microporella

Extraction of traits from morphological modules of Microporella species

Images

The images are SEM images of Microporella collected from New Zealand. These specimens are both modern and paleo (~2.3 Mya).

The images were collected by K. Voje, L.H. Liow, E. Di Martino, and others as part of the WABO expeditions. The specimens were imaged by M. Ramsfjell and E. Di Martino.

Images are stored on a shared lab computer and will be made available with the publication of this project.

Metadata

Imaging metadata

The metadata file "Microporella_SEMs_EDM+Mali_05.06.2024.csv" contains information about:

• Date: date of image in MM/DD/YYY format
• Image_ID: a unique specimen number
• Formation: formation from which specimens came
• Age: age of the formation in stages, not years
• Sample_ID: unique number assigned to collection sample
• Shell: unique number assigned to shell within collection sample (Sample_ID)
• Colony: unique number assigned to the colony on the shell (can be multiple on one shell)
• Genus: genus of the binomial
• Species: species of the binomial

We also compared the images to those used in Liow et al. 2024 (see dataset here)

Traits

We extracted linear measurements from landmarks images of zooids.

The image below is from Di Martino et al. 2023 and is of Microporella discors.

There are a total of 22 landmarks, numbered 1 to 14, 1_O to 4_O, and 1_A to 1_A.

The measurements were based off Di Martino & Liow 2022 and Schack et al. 2020.

Landmarks

Ovicell (green shading):

• 1_V: centroid

Autozooid (yellow shading):

• 1_U: centroid

Ascopore (lime green shading):

• 1_P: centroid

Operculum (pink shading):

• 1_O: centroid

Avicularia (purple shading):

• 1_A: centroid

Measurements

Linear measurements were extracted using dimensions of the mask (see DeepBryo_micro).

Ovicell (green shading):
 Shape:
  maximum width
  maximum length
  area

Autozooid (yellow shading):
 Shape:
  zooid length (height)
  zooid width
  area

Ascopore (lime green shading, black lines):
 Shape:
  area
 Position on autozooid
  distance from distal wall (intersection between vertical midline of ascopore mask and autozooid mask)
  distance from lateral wall (intersection between horizontal midline of aspcopore mask and autozooid mask)

Operculum (pink shading):
 Shape:
  area
  length
  width
 Amount covered by ovicell

Avicularia (purple shading):
 Shape:
  length
  height
  area
 Position on autozooid
  distance from distal wall (intersection between vertical midline of avicularium mask and autozooid mask)
  distance from lateral wall (intersection between horizontal midline of avicularium mask and autozooid mask

Automation

We use DeepBryo, a tool developed by Di Martino et al. 2023 and which we forked for our project (DeepBryo_micro), to extract measurements. This code provides segmentation of morphological features of Microporella colonies. We modified the code to also output minimum bounding box and polygon coordinates to: extract relative position of avicularia and ascopores on autozooids, match masks of the ascopore, avicularia, operculum, and ovicell to the autozooid, check for any errors in segementation.

The output of the machine learning pipeline is a csv file of (trimmed to the columns of interest):

• index of the structure id
• image_id: image name
• category: type of structure (i.e., autozooid, ascopore, operculum, avicularia, or ovicell)
• area: area of mask
• circularity
• majorAxis: height of structure
• minorAxis: width of structure
• center_x: x-axis for centroid
• center_y: y-axis for centroid
• polygon: json of polygon vertices
• min_bbox_points: bbox coordinates for top, left, bottom, right
• unit: if pixels or scaled

Data processing

We check the metadata for every image in the scripts microporella_imageMetadata.R and microporella_metadata.R. (See note on processing)

In the code, "fileNames.R", reads in the image names and associated metadata file name and creates the dataset, "image.filter.csv".