Delta thermal radiomics: An application in dairy cow teats

Parminder S. Basran, C. DiLeo, Youshan Zhang, Ian Porter, Matthias WielandJDS 2022 · DOI

TL;DR

This paper introduces "delta thermal radiomics," a novel method to analyze infrared thermal images of dairy cow teats. By applying radiomics techniques to pre- and post-milking thermal images, the approach can quantitatively detect subtle changes in teat surface temperature, offering a non-invasive way to monitor udder health and physiological responses to milking.

What Problem Does This Paper Address?

Monitoring the health and physiological status of dairy cows is crucial for animal welfare and milk production. Teat health, in particular, is a key indicator, and changes in blood flow and temperature can signal issues like inflammation or stress. Traditional methods for assessing these changes can be invasive or subjective. While infrared thermography can visualize surface temperature, extracting meaningful, quantitative data from these images, especially subtle changes over time, has been challenging. This paper addresses the need for a non-invasive, quantitative method to analyze thermal images for detecting short-term physiological changes in dairy cow teats, specifically those occurring around milking.

Methods

The core of this work is the application of radiomics to thermal imaging, termed "delta thermal radiomics." The workflow involves:

  1. Image Acquisition: Infrared thermal images of dairy cow hind teats are captured using a digital thermal camera.
  2. Image Conversion: The multi-channel colorized thermal images (JPG) are converted into single-channel medical images in the DICOM format. In this conversion, pixel intensities are mapped to represent temperature values.
  3. Segmentation: The left and right hind teats are manually segmented by two independent investigators using open-source medical image analysis software (3DSlicer). This step defines the "region of interest" for analysis.
  4. Radiomics Calculation: An open-source radiomics package (Pyradiomics) is used to extract a large number of quantitative image biomarkers (features) from the segmented teat regions. These features include first-order statistics (e.g., mean, skewness, entropy) and higher-order texture features (e.g., GLCM, GLDM, NGTDM).
  5. Delta Analysis: To assess changes, radiomic features are computed for images taken before and after milking. "Delta radiomics" refers to the analysis of the differences or changes in these features between these two time points.
  6. Statistical Analysis: Student's t-test is used to determine statistical significance (P < 0.01) for differences in radiomic features between pre- and post-milking states. Cohen's d is employed to quantify the effect size (d > 1.2 indicates a strong effect).

The novelty lies in applying radiomics, a technique more commonly used in medical imaging for disease detection, to thermal images for physiological monitoring in an agricultural context, specifically focusing on the change (delta) in features over a short interval.

Datasets

The study utilized thermal images from:

  • 25 dairy cows for initial segmentation evaluation.
  • 18 dairy cows (36 teats total) for pre- and post-milking delta radiomics analysis.

The paper does not introduce a new benchmark dataset but rather uses data collected specifically for this study.

Key Insights and Ideas

  • Radiomics Unlocks Hidden Thermal Information: The study demonstrates that radiomics can extract quantitative features from thermal images that go beyond simple average temperature, revealing subtle patterns related to blood flow and tissue characteristics.
  • Milking Induces Measurable Thermal Changes: The "delta thermal radiomics" approach successfully identified significant changes in teat surface temperature patterns before and after milking, suggesting it can capture physiological responses to this intervention.
  • Consistency is Key: High Dice Similarity Scores (DS) between manual segmentations by different investigators (average 0.952) indicate that manual segmentation of teats is reproducible, which is crucial for reliable radiomic analysis.
  • Robustness to Segmentation Variability: The radiomic features computed were largely insensitive to minor variations in manual segmentation, suggesting the method is robust.
  • Beyond Average Temperature: While average teat temperature increased post-milking, other features like skewness and entropy showed more pronounced and potentially more informative changes, highlighting the value of a multi-feature approach.

Key Results

  • Segmentation Reliability: The average Dice Similarity Score (DS) between two investigators' manual segmentations was 0.952 (95% CI: 0.913–0.982), indicating high consistency.
  • Inter-Investigator Feature Consistency: No significant differences (P > 0.07, Cohen's d < 0.36) were found in radiomic features computed from segments by different investigators.
  • Pre- vs. Post-Milking Changes: Out of 109 analyzed image biomarkers, 17 were simultaneously statistically significant (P < 0.01) and exhibited a strong effect size (Cohen's d > 1.2) when comparing pre- and post-milking states.
  • Mean Temperature Increase: Feature maps indicated an increase in mean teat temperature after milking.
  • Morphological Changes: Significant increases were observed in major axis length (31%) and surface area (50%) of the segmented teats post-milking, likely due to longitudinal shear forces during milking.

Limitations and Future Work

  • Manual Segmentation Bottleneck: The most time-consuming part of the current workflow is the manual segmentation of teats.
  • Spatial Calibration: Thermal images are not spatially calibrated in an absolute sense, limiting the direct interpretation of morphometric features like length and area.
  • Temperature Resolution: The current temperature resolution of 0.1°C might limit the detection of very subtle temperature features.
  • Noise and Interpolation: Impulse noise and potential interpolation errors during image conversion can affect analysis.

Future Work:

  • Automated Segmentation: Developing automated segmentation techniques (e.g., using U-Nets) to reduce the time and effort required.
  • Standardization: Further standardization of imaging protocols and post-processing techniques.
  • Longitudinal Studies: Applying delta thermal radiomics to monitor longer-term changes in teat health over weeks or months.
  • Association with Physiological Variables: Investigating the direct association between thermal radiomic features and specific physiological parameters like blood flow or inflammatory markers.
  • Clinical Validation: Expanding studies to include teats with various health conditions to validate the diagnostic potential.

Summary updated

Abstract

We describe a novel approach for analyzing thermal images by way of radiomics (i.e., thermal radiomics) and how it can be used to monitor short-term temperature changes of dairy cow hind teats; that is, delta thermal radiomics. The heat generated from metabolic activities and blood-flow patterns can be visualized using thermal radiography of the skin surface. The hind teats from 25 dairy cows were imaged with a digital thermal camera and the images were converted to medical images (DICOM format) by mapping the multi-channel colorized thermal image to a monochromatic image whose intensities represent temperature. The 50 teats (left and right hind) were then manually segmented by 2 investigators. Radiomics analysis, which is a common method of extracting semantic and nonsemantic image biomarkers from medical images for machine learning, was performed. To evaluate whether this approach can detect pre- and postmilking differences, 18 cows were imaged before and after milking, the teats were manually segmented, and radiomic calculations were performed. Student's <i>t</i>-test was used to provide an estimate of the likelihood of whether postmilking thermal image biomarkers are the same as premilking thermal image biomarkers, and Cohen's <i>d</i> was used to evaluate the size of the effect (<i>d</i> > 1.2). To evaluate uncertainties from manual segmentation, the Dice similarity score (DS) between the 2 investigators' segments was computed. The average DS (95% confidence limit) was 0.952 (0.913-0.982) when comparing the 2 investigators' segmentations. There was no significant difference in DS when comparing the left and right segmented teats, suggesting that teats can be segmented consistently. No differences (<i>d</i> < 0.36) were observed when comparing image biomarkers from one investigator's segments with the other's, suggesting that image biomarkers computed from one investigator's segmentation of teats are not likely to differ from those computed from the other investigator. When comparing image biomarkers before and after milking, 109 image biomarkers were analyzed, and 17 image biomarkers were simultaneously significant and exhibited effect size. Thus, delta thermal radiomics offers a noninvasive and quantitative method of monitoring skin temperature changes in humans and animals after an intervention. The advantage of this approach is that it can reveal both perceptible and imperceptible surface temperature features that may be useful for detecting and managing dairy teat health.
BibTeX
@inproceedings{basran2022delta,
  author = {Parminder S. Basran and C. DiLeo and Youshan Zhang and Ian Porter and Matthias Wieland},
  title = {{Delta thermal radiomics: An application in dairy cow teats}},
  booktitle = {JDS},
  year = {2022},
  doi = {10.3168/jdsc.2021-0179},
}
Figure