Codes Overview:

This project is broken into two sections: Python and R.

The Python code section has four types of codes:

1. Code to count the specific color of pixels in an image that matches the color key (see Fig. 3).
2. Code to convert and create new variables (i.e., percent area variables) for one map value and batch-process multiple data frames.
3. Code to batch-process images and convert and create new variables at once.
4. Predictive Analysis

The R code is broken up into four sections:

1. Descriptive Statistics
2. Exploratory Data Analysis
3. Statistical Testing Analysis - ANOVA and Linear Regression
4. Multivariate Analysis - Principal Component Analysis (PCA)

Project Background

In the field of human evolution, it is common to hear and read that hominins, our early human ancestors, have an 'ape-like' growth pattern. However, visual observations of growth in different ape species (i.e., gorillas, chimpanzees, orangutans, gibbons, etc) show that each ape species have their own unique growth pattern.

We also know that modern humans have a unique growth pattern--rapid growth as infants, slow growth during childhood followed by a growth spurt during puberty--compared to apes and hominins.

However, we cannot visually observe the growth patterns in hominins as they are all extinct, and all that remains are parts of their fossilized skeletal systems.

Luckily for people studying human evolution, the bones of the skeletal system are like tree rings, keeping a record of the individual's growth. We can map the patterns of growth based on how fast or slow bone tissue is deposited in a bone cross-section to understand how an individual grew in life after death (Fig. 1).

Unfortunately, there are few comparative datasets on bone microanatomy from our closest living relatives--apes and monkeys--and few comparative analyses of bone microanatomy growth patterns between ape species (McFarlin et al., 2016; Tuosto et al., 2019).

Understanding the growth pattern differences in our closest living primate relatives can help us determine the potential growth patterns our early human ancestors followed.

##Add image of bone cross-section

Project Objective

The objective of this project is to explore the differences in growth patterns in primate microanatomy and build a predictive model to provide the probability of a new tissue map growing more like which primate.

Code Sections

Python Section 1: Count Pixels

This code will count the number of pixels of the specific colors in the bone maps and provide the pixel results as a table. This code section is good if you are looking for anomalies or have large image files and low computing power.

Python Section 2: Create New Variables

This code will calculate the proportions of the bone tissue types from the raw pixels and create 4 new variables from the existing variables, C.Ar, %CA, %PCA, and %ECA (see Data Description for more information). Converting the raw pixel values into proportions allows us to control for body size differences between the primates and image size differences. This code file has code to process 1 image value at a time or to batch-process multiple image values.

Python Section 3: Batch-Processing Images and Values

This code section is a combination the Python section 1 & 2, just automated for multiple images.

R Section 1: Descriptive Statistics

R Section 2: Exploratory Data Analysis

R Section 3: Statistical Testing Analysis - ANOVA and Linear Regression

R Section 4: Multivariate Analysis - Principal Component Analysis (PCA)

Principal component analysis (PCA) was done to visualize and summarize the variance within species differences by age and sex and to detect broad species growth pattern differences.

R Section 5: Predictive Analysis

Data set:

simulation_primate_bone.csv

Data Description:

Python Image The image provided for the Python code is a random bone histological image that was mapped by K. Tuosto and is not part of the images used for the project publication (Tuosto et al., in prep).

R Data The data provided for practice is simulated data of the data currently being prepared for publication (Tuosto et al., in prep). There are # observations and 21 variables.

Note: Genus has been modified for the practice set.

1. ID = Randomized
2. GENUS = Ape 1, Ape 2, Ape 3, Monkey 1
3. SEX = M (male), F (female), U (unknown)
4. AGE = Ages based on dental eruption and skeletal fusion, 6 age classes from youngest to oldest: PDICED, DECID, M1, M2, M3, EPIPH
5. WOV = Fastest forming bone regardless of bone cortex, mostly observed on the outer cortex
6. FLC = Fast-forming bone regardless of bone cortex, mostly observed on the outer cortex
7. FLC-LZPO = Fast-forming bone transitioning to slower-forming bone on the outer cortex
8. PF = Moderate speed forming bone on the outer cortex
9. PF-LAM = Slow-forming bone on the outer cortex
10. LAM = Slowest forming bone on the outer bone cortex
11. ECCC = Transition bone type
12. EPF = Moderate speed forming bone on the inner cortex
13. EPF-LAM = Slow-forming bone on the inner cortex
14. ELAM = Slowest forming bone on the inner bone cortex
15. SP = Muscle attachment fibers
16. INTER = Bone tissue in-between osteonal bone tissue
17. HAV = Osteonal bone tissue
18. C.Ar = Cortical bone area
19. %CA = Percent cortical bone area
20. %PCA = Percent periosteal cortical bone area (i.e., outer bone cortex)
21. %ECA = Percent endosteal cortical bone area (i.e., inner bone cortex)

For a description of each bone tissue type, see the supplemental materials from McFarlin et al., 2016.

Tissue Mapping

Please refer to McFarlin et al., 2006/McFarlin et al., 2008, and McFarlin et al., 2016 publications for information on how to tissue map cortical bone (Fig. 2).

Or if you would like guidance on the tissue mapping method, please feel free to contact me on GitHub.

Fig. 2 Example of cortical bone tissue type mapping. For cortical bone tissue mapping you will need a brightfield (LM) and circularly-polarized light (CPL) or polarized light (PL) images layered on top of each other with the CPL/PL at a reduced opacity. Mapping of the various tissue types should be done in their own layer. Note, interstitial bone tissue (INT) is missing from the color table (Tuosto et al., 2019 poster presentation at the International Symposium on Palaeohistology, South Africa).