Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1919
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-19T18:13:34.669649Z
status: published
lens: g_space
sector: health
headline: Sheep Pulmonary Image Recognition (SPIRO): an explainable deep learning framework for macroscopic assessment of ovine pulmonary lesions
dek: Accurate detection and classification of pulmonary diseases in sheep are critical for ensuring animal welfare, food safety, and effective disease surveillance in abattoir settings. In this study, we propose a two-stage, explainable, histopathology-informed deep learning framework for the automated classification of ovine lung lesions. The dataset comprises 1,052 sheep lung images, annotated through a dual-level diagnostic process involving macroscopic inspection and histopathological confirmation, and classified…
gain_title: A YOLOv11 plus EfficientNetB0 pipeline classified ovine lungs as healthy, inflammatory, or neoplastic with 90% test accuracy, supporting objective post-mortem assessment for abattoir surveillance.
problem_title: (none)
trace_subject: (none)
gain_reading: A YOLOv11 plus EfficientNetB0 pipeline classified ovine lungs as healthy, inflammatory, or neoplastic with 90% test accuracy, supporting objective post-mortem assessment for abattoir surveillance.
gain_evidence: The YOLOv11-EfficientNetB0 pipeline achieved an overall test accuracy of 90% and a weighted F1-score of 0.88. | objective, scalable, and biologically grounded post-mortem assessment of sheep lungs
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers built an image-based system that first locates the lung in a photograph and then sorts it into healthy, inflammatory, or neoplastic categories. It was trained and tested on 1,052 sheep lung images labeled through macroscopic inspection and histopathological confirmation.

The work matters because abattoir lung assessment supports animal welfare, food safety, and disease surveillance, and an automated, interpretable tool could make it more objective. The results are a feasibility demonstration, and performance on healthy and inflammatory lungs was clearly weaker than on tumors.
limitation: The source describes feasibility rather than deployment, and performance was noticeably lower for healthy and inflammatory lungs than for neoplastic ones.
tag: Evidence-backed gain
key_points: The dataset contained 1,052 sheep lung images, each labeled as healthy, inflammatory, or neoplastic. | VGG16, VGG19, ResNet50, and EfficientNetB0 were benchmarked, and EfficientNetB0 had the highest standalone performance. | Neoplastic lesions reached precision 0.98, recall 0.99, and F1 0.99. | Healthy and inflammatory lungs scored F1 values of 0.83 and 0.82, respectively. | Grad-CAM maps indicated predictions were mainly driven by diagnostically relevant lung regions.
rundown: The framework has two stages: YOLOv11 localizes lung regions, and EfficientNetB0 classifies the cropped images. Four convolutional network architectures were compared under identical conditions before the best one was chosen.

The pipeline reached an overall test accuracy of 90% and a weighted F1-score of 0.88. Tumor detection was strongest, while distinguishing healthy from inflammatory lungs was less accurate.

Interpretability was assessed with Grad-CAM visualizations, which suggested the model focused on relevant pulmonary areas. The authors propose potential uses in abattoir surveillance, veterinary pathology, and flock health monitoring.
sources:
- peer_reviewed | The Veterinary Journal | https://doi.org/10.1016/j.tvjl.2026.106936 | 2026-10-14
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

e960c17de4c72c4fda365fd6393aeea9ec900dee37f0b32169eb72341ea8c28d

previous

0000000000000000000000000000000000000000000000000000000000000000