A reliable and low-cost deep learning model integrating convolutional neural network and transformer structure for fine-grained classification of chicken Eimeria species

CarcassFormer: an end-to-end transformer-based framework for simultaneous localization, segmentation and classification of poultry carcass defect

In the food industry, assessing the quality of poultry carcasses during processing is a crucial step. This study proposes an effective approach for automating the assessment of carcass quality without requiring skilled labor or inspector involvement. The proposed system is based on machine learning (ML) and computer vision (CV) techniques, enabling automated defect detection and carcass quality assessment. To this end, an end-to-end framework called CarcassFormer is introduced. It is built upon a Transformer-based architecture designed to effectively extract visual representations while simultaneously detecting, segmenting, and classifying poultry carcass defects. Our proposed framework is capable of analyzing imperfections resulting from production and transport welfare issues, as well as processing plant stunner, scalder, picker, and other equipment malfunctions. To benchmark the framework, a dataset of 7,321 images was initially acquired, which contained both single and multiple carcasses per image. In this study, the performance of the CarcassFormer system is compared with other state-of-the-art (SOTA) approaches for both classification, detection, and segmentation tasks. Through extensive quantitative experiments, our framework consistently outperforms existing methods, demonstrating re- markable improvements across various evaluation metrics such as AP, AP@50, and AP@75. Furthermore, the qualitative results highlight the strengths of CarcassFormer in capturing fine details, including feathers, and accurately localizing and segmenting carcasses with high precision. To facilitate further research and collaboration, the source code and trained models will be made publicly available upon acceptance.

Automated Image Analysis for Detection of Coccidia in Poultry

Coccidiosis, caused by the protozoan Eimeria sp., is one of the most common and costly diseases impacting the poultry industry. To establish effective control measures, it is essential to identify these protozoa. Typical methods for identifying and determining the severity of the protozoal infection include intestinal lesion scoring or enumeration of the protozoal oocysts in fecal samples. Standard analysis methods require highly skilled technicians or veterinarians to manually identify and manually enumerate these protozoal parasites. This process is labor intensive, time-consuming, and susceptible to human error. None of the current methods available, including molecular flow cytometry or even digital image analysis, can determine if an oocyst is sporulated or not. Oocysts are not infectious until they sporulate. The goal of this study was to design an automated model using Artificial Intelligence (AI) to expedite the process of enumeration, improve the efficiency and accuracy of the species identification, and determine the ability of the oocysts to infect. To this end, we trained and evaluated computer vision models based on the Mask RCNN neural network architecture. A model was trained to detect and differentiate three species and to determine sporulation for each (totaling six detection groups). This model achieved a mean relative percentage difference (RPD) of 5.64%, representing a slight overcount compared to manual counts, averaging across all groups. The mean RPD for each group individually fell within a range from -33.37% to 52.72%. These results demonstrate that these models were speedy and had high agreement with manual counts, with minimal processing of field-quality samples. These models also could differentiate the sporulation status of the oocysts, providing critical diagnostic information for potential field applications.

Amplicon sequencing allows differential quantification of closely related parasite species: an example from rodent Coccidia (Eimeria)

BACKGROUND

Quantifying infection intensity is a common goal in parasitological studies. We have previously shown that the amount of parasite DNA in faecal samples can be a biologically meaningful measure of infection intensity, even if it does not agree well with complementary counts of transmission stages (oocysts in the case of Coccidia). Parasite DNA can be quantified at relatively high throughput using quantitative polymerase chain reaction (qPCR), but amplification needs a high specificity and does not simultaneously distinguish between parasite species. Counting of amplified sequence variants (ASVs) from high-throughput marker gene sequencing using a relatively universal primer pair has the potential to distinguish between closely related co-infecting taxa and to uncover the community diversity, thus being both more specific and more open-ended.

METHODS

We here compare qPCR to the sequencing-based amplification using standard PCR and a microfluidics-based PCR to quantify the unicellular parasite Eimeria in experimentally infected mice. We use multiple amplicons to differentially quantify Eimeria spp. in a natural house mouse population.

RESULTS

We show that sequencing-based quantification has high accuracy. Using a combination of phylogenetic analysis and the co-occurrence network, we distinguish three Eimeria species in naturally infected mice based on multiple marker regions and genes. We investigate geographical and host-related effects on Eimeria spp. community composition and find, as expected, prevalence to be largely explained by sampling locality (farm). Controlling for this effect, the novel approach allowed us to find body condition of mice to be negatively associated with Eimeria spp. abundance.

CONCLUSIONS

We conclude that amplicon sequencing provides the underused potential for species distinction and simultaneous quantification of parasites in faecal material. The method allowed us to detect a negative effect of Eimeria infection on the body condition of mice in the natural environment.