by the ratio between the shortest and longest diameters of the object, providing insights into its elongated or compact nature

Nucleus Position

This feature assesses how centrally the nucleus is positioned within the cytoplasm, offering information on the object's spatial distribution

Texture

Contrast

It quantifies the local variations in pixel intensity within the image. High contrast values indicate significant variations

Correlation

This feature measures the likelihood of specific pixel pairs occurring, providing insights into the pixel relationship

Energy

Energy is a statistical measure of the image's randomness or entropy, indicating the complexity of pixel patterns

Homogeneity

Homogeneity assesses the similarity of pixel intensities across the image, offering insights into the uniformity of texture