← Articles

Low et al: Machine-learning-based prediction of functional recovery in deep-pain-negative dogs after decompressive thoracolumbar hemilaminectomy for acute intervertebral disc extrusion
Veterinary Surgery 4, 2025

🔍 Key Findings

The study included 162 deep-pain-negative dogs undergoing decompressive surgery (hemilaminectomy) for acute thoracolumbar intervertebral disc extrusion (IVDE).

Ambulatory recovery occurred in 53.1% of dogs (86/162).

The best performing machine-learning model was XGBoost, with an AUC of 0.9502 and accuracy of 89.1%, outperforming Ridge, AdaBoost, and Naive Bayes models.

Preoperative-only XGBoost models were less accurate, with AUC dropping to 0.8271 and accuracy to 71.9%.

Top predictive features (by SHAP analysis) included:

  1. T2-weighted to L2 spinal cord signal ratio (lower values predicted better outcome)
  2. Use of fenestration (presence associated with better recovery)
  3. Hospitalization duration
  4. Imaging modality used
  5. Duration of nonambulatory status

Machine learning provided better insight into prognostic factors than traditional statistical methods.

Simini Surgery Review Podcast

How critical is this paper for crushing the Boards?

🚨 Must-know. I’d bet on seeing this.

📚 Useful background, not must-know.

💤 Skip it. Doubt it’ll ever show up.

Thanks for the feedback!
‍We'll keep fine-tuning the articles vault.
Oops — didn’t go through.
‍Mind trying that again?
← Articles

Low et al: Machine-learning-based prediction of functional recovery in deep-pain-negative dogs after decompressive thoracolumbar hemilaminectomy for acute intervertebral disc extrusion
Veterinary Surgery 4, 2025

🔍 Key Findings

The study included 162 deep-pain-negative dogs undergoing decompressive surgery (hemilaminectomy) for acute thoracolumbar intervertebral disc extrusion (IVDE).

Ambulatory recovery occurred in 53.1% of dogs (86/162).

The best performing machine-learning model was XGBoost, with an AUC of 0.9502 and accuracy of 89.1%, outperforming Ridge, AdaBoost, and Naive Bayes models.

Preoperative-only XGBoost models were less accurate, with AUC dropping to 0.8271 and accuracy to 71.9%.

Top predictive features (by SHAP analysis) included:

  1. T2-weighted to L2 spinal cord signal ratio (lower values predicted better outcome)
  2. Use of fenestration (presence associated with better recovery)
  3. Hospitalization duration
  4. Imaging modality used
  5. Duration of nonambulatory status

Machine learning provided better insight into prognostic factors than traditional statistical methods.

Simini Surgery Review Podcast

Know What Matters in the Literature - and Why

We distill peer-reviewed surgical studies into clinically relevant summaries and
exam-style questions, so you can make informed decisions with confidence.

Free Access. No Spam. Just Smarter Surgical Learning

Multiple Choice Questions on this study

In Low 2025 et al., on machine-learning outcomes in IVDE, what was the most important predictor of outcome according to SHAP value analysis?

A. Age
B. Body weight
C. T2W:L2 spinal cord signal ratio
D. Fenestration performed
E. Time to surgery

Answer: T2W:L2 spinal cord signal ratio

Explanation: Lower T2W:L2 signal ratio was the strongest predictor of ambulation recovery.
In Low 2025 et al., on machine-learning outcomes in IVDE, which factor was associated with improved functional recovery?

A. Higher T2W:L2 ratio
B. Absence of fenestration
C. Longer hospitalization
D. Fenestration performed
E. Delayed decompression

Answer: Fenestration performed

Explanation: Fenestration was among the top predictive features associated with better recovery.
In Low 2025 et al., on machine-learning outcomes in IVDE, which model achieved the highest predictive performance for ambulation recovery?

A. AdaBoost
B. Naive Bayes
C. Ridge Regression
D. XGBoost
E. Random Forest

Answer: XGBoost

Explanation: XGBoost had the highest AUC (0.9502) and accuracy (89.1%) in this cohort.
In Low 2025 et al., on machine-learning outcomes in IVDE, what was the performance of the XGBoost model when using only preoperative variables?

A. AUC 0.6504, Accuracy 59.4%
B. AUC 0.7102, Accuracy 65.2%
C. AUC 0.8271, Accuracy 71.9%
D. AUC 0.8943, Accuracy 76.1%
E. AUC 0.9108, Accuracy 80.3%

Answer: AUC 0.8271, Accuracy 71.9%

Explanation: The preoperative-only XGBoost model had AUC 0.8271 and accuracy 71.9%.
In Low 2025 et al., on machine-learning outcomes in IVDE, what proportion of deep-pain-negative dogs recovered ambulation after decompressive surgery?

A. 36.4%
B. 45.7%
C. 53.1%
D. 62.2%
E. 69.8%

Answer: 53.1%

Explanation: Ambulation recovery occurred in 86 out of 162 dogs, a 53.1% success rate.

Access the full library of surgical summaries and exam-style questions.

Get Instant Access

Educational content developed independently and supported by Simini.

The maker of Simini Protect Lavage.

Review More Key Summaries from this Veterinary Surgery Journal Issue

Jenkins

 et al: 

Medial epicondylar fissure fracture as a complication of transcondylar screw placement for the treatment of humeral intracondylar fissure

Neal

 et al: 

The effect of an aiming device on the accuracy of humeral transcondylar screw placement

Low

 et al: 

Machine-learning-based prediction of functional recovery in deep-pain-negative dogs after decompressive thoracolumbar hemilaminectomy for acute intervertebral disc extrusion

Welsh

 et al: 

Biomechanical analysis of orthogonal and unilateral locking plate constructs in a fracture gap model

Brockman

 et al: 

Improvement in short‐term outcome over time, in a single center embarking on a canine mitral valve repair program using a structured multidisciplinary approach