JDS_Supplementary_Tables_Touil_et_al._Reticuloruminal_pH_and_Subacute_Ruminal_Acidosis_prediction.pdf
Bibliographic record
Abstract
The prolonged occurrence of low reticuloruminal pH (rpH) can lead to subacute ruminal acidosis (SARA), a condition detrimental to cow health and costly for the dairy industry. This research project aims to predict rpH and SARA by utilizing artificial intelligence (AI) applied to Fourier transform infrared spectroscopy (FTIR) spectra derived from routine Dairy Herd Improvement (DHI) milk analyses of individual cows.In this research, 107 multiparous Holstein cows from 12 commercial farms in Québec, Canada, were selected. The rpH of these cows was continuously monitored for 150 days using wireless boluses. Concurrently, 1,744 individual milk samples were collected, with 872 samples taken in the morning (AM) and 872 in the afternoon (PM). These samples were analyzed to obtain FTIR spectra.The FTIR spectra and rpH data were merged to form three equally balanced datasets for AI model development: one containing AM samples, one with PM samples, and a combined dataset of both AM and PM samples, each comprising 872 samples. Various spectral pre-processing methods were evaluated, including the first derivative and filtering with three different sets of spectra.Several AI algorithms were employed to predict rpH and SARA, including partial least squares (PLS), random forest (RF), and gradient boosting (GB). In total, 36 different models were developed and evaluated for their prediction performance using three cross-validation (CV) methods: classic 15-fold, leave-one-farm-out (LOFO), and leave-cows-out (LCO) CV.For rpH prediction, classic CV yielded the highest performance, with median R² values of 0.26 for the AM dataset, 0.24 for the PM dataset, and 0.26 for the combined AM/PM dataset. However, these results are likely overly optimistic, as models evaluated with LOFO or LCO CV did not exceed R² values of 0.08.In contrast, SARA prediction showed more promising results. The best models evaluated using classic CV achieved accuracies of 72% for the AM dataset, 72% for the PM dataset, and 71% for the combined AM/PM dataset. Performance achieved using LOFO and LCO CV methods were comparable to the classic CV, with accuracies of 73%, 79%, and 72% for LOFO, and 71%, 70%, and 75% for LCO, for the AM, PM, and combined datasets, respectively.These findings suggest that while rpH prediction from FTIR spectra is not feasible, SARA can be predicted with over 70% accuracy using routine DHI milk samples of individual cows. This research highlights the potential of AI and FTIR spectroscopy in enhancing the management of cow health and improving economic outcomes in the dairy industry.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.035 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.003 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.003 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.741 | 0.211 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; the direct Gemma label and the distilled Codex classifier agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".