Quantitative Analysis of Fish Morphology Through Landmark and Outline-based Geometric Morphometrics with Free Software

Software and datasets

1. Software:

   1. tpsDig2 Version 2.32 (https://www.sbmorphometrics.org/soft-dataacq.html, accessed January 31, 2023)

   2. tpsUtil Version 1.82 (https://www.sbmorphometrics.org/soft-tps.html, accessed January 31, 2023)

   3. tpsRelw Version 1.75 (https://www.sbmorphometrics.org/soft-utility.html, accessed January 31, 2023)

   4. ImageJ 1.54i (https://imagej.net/ij/download.html, accessed March 13, 2023) [16]

   5. MorphoJ Version 1.08.01 (https://morphometrics.uk/MorphoJ_page.html)

   6. R programs (https://cran.r-project.org)

   7. RStudio (https://posit.co/)

   8. R package of Momocs (https://cran.r-project.org/web/packages/Momocs/index.html)

   This protocol is running on Windows 11 (64-bit). Taking into account the compatibility of the operating system, the software tutorial, and the possible requirement of preinstalling a version of Java, please download and install all the necessary software: tpsUtil version 1.82, tpsDig2 version 2.32, tpsRelw version 1.75 [17], MorphoJ version 1.08.01 [18], R version 4.3.2 [19], and RStudio 2023.09.1 [20].

2. Website

   1. Pixelcut (https://create.pixelcut.ai/background-remover)

   2. Photoroom (https://www.photoroom.com/tools/background-remover)

   The two AI-based background-remover tools are used to extract fish by removing the image background.

3. Digitized images

   Four groups of largemouth bass (Micropterus salmoides) images were utilized in this study. Two groups were obtained by photographing fish cultured in farm ponds located in Foshan city, while the other two groups of images were sourced from the internet. When photographing the fish, the digital camera was fixed in position with the lens perpendicular to the ground. The fish was placed horizontally on a solid-colored background directly beneath the camera. If necessary, soft materials were used to adjust the position of the fish to ensure its body axis was horizontal and the head was facing left. The photos were taken in macro mode after focusing and were stored in .jpeg format. The size of the photos depended on the camera's capabilities, with sizes between 2 and 10 MB considered appropriate. The internet images were sourced from Microsoft Bing Images (https://cn.bing.com/images/feed?form=Z9LH, accessed April 4, 2024) and Google Images (https://www.google.com.hk/imghp?hl=en&ogbl, accessed April 4, 2024) using the query terms “largemouth bass,” “Micropterus salmoides,” and “largemouth bass (Micropterus salmoides)”. All images used in the current experiment included fish with sufficient resolution, showing a normal appearance and an integrated outline in left/right lateral views [21]. For self-captured JPEG digital photographs, the file size was greater than 2266 KB. The internet-sourced images, whether in .jpg, .png, or other file formats, were converted to .jpeg (.jpg) format, with a minimum size of 14 KB used in the present research. Finally, four groups of mature largemouth bass images were included in this research: LB-FF-FF (feeding with frozen bait, n = 44), LB-FF-FS (feeding with artificial feed, n = 42), LB-IN-DH (internet-sourced realistic painting, n = 23), and LB-IN-PH (internet-sourced picture, n = 30).

Procedure

The procedure for conducting GM primarily involves the following steps: data collection, landmark placement, digitization, Procrustes superimposition, shape analysis, statistical testing, visualization, interpretation, and biological inference.

1. Digitization of fish image data through landmarking and file format conversion

   1. Image preparation

      Typically, original images with scale in .jpeg format can be used for landmarking. To maintain consistency, the background of the images was first removed using an AI-based online tool, and then the images were used for placing landmarks and extracting outlines.

      The two online tools (Pixelcut and Photoroom, accessed April 7, 2024) are both image background removers that offer the free function of downloading background-removed images at standard resolution. The resolution meets the needs of the following analyses. By default, the background is set to be transparent, and the image data quality is determined by the original image.

      For Pixelcut: Open the website → Upload image → Download → Download standard quality (1,080 × 720 px).

      For Photoroom: Open the website → Start from a photo → Download → Standard resolution (1,280 × 1 280 px).

      Taking the fish images from the first group (LB-FF-FF) as example data, the background-removed images are collected and transferred to a designated location (F:\Directory\Images\LB-FF-FF). If the images are not easily distinguishable from one another prior to landmarking, it is best to rename the images.

   2. Landmark placement

      Fourteen landmarks (Figure 1) were digitized using tpsDig2 software from each specimen's image [22]. It is recommended that the total number of landmarks should be less than half or a third of the number of individuals [23]. Prior to landmarking, a .tps file should be constructed using tpsUtil. This file serves as a link to manipulate all the image files within a specific group or classification.

      Open tpsUtil → Operation → Build tps file from images → Input directory → Input → Click any one image and then click open → Output file → Name the output file with .tps as extensions (LB-FF-FF.tps) → Actions → Setup → Check images with Actions → Select Include path → Create → Close.

      Landmarking using tpsDig2 and recording the landmarks with x- and y-coordinates in .tps file (Figure 2).

      Open tpsDig2 → File → Input source → File → .tps file (LB-FF-FF.tps) → Open → Digitalize landmarks → Save data to TPS file → (LB-FF-FF.tps) → Save → Overwrite (Supplemental File 1. LB-FF-FF.tps).

      
      

      
      Figure 1. Landmarks placement of largemouth bass used in this study. The definition of landmarks depends on the biological interpretation specific to each research scope.

      
      

      
      Figure 2. Process of landmark placing using software tpsDig2. A. Input source .tps file. B. Save data to .tps file after finishing landmarking.

      
      

   3. Set scale according to measurement methodology.

      1. Scale image one by one

         Open tpsDig2 → Image edit tools → Measure → Scale factor → Reference length (1-Centimeters) → Set scale → Draw a line with the same reference length → OK → Back to tpsDig2 window → Save data to TPS file → (LB-FF-FF.tps) → Save → Overwrite.

      2. Scale as a variable for all images

         Open tpsUtil → Operation → Add variable → Input file → Input → Click (LB-FF-FF.tps) and open → Output file → Name the output file with .tps as extensions (LB-FF-FF _allscale.tps) → Actions → Setup → Fill Variable keyword (Scale) → Fill Value → Create → Close.

   4. Compile files (.tps)

      If the images are not landmarked together with all the required data in one document, the .tps files should first be collected in one folder and then compiled into a single file by appending them.

      Open tpsUtil → Operation → Append files → Input file → Input → Click (Group1.tps) and open → Output file → Name the output file with .tps as extensions (Expreiment1-Appendgroupfile.tps) → Actions → Setup → Tick .tps files → Create → Close.

   5. The format of .tps file in GM

      In GM, .tps files are commonly used to store landmark data. These files contain both landmark coordinates and additional information about the specimen. The .tps file format was first developed by Fred L. Bookstein in the 1990s. To check the contents of a .tps file, you can open it as a .txt document.

      The general format of a TPS file is as follows:

      LM= 14

      19.00000 170.00000

      81.00000 206.00000

      616.00000 217.00000

      …

      774.00000 155.00000

      IMAGE= F:\Directory\Images\ LB-FF-FF \IMG_2852.jpg

      ID=0

      VARIABLES=Scale=0.018155

      …

      A .tps file consists of two parts: a header (LM = 14) and a data section. The header provides information about the number of landmarks, while the data section contains the actual landmark coordinates. Each line in the data section represents a single landmark with its x-coordinate and y-coordinate. Additionally, other information such as image ID, image directory, specimen ID, and scale can be included at the beginning or end of each specimen's data.

   6. Convert .tps format file to .nts format for use in MorphoJ

      Open tpsUtil → Operation → Convert tps/nts coordinates file → Input file → Input → Click (LB-FF-FF_allscale.tps) and open → Output file → Name the output file with .nts as extensions (LB-FF-FF_allscale.nts) → Actions → Create → (Tick “use scale factor” [this will convert the pixel coordinates into standard units of measurements], 2D landmarks and Image name) → Create → Close.

   7. The format of .nts file in GM

      The .nts (Numerical Taxonomic System) file format is a legacy format used for storing landmark data, developed by Fred Bookstein in the early 1980s. .nts files can be created and edited using various software programs, including MorphoJ, tpsDig, and tpsRelw.

      The general format of a .nts file is as follows:

      1 44L 28L 0

      IMG_2852.jpg IMG_2854.jpg…IMG_2976.jpg

      X1 Y1 X2 Y2…X14 Y14

      176.00000 384.00000 223.00000 419.00000 … 854.00000 376.00000

      …

      The parameters in the .nts file must adhere to specific guidelines. The first code in the parameter line should be "1". The second parameter, 44L, denotes the number of specimens, while the third parameter, 28L, represents the total count of x and y coordinates for each specimen. The fourth code in the parameter line is "0,” indicating the absence of missing data; if missing data exists, this code should be "1." It's important to note that some editing or removing redundant blank spaces may be necessary if a file does not solely consist of a single data matrix.

      
      

2. Shape variation analysis and visualization with MorphoJ

   1. Create project and dataset [24]

      File → New Project → Name for the new project (Project-images1) → Dimensionality of the data (2 dimensions) → Object symmetry? (no) → Name for the new dataset (Dataset- LB-FF-FF) → File type (NTSYSpc) → Add documents (LB-FF-FF_allscale.nts) → Create Dataset (Figure 3A) → File → Save Project → Name the Project (Project-LMB1.morphoj)

      
      

      
      Figure 3. Creating a project and extracting a classifier in MorphoJ software. A. Creating a project and dataset. B. Extracting the classifier from ID strings by counting positively from left to right.

      
      

   2. Extract and add classifiers

      Preliminaries → Extract new classifier from ID strings → Name for a new classifier (fish_source) → First character (21) → Last character (28) → Execute (Figure 3B).

      Repeat the above steps to add more classifiers and the added classifiers can be edited from: Preliminaries → Edit classifiers (or directly click Graphics).

   3. Outliers check

      Preliminaries → Find outliers.

      Repeat the above steps to create other datasets

   4. Procrustes fit

      MorphoJ uses a full Procrustes fit. For most circumstances, there is very little difference from generalized Procrustes analysis (GPA).

      Preliminaries → New Procrustes Fit → Select Align by principal axes → Perform Procrustes Fit.

   5. Export dataset for use in other software

      Project Tree → Click Dataset (Dataset-images1.NTS) (Dataset-LB-FF-FF) → Click File in menu → Export Dataset → Select Data types and Classifiers (click with Ctrl or Shift for more than one type) → Name the Dataset as .txt file (Dataset-images1.NTS.txt) → Save.

   6. Principal component analysis (PCA)

      Purpose: Reduces the dimensionality of the shape data and identifies the main axes of variation.

      Output: Principal components (PCs) that describe the major patterns of shape variation, and scatter plots of specimens in PC space.

      Preliminaries → Generate Covariance Matrix → Selected dataset (Dataset-images1.NTS) → Execute → Variation → PCA → Results (without ticking the box “Pooled within-group covariances.”).

      Eigenvalues → Variance explained by each PC.

      PC scores → Visualization of individuals in the shape space.

   7. Combine datasets

      Click on the dataset in the Project Tree → Preliminaries → Combine Datasets → Name for the new dataset → Select the Start dataset and other datasets → Execute.

   8. Discriminant function analysis (DFA)

      Purpose: Classifies specimens into predefined groups based on shape.

      Output: Discriminant functions, classification accuracy, scatter plots of specimens in discriminant space.

      Click on the combined dataset → Comparison menu → Discriminant Function → Name for the discriminant function analysis → Select Dataset and Datatype (combined dataset) → Classifier to be used as grouping criterion (fish_source) → Select Pairs of groups to be included (or tick “Include all pairs of groups”) → Permutation runs:1000 → Execute.

   9. Canonical variate analysis (CVA)

      Purpose: Maximizes separation between predefined groups (e.g., species, populations) to find axes that best differentiate the groups.

      Output: Canonical variates (CVs), scatter plots showing group separation (Figure 4), classification rates.

      Click the combined dataset → Comparison menu → Canonical Variate Analysis → Name the CVA → Select the Dataset and Data type → Select classifier variable to use for grouping (fish_source) → Tick Permutation tests → Number of iterations = 10,000 → Execute.

      
      

      
      Figure 4. Scatterplot of the CV scores showing the shape features that best distinguish multiple groups of specimens. The color of points represents different specimen groups.

      
      

   10. Visualization of DFA shape difference combined with outline data

       1. Extract fish outline

          Open ImageJ → File → Open → Select the background removed image → Open → Image → 8-bt → Image → Adjust → Brightness/Contrast (Figure 5A) → Adjust the minimum and maximum value → Apply → Apply Lookup Table → OK → Process → Binary → Make Binary → Process → Binary → Fill holes → Process → Binary → Outline → Wand (tracing) tool → Click the outline of the fish → File → Save As → Jpeg → Save.

          
          

          
          Figure 5. Extracting an outline and creating an outline file for use in MorphoJ software. A. Adjusting brightness and contrast while extracting the outline using ImageJ software. B. Creating the outline file through the outline object tool and saving it as xy coordinates using tpsDig2 software.

          
          

       2. Create outline file.

          Open tpsDIG2 → File → Input source → File → Select the outline image → Open → Outline object → Right click to save as “Save as XY coords.” (Figure 5B) → File → Save just outlines → Save the outline as .txt files → Copy the coords. to Excel or other table processing software → Insert landmarks before the coodrs. → Insert a column to number the landmarks and coords. (such as “0” and “1,” respectively) → Copy the data back to .txt file to create a file.

       3. Import outline file.

          Click the dataset → File → Import Outline File → Name for the outline → Select the outline file → Open → Jump to Graphics of DFA → Shape difference → Right click → Change the type of graph → Warped Outline Drawing (Figure 6).

          
          

          
          Figure 6. Visualization of discriminant function analysis (DFA) shape difference. Combining with outline data, the shape differences show general changes from the first to the second group. Lollipop graph and transformation grid graph can also be used to display specific variations.

          
          

3. Outline analysis and visualization with the Momocs R package

   The analysis of semi-landmark data follows a procedure similar to that previously described. The outlined data analysis is subjected to the following steps using the Momocs R package [25].

   1. ### Starting

      
      

      setwd("F:/Directory/GM-momocsu/IMGS") # Working directory

      library(Momocs)

      
      

   2. ### Collection of outline coordinates data

      1. ## Extract outline coordinates from image files

         
         

         lf1 <- list.files('F:/Directory/GM-momocsu/IMGS/LB-FF-FF-FIJI', full.names = TRUE)

         lf1

         coo1 <- import_jpg(lf1) # Data storage directory/Coo1

         
         

      2. ## Export each outline coordinates as a .csv file

         # For coo1: Data storage directory/Coo1

         setwd("F:/Directory/GM-momocsu/IMGS/outputdata/coo1_ff")

         coo_list <- lapply(1:length(coo1), function(i) {

          df <- data.frame(

           x = coo1[[i]][, 1],

           y = coo1[[i]][, 2],

           image = basename(lf1[i])

          )

          return(df)

         })

         #

         for (i in seq_along(coo_list)) {

          filename <- paste0("outline_", i, ".csv")

          write.csv(coo_list[[i]], file = filename, row.names = FALSE)

         }

         #

         
         

      3. ## Import and combine individual data files

         
         

         library(dplyr)

         imported_files1 <- lapply(paste0("outline_", seq_len(length(lf1)), ".csv"), read.csv)

         combined_df1 <- bind_rows(imported_files1)

         
         

      4. ## Export as a combined .csv file

         
         

         write.csv(combined_df1, file = "combined_outlines1.csv", row.names = FALSE) # (Supplemental File 2. combined_outlines1.csv).

         
         

   3. ### Loading data, converting to an Out object and inspecting

      
      

      setwd("F:/Directory/GM-momocsu/IMGS/outputdata/coo1_ff")

      
      

      1. ## Read the .csv file

         
         

         coo1_ff <- read.csv("combined_outlines1.csv")

         
         

         # Convert the data frame to a list of matrices

         
         

         coo1_list <- lapply(split(coo1_ff[, c("x", "y")], coo1_ff$image), as.matrix)

         
         

      2. ## Convert the list of matrices to an Out object

         
         

         out1 <- Out(coo1_list)

         
         

      3. ## Plots outlines for inspection

         # Panels of outlines (Figure 7)

         
         

         panel(out1, c(6, 8),names=TRUE, cex.names=0.5)

         
         

         # Plot one of the Out

         
         

         coo_plot(out1[3])

         
         

         
         Figure 7. Outlines of largemouth bass from one experimental group. The panel of outlines can be used for inspecting the extraction.

         
         

      4. ## Check if the outline is closed

         
         

         coo_is_closed(out1)

         
         

   4. ### Outline data normalization

      1. ## Outline smoothing

         # Extract coordinates from the Out object

         
         

         coo_list_out1 <- lapply(out1$coo, function(coo) as.matrix(coo))

         
         

         # Apply smoothing to each set of coordinates

         
         

         smoothed_coo_list_out1 <- lapply(coo_list_out1, function(coo) coo_smooth(coo, n = 5))

         
         

         # Convert the smoothed coordinates back to an Out object

         
         

         smoothed_out1 <- Out(smoothed_coo_list_out1)

         
         

         # Check for one of the smoothed outline

         
         

         coo_plot(smoothed_out1[2], main = "iterations=5")

         
         

      2. ## Centering and scaling the outlines

         # Center each set of coordinates

         
         

         centered_matrices_out1 <- lapply(coo_list_out1, coo_center)

         
         

         # Scale each set of coordinates

         
         

         scaled_matrices_out1 <- lapply(centered_matrices_out1, coo_scale, scale = 1)

         
         

         # Reconstruct the Out object

         
         

         centered_scaled_out1 <- Out(scaled_matrices_out1)

         
         

         # Check for centering and scaling

         
         

         stack(centered_scaled_out1)

         
         

      3. ## Sampling pseudo-landmarks and reversing anticlockwise coordinates

         # Sampling pseudo-landmarks

         
         

         out1_resample <- coo_sample(centered_scaled_out1, 1000)

         
         

         # Test if all shapes are developing consistently clockwise

         
         

         coo_likely_clockwise(out1_resample)

         
         

         # If not, the coordinates can be reversed

         
         

         out1_resample_rev <- out1_resample

         
         

         # Prepare Coo data for reversing

         
         

         out1_rev_coo <- out1_resample_rev$coo

         
         

         # Iterate over each outline

         
         

         for (i in seq_along(out1_rev_coo)) {

         
         

         # Extract the coordinates for this outline

         
         

         coords <- out1_rev_coo[[i]]

         
         

         # Check if the outline is likely not clockwise

         
         

         if (!coo_likely_clockwise(coords)) {

         
         

         # Reverse the coordinates

         
         

         reversed_coords <- coo_rev(coords)

         
         

         # Update the outline within the list

         
         

         out1_rev_coo[[i]] <- reversed_coords

         }

         }

         #

         
         

         # Check if all outlines are clockwise

         
         

         out1_rev_coor <- Out(out1_rev_coo)

         coo_likely_clockwise(out1_rev_coor) # All TRUE

         
         

   5. ### Adjust data by defining landmarks, sliding coordinates to create Coe objects

      1. ## Define landmarks on the sampled-reversed outlines

         
         

         out1_rev_coor_ldk <- def_ldk(out1_rev_coor,4)

         
         

      2. ## Sliding coordinates

         
         

         out1_rev_coor_ldks <- coo_slide(out1_rev_coor_ldk,ldk = 1)

         
         

      3. ## Procrustes superimposition

         
         

         out1_rev_coor_ldks_fgp <- fgProcrustes(out1_rev_coor_ldks)

         
         

      4. ## Creation Coe objects by Elliptical Fourier transform

         
         

         out1_rev_coor_ldks_fgp_ef <- efourier(out1_rev_coor_ldks_fgp, 12, norm = FALSE) # In some cases, Elliptical Fourier Transform may address the limitations of traditional methods [26].

         
         

   6. ### Principal component analysis (PCA) and successive statistics

      1. ## PCA for the group out1

         
         

         out1_rev_coor_ldks_fgp_ef_pca <- PCA(out1_rev_coor_ldks_fgp_ef)

         summary(out1_rev_coor_ldks_fgp_ef_pca)

         boxplot(out1_rev_coor_ldks_fgp_ef_pca, fac = NULL, nax = 1:12)

         scree_plot(out1_rev_coor_ldks_fgp_ef_pca)

         
         

      2. ## For PCA of combined outlines

         # Repeating the previous steps to develop data for other groups

         # Combining group outlines

         
         

         outall_rev_coor_ldks_fgp <-combine(out1_rev_coor_ldks_fgp,

                                out2_rev_coor_ldks_fgp,

                                out3_rev_coor_ldks_fgp,

                                out4_rev_coor_ldks_fgp)

         
         

         # Creating Coe objects

         
         

         outall_rev_coor_ldks_fgp_ef <- efourier(outall_rev_coor_ldks_fgp, 12, norm = FALSE)

         
         

         # PCA and visualization

         
         

         outall_rev_coor_ldks_fgp_ef_pca <- PCA(outall_rev_coor_ldks_fgp_ef)

         summary(outall_rev_coor_ldks_fgp_ef_pca)

         boxplot(outall_rev_coor_ldks_fgp_ef_pca, fac = NULL, nax = 1:12)

         scree_plot(outall_rev_coor_ldks_fgp_ef_pca) # (Figure 8)

         
         

         
         Figure 8. Principal component analysis (PCA) shows the proportion of each component for combined outlines. The number on top of each column of the bar chart represents the accumulated proportion of the 24 components.

         
         

   7. ### Mean shape analysis and Thin plate spline (TPS) analysis

      1. ## Mean shape analysis and visualization

         
         

         out1_rev_coor_ldk_fgps_ef_mean <- (MSHAPES((out1_rev_coor_ldk_fgps_ef), fac = NULL, FUN = mean, nb.pts = 1000))

         
         

         # Visualization of mean shape

         
         

         coo_plot(out1_rev_coor_ldk_fgps_ef_mean)

         
         

         # Visualization of mean shape and group outlines (Figure 9)

         
         

         out1_rev_coor_ldks_fgp %>% coo_center %>% stack

         coo_draw(out1_rev_coor_ldks_fgp_ef_mean,border='forestgreen',lwd = 3)

         
         

         # Visualization of mean shapes differences

         
         

         coo_arrows(out1_rev_coor_ldks_fgp_ef_mean,
         out2_rev_coor_ldks_fgp_ef_mean,

                   length = coo_centsize(out1_c)/2, angle = 20, code = 2)

         coo_draw(out2_rev_coor_ldks_fgp_ef_mean, border = "blue", centroid = T,

                    first.point = F, zoom = 0.7, lwd = 2)

         
         

         
         Figure 9. Visualization of the mean shape and stacked outlines from the experimental Group 1. The green line represents the mean shape of the outlines from Group 1, while the grey lines represent the individual outlines from Group 1.

         
         

      2. ## Thin plate spline analysis (TPS) (Figure 10)

         
         

         tps_grid(out1_rev_coor_ldks_fgp_ef_mean,
         out3_rev_coor_ldks_fgp_ef_mean, over = 1.2,

              amp=1, grid.size=20,shp.border = c("red", "blue"),

              shp.lwd = c(2, 2),legend = F)

         
         

         
         Figure 10. Visualization of the mean shape difference between experimental Group 1 and Group 3 through thin plate spline (TPS) analysis. The red line represents the mean shape of outlines for Group 1, while the blue line represents the mean shape of outlines for Group 3.

Validation of protocol

This protocol or parts of it has been used and validated in the following research articles:

1. Santos et al. [27]. Geometric morphometrics as a tool to identify species in multispecific flatfish landings in the Tropical Southwestern Atlantic. Fish Res. (Figure 2, 3).

2. Caillon et al. [28]. A morphometric dive into fish diversity. Ecosphere. (Figure 1, 3).

3. Rabe et al. [29]. Geometric morphometric analysis of an ontogenetic cranial series of the Permian dicynodont Diictodon feliceps. Proc R Soc B. (Figure 2).

General notes and troubleshooting

1. Based on experience from the current research, the size of an image used for outline extraction should be no smaller than 14 KB.

2. For research requiring scale, landmarking should be performed for measurements before removing the image background.

3. If a landmark is misplaced, all subsequent landmarks in that image must be crossed out. The researcher needs to correct the misplacement by re-establishing the landmarks from scratch.

4. For images of fish without an original scale, real measurements cannot be accomplished. In the methods section for setting the scale for measurement, the scale can be set to “scale=1.”

5. Based on the current versions of tpsDig2 (Version 2.32) and tpsUtil (Version 1.82), the file path is included in the name of each image during the process of creating the .tps file and placing landmarks. Sometimes, the image name in the .tps file may be presented as its original name. Therefore, prior to landmarking, images from each group should be appropriately named to differentiate them from one another and to provide sufficient information regarding their biological characteristics. For example, the images can be named LB_Foshan_group1_female_mature_bait1_IMG_1001.jpeg. The names can be shortened as necessary by abbreviating words within the name.

6. The xy coordinates of the outline can also be created using ImageJ software, though the number of coordinates is typically much less than that generated through tpsDig2.

7. For outline analysis, reversing the anticlockwise coordinates and defining landmarks to create Coe objects are two key steps for samples with significant shape variations among images in a group.

8. When applying this protocol to GM analyses in other species, landmarking, outline extraction, and the selection of analysis methods should correlate with biological interpretation.

Acknowledgments

The author thanks Dr. Tingwei Zhang for assisting in taking photos of fish. This work was supported by the Central Public-interest Scientific Institution Basal Research Fund, CAFS (2024SJRC9), the Project of Innovation Team of Survey and Assessment of the Pearl River Fishery Resources (2023TD10) and the National Natural Science Foundation of China (31600446).

Competing interests

The author declares no competing interests.

Ethical considerations

All animal study protocols were approved by the Laboratory Animal Ethics Committee of Pearl River Fisheries Research Institute, CAFS (LAEC-PRFRI-20160323).

References

1. Polly, P. D. and Motz, G. J. (2016). PATTERNS AND PROCESSES IN MORPHOSPACE: GEOMETRIC MORPHOMETRICS OF THREE-DIMENSIONAL OBJECTS. The Paleontological Society Papers 22: 71–99. https://doi.org/10.1017/scs.2017.9
2. Dwivedi, A. K. and De, K. (2023). Role of Morphometrics in Fish Diversity Assessment: Status, Challenges and Future Prospects. Natl Acad Sci Lett. 47(2): 123–126. https://doi.org/10.1007/s40009-023-01323-x
3. Lawing, A. M. and Polly, P. D. (2009). Geometric morphometrics: recent applications to the study of evolution and development. J Zool. 280(1): 1–7. https://doi.org/10.1111/j.1469-7998.2009.00620.x
4. Collyer, M. L., Davis, M. A. and Adams, D. C. (2020). Making Heads or Tails of Combined Landmark Configurations in Geometric Morphometric Data. Evol Biol. 47(3): 193–205. https://doi.org/10.1007/s11692-020-09503-z
5. Palmqvist, P. (2022). Recensión. Fred L. Bookstein, 1991. Morphometric Tools for Landmark Data: Geometry and Biology. Cambridge University Press, Cambridge. ISBN 0-521-38385-4. Span J Palaeontol. 7(2): 166. https://doi.org/10.7203/sjp.25045
6. Bookstein, F. L. (2018). Geometric morphometrics: its geometry and its pattern analysis, in: Bookstein, F.L. (Ed.), A Course in Morphometrics for Biologists. Cambridge University Press, Cambridge, pp. 322–493. https://doi.org/10.1017/9781108120418.006
7. Wärmländer, S. K. T. S., Garvin, H., Guyomarc'h, P., Petaros, A. and Sholts, S. B. (2018). Landmark Typology in Applied Morphometrics Studies: What's the Point?. Anat Rec. 302(7): 1144–1153. https://doi.org/10.1002/ar.24005
8. Webster, M. and Sheets, H. D. (2010). A Practical Introduction to Landmark-Based Geometric Morphometrics. In Alroy, J., Hunt, G. (Eds.), Quantitative methods in Paleobiology. Cambridge University Press, Cambridge, pp. 16: 163–188. https://doi.org/10.1017/s1089332600001868
9. Thompson, D. A. (1917). On growth and form. Abridged and edited by j. T. Bonner, Cambridge University Press, 1961. Cambridge University Press, Cambridge.
10. Bookstein, F. L. (2024). Quadratic Trends: A Morphometric Tool Both Old and New. Evol Biol. 51(1): 1–44. https://doi.org/10.1007/s11692-023-09621-4
11. Bookstein, F. L. (2023). Reworking Geometric Morphometrics into a Methodology of Transformation Grids. Evol Biol. 50(3): 275–299. https://doi.org/10.1007/s11692-023-09607-2
12. Hu, H., Bjarnason, A. and Benson, R. (2021). 3D geometric morphometric protocol - quantifying the morphology of living and extinct vertebrates. Bio-101: e1010664. https://doi.org/10.21769/BioProtoc.1010664
13. Wang, C., Fang, Z. and Chen, X. J. (2022). Advances in the application of biometrics-based geometric morphometrics in fisheries. Marine Fisheries. 44 (1): 112–128. https://doi.org/10.13233/j.cnki.mar.fish.2022.01.004
14. Tong, Y., Zhang, M., Jenkins Shaw, J., Wan, X., Yang, X., Bai, M. (2021). A geometric morphometric dataset of stag beetles. Biodivers Sci. 29(9): 1159–1164. https://doi.org/10.17520/biods.2021160
15. Moccetti, P., Rodger, J. R., Bolland, J. D., Kaiser-Wilks, P., Smith, R., Nunn, A. D., Adams, C. E., Bright, J. A., Honkanen, H. M., Lothian, A. J., et al. (2023). Is shape in the eye of the beholder? Assessing landmarking error in geometric morphometric analyses on live fish. PeerJ. 11: e15545. https://doi.org/10.7717/peerj.15545
16. Schneider, C. A., Rasband, W. S. and Eliceiri, K. W. (2012). NIH Image to ImageJ: 25 years of image analysis. Nat Methods. 9(7): 671–675. https://doi.org/10.1038/nmeth.2089
17. Rohlf, F. J. (2021). https://sbmorphometrics.org/index.html [Accessed January 31, 2023]
18. Klingenberg, C. P. (2010). MorphoJ: an integrated software package for geometric morphometrics. Mol Ecol Resour. 11(2): 353–357. https://doi.org/10.1111/j.1755-0998.2010.02924.x
19. R Core Team (2023). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org. [Accessed October 31, 2023]
20. Posit team. (2023). RStudio: Integrated Development Environment for R. Posit Software, PBC, Boston, MA. URL http://www.posit.co/. [Accessed October 31, 2023]
21. Karki, N. P., Colombo, R. E., Gaines, K. F. and Maia, A. (2020). Exposure to 17β estradiol causes erosion of sexual dimorphism in Bluegill (Lepomis macrochirus). Environ Sci Pollut Res. 28(6): 6450–6458. https://doi.org/10.1007/s11356-020-10935-5
22. Elmer, K. R., Kusche, H., Lehtonen, T. K. and Meyer, A. (2010). Local variation and parallel evolution: morphological and genetic diversity across a species complex of neotropical crater lake cichlid fishes. Philos Trans R Soc Lond B Biol Sci. 365(1547): 1763–1782. https://doi.org/10.1098/rstb.2009.0271
23. Savriama, Y. (2018). A Step-by-Step Guide for Geometric Morphometrics of Floral Symmetry. Front Plant Sci. 9: e01433. https://doi.org/10.3389/fpls.2018.01433
24. Crampton, D. A., Giacomini, G. and Meloro, C. (2024). Mandibular morphology in four species of insectivorous bats: the impact of sexual dimorphism and geographical differentiation. J Zool. 323(4): 331–345. https://doi.org/10.1111/jzo.13177
25. Bonhomme, V., Picq, S., Gaucherel, C. and Claude, J. (2014). Momocs: Outline Analysis UsingR. J Stat Softw. 56(13): ei13. https://doi.org/10.18637/jss.v056.i13
26. García-Bustos, M., García Bustos, P. and Rivero, O. (2024). New Methods for Old Questions: The Use of Elliptic Fourier Analysis for the Formal Study of Palaeolithic Art. J Archaeol Method Theory.  https://doi.org/10.1007/s10816-024-09656-7
27. Santos, S. R., Pessôa, L. M. and Vianna, M. (2019). Geometric morphometrics as a tool to identify species in multispecific flatfish landings in the Tropical Southwestern Atlantic. Fish Res. 213: 190–195. https://doi.org/10.1016/j.fishres.2019.01.017
28. Caillon, F., Bonhomme, V., Möllmann, C. and Frelat, R. (2018). A morphometric dive into fish diversity. Ecosphere 9(5): e2220. https://doi.org/10.1002/ecs2.2220
29. Rabe, C., Marugán-Lobón, J., Smith, R. M. H. and Chinsamy, A. (2024). Geometric morphometric analysis of an ontogenetic cranial series of the Permian dicynodont Diictodon feliceps. Proc R Soc B. 291(2027): e0626. https://doi.org/10.1098/rspb.2024.0626

Supplementary information

The following supporting information can be downloaded here:

1. Supplemental File 1. LB-FF-FF.tps
2. Supplemental File 2. combined_outlines1.csv