Implementation barriers and facilitators of Moyo foetal heart rate monitor during labour in public hospitals in Nepal.

BACKGROUND: Globally, every year, approximately 1 million foetal deaths take place during the intrapartum period, fetal heart monitoring (FHRM) and timely intervention can reduce these deaths. OBJECTIVE: This study evaluates the implementation barriers and facilitators of a device, Moyo for FHRM. METHODS: The study adopted a qualitative study design in four hospitals in Nepal where Moyo was implemented for HRM. The study participants were labour room nurses and convenience sampling was used to select them. A total of 20 interviews were done to reach the data saturation. The interview transcripts were translated to English, and qualitative content analysis using deductive approach was applied. RESULTS: Using the deductive approach, the data were organised into three categories i) changes in practice of FHRM, ii) barriers to implementing Moyo and iii) facilitators of implementing Moyo. Moyo improved adherence to intermittent FHRM as the device could handle higher caseloads compared to the previous devices. The implementation of Moyo was hindered by difficulty to organise training ondevice during non-working hours, technical issue of the device, nurse mistrust towards the device and previous experience of poor implementation to similar innovations. Facilitators for implementation included effective training on how to use Moyo, improvement in intrapartum foetal monitoring and improvement in staff morale, ease of using the device, Plan Do Study Act (PDSA) meetings to improve use of Moyo and supportive leadership. CONCLUSION: The change in FHRM practice suggests that the implementation of innovative solution such as Moyo was successful with adequate facilitation, supportive staff attitude and leadership.

Main findings: Before the Moyo implementation, foetal heart rate monitoring was sub-optimal in the hospitals, which changed after introduction of the device, as it helped early display of foetal heart rate in the monitor and supported communication with women during the labour and delivery.Added knowledge: Implementation of Moyo in low-resource setting requires an interdisciplinary approach with continuous support to health care providers on how to correctly read Moyo, maintenance of device and management of false reading.Global health impact for policy and action: The global efforts to accelerate reduce preventable intrapartum related neonatal death requires contextual understanding of clinical context for effective implementation of Moyo.

Inter-facility transfers for emergency obstetrical and neonatal care in rural Madagascar: a cost-effectiveness analysis.

CONTEXT: There is a substantial lack of inter-facility referral systems for emergency obstetrical and neonatal care in rural areas of sub-Saharan Africa. Data on the costs and cost-effectiveness of such systems that reduce preventable maternal and neonatal deaths are scarce. SETTING: We aimed to determine the cost-effectiveness of a non-governmental organisation (NGO)-run inter-facility referral system for emergency obstetrical and neonatal care in rural Southern Madagascar by analysing the characteristics of cases referred through the intervention as well as its costs. DESIGN: We used secondary NGO data, drawn from an NGO's monitoring and financial administration database, including medical and financial records. OUTCOME MEASURES: We performed a descriptive and a cost-effectiveness analysis, including a one-way deterministic sensitivity analysis. RESULTS: 1172 cases were referred over a period of 4 years. The most common referral reasons were obstructed labour, ineffective labour and eclampsia. In total, 48 neonates were referred through the referral system over the study period. Estimated cost per referral was US$336 and the incremental cost-effectiveness ratio (ICER) was US$70 per additional life-year saved (undiscounted, discounted US$137). The sensitivity analysis showed that the intervention was cost-effective for all scenarios with the lowest ICER at US$99 and the highest ICER at US$205 per additional life-year saved. When extrapolated to the population living in the study area, the investment costs of the programme were US$0.13 per person and annual running costs US$0.06 per person. CONCLUSIONS: In our study, the inter-facility referral system was a very cost-effective intervention. Our findings may inform policies, decision-making and implementation strategies for emergency obstetrical and neonatal care referral systems in similar resource-constrained settings.

Artificial Intelligence-Assisted Cancer Status Detection in Radiology Reports.

Cancer research is dependent on accurate and relevant information of patient's medical journey. Data in radiology reports are of extreme value but lack consistent structure for direct use in analytics. At Memorial Sloan Kettering Cancer Center (MSKCC), the radiology reports are curated using gold-standard approach of using human annotators. However, the manual process of curating large volume of retrospective data slows the pace of cancer research. Manual curation process is sensitive to volume of reports, number of data elements and nature of reports and demand appropriate skillset. In this work, we explore state of the art methods in artificial intelligence (AI) and implement end-to-end pipeline for fast and accurate annotation of radiology reports. Language models (LM) are trained using curated data by approaching curation as multiclass or multilabel classification problem. The classification tasks are to predict multiple imaging scan sites, presence of cancer and cancer status from the reports. The trained natural language processing (NLP) model classifiers achieve high weighted F1 score and accuracy. We propose and demonstrate the use of these models to assist in the manual curation process which results in higher accuracy and F1 score with lesser time and cost, thus improving efforts of cancer research. SIGNIFICANCE: Extraction of structured data in radiology for cancer research with manual process is laborious. Using AI for extraction of data elements is achieved using NLP models' assistance is faster and more accurate.

Using Computer Vision to Annotate Video-Recoded Direct Observation of Physical Behavior.

Direct observation is a ground-truth measure for physical behavior, but the high cost limits widespread use. The purpose of this study was to develop and test machine learning methods to recognize aspects of physical behavior and location from videos of human movement: Adults (N = 26, aged 18-59 y) were recorded in their natural environment for two, 2- to 3-h sessions. Trained research assistants annotated videos using commercially available software including the following taxonomies: (1) sedentary versus non-sedentary (two classes); (2) activity type (four classes: sedentary, walking, running, and mixed movement); and (3) activity intensity (four classes: sedentary, light, moderate, and vigorous). Four machine learning approaches were trained and evaluated for each taxonomy. Models were trained on 80% of the videos, validated on 10%, and final accuracy is reported on the remaining 10% of the videos not used in training. Overall accuracy was as follows: 87.4% for Taxonomy 1, 63.1% for Taxonomy 2, and 68.6% for Taxonomy 3. This study shows it is possible to use computer vision to annotate aspects of physical behavior, speeding up the time and reducing labor required for direct observation. Future research should test these machine learning models on larger, independent datasets and take advantage of analysis of video fragments, rather than individual still images.

Validation of a Predictive Model of Vaginal Birth after Caesarean Section in Pakistani Women.

OBJECTIVE: To determine the frequency of successful trials of vaginal birth after Caesarean section (VBAC) using the Flamm and Geiger model. STUDY DESIGN: Observational study. Place and Duration of the Study: Obstetrics and Gynaecology Unit II of Dr. Ruth K.M. Pfau Civil Hospital, Karachi, Pakistan, from August 2022 to January 2023. METHODOLOGY: Women with singleton pregnancy having cephalic presentation, previous one lower segment caesarean section (LSCS), and without any contraindication for vaginal delivery were included. Women bearing foetus having estimated weight >3.5kg, morbid obesity, multiple pregnancies, non-cephalic presentation, placenta praevia, abruptio placentae, uncontrolled maternal comorbidities, or had previous two or more Caesarean sections were excluded. Flamm and Geiger score was applied to record observations of successful and unsuccessful trials. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the score were calculated by ROC curve, along with its area under the curve (AUC) and Youden's index curve, with 95% confidence interval. RESULTS: This study included 258 participants. Successful vaginal delivery was possible in 125 (48.4%) participants, whereas 133 (51.6%) underwent emergency Caesarean section. Cervical dilatation and effacement were the main factors assessing the success. The Flamm and Geiger score of >5 had an area under the curve of 0.813 (0.762-0.864). The score's sensitivity was 75% (67-82%), specificity 76% (68-82%), and PPV of 75% with an accuracy of 76%. CONCLUSION: The Flamm and Geiger score of >5 demonstrated high sensitivity, specificity, PPV, and accuracy in predicting later successful vaginal birth. It is recommended as a promising and valuable tool for assessing VBAC's success in low-resource countries. KEY WORDS: Vaginal birth after caesarean, Flamm and Geiger Model, Indication of caesarean, trial of labour after caesarean, VBAC risk scoring criteria.

Management of impacted fetal head at cesarean delivery.

Globally, more than 1 in 5 women give birth by cesarean delivery, and at least 5% of these births are at full cervical dilatation. In these circumstances, and when labor has been prolonged in the first stage of labor, the fetal head can become low and wedged deep in the woman's pelvis, making it difficult to deliver the baby. This emergency is known as impacted fetal head. These are technically challenging births associated with serious risks to both the woman and the baby. The difficulty in disimpacting the fetal head increases maternal risks of hemorrhage and injury to adjacent organs and may have long-term consequences for future pregnancies. In addition, there can be associated neonatal consequences, such as skull fractures, brain hemorrhage, hypoxic brain injury, and, rarely, perinatal death. Globally, maternity staff are increasingly encountering this emergency, with studies in the United Kingdom suggesting that impacted fetal head may complicate as many as 1 in 10 emergency cesarean deliveries. Moreover, there has been a sharp increase in reports of perinatal brain injuries associated with impaction of the fetal head at cesarean delivery. When an impacted fetal head occurs, the maternity team can employ a range of approaches to help deliver the fetal head, including an assistant (another obstetrician or midwife) pushing the head up from the vagina, delivering the baby feet first (reverse breech extraction), administering tocolysis to relax the uterus, and using a balloon cephalic elevation device (Fetal Pillow) to elevate the baby's head. However, there is currently no consensus on how best to manage these births, resulting in a lack of confidence among maternity staff, variable practice, and potentially avoidable harm in some circumstances. This article examined the evidence for the prevention and management of this critical obstetrical emergency and outlined recommendations for best practices and training.

Management of an acute intracranial emergency in pregnancy.

We report a case of a mid-20s primigravida at 37 weeks' gestation who presented with severe headache and acute neurological deterioration. The CT brain scan showed hydrocephalus caused by a colloid cyst in the third ventricle. The patient underwent emergency placement of an external ventricular drain for decompression of acute hydrocephalus. Four hours later, labour commenced spontaneously, and in view of her neurological status, a decision was taken to perform a caesarean section under general anaesthetics. Four days postpartum, the patient underwent an endoscopic removal of the cyst. Intracranial emergencies during pregnancy are rare and challenging to manage. The mortality rate can be significant. Diagnostic and surgical intervention should not be delayed because of pregnancy. An individualised treatment approach is required with multidisciplinary input. The collaborative efforts of our multidisciplinary team resulted in prompt diagnosis and surgical treatment in this case that resulted in both a healthy mother and child.

A new classification allowing assessment of instrumental vaginal-birth practices.

BACKGROUND: Instrumental vaginal birth, a very common intervention in obstetrics, concerns nearly one in eight women in France. Instrumentally assisted vaginal childbirth can be for maternal and/or fetal indications. Although it reduces recourse to caesarean section, it is subject to risks. Practices concerning instrumental birth are disparate, varying among different practitioners, maternity units and countries, and it is essential to be able to evaluate them. Our objective was to create a classification tool of women requiring instrumental birth to facilitate the analysis of practices within our maternity unit as well as to enable temporal and geographical comparisons. MATERIALS AND METHODS: We propose a simple and robust classification based on the same principles as Robson's classification. It is made up of seven totally inclusive and mutually exclusive groups. Our classification was refined and validated using the Delphi method by a panel of 14 experts from throughout France, and tested in our maternity unit using data from throughout 2021. RESULTS: The seven clinically relevant groups are based on five obstetric criteria (multiplicity, presentation, gestational age, previous type of birth, induction of labor). To classify each woman in a group, five successive questions are posed in a predefined order. The classification has been validated by the experts with highly satisfactory overall agreement. CONCLUSION: In order to improve the quality of care, we propose a tool to standardize the evaluation of instrumental vaginal birth practice (called the "Isère classification", after the county where we work in south-eastern France). It will also facilitate the comparison the practices among different maternity units in a network, a country or even among different countries.

The Potential Role of Wearable Inertial Sensors in Laboring Women with Walking Epidural Analgesia.

There is a growing interest in wearable inertial sensors to monitor and analyze the movements of pregnant women. The noninvasive and discrete nature of these sensors, integrated into devices accumulating large datasets, offers a unique opportunity to study the dynamic changes in movement patterns during the rapid physical transformations induced by pregnancy. However, the final cut of the third trimester of pregnancy, particularly the first stage of labor up to delivery, remains underexplored. The growing popularity of "walking epidural", a neuraxial analgesia method allowing motor function preservation, ambulation, and free movement throughout labor and during delivery, opens new opportunities to study the biomechanics of labor using inertial sensors. Critical research gaps exist in parturient fall prediction and detection during walking epidural and understanding pain dynamics during labor, particularly in the presence of pelvic girdle pain. The analysis of fetal descent, upright positions, and their relationship with dynamic pelvic movements facilitated by walking during labor is another area where inertial sensors can play an interesting role. Moreover, as contemporary obstetrics advocate for less restricted or non-restricted movements during labor, the role of inertial sensors in objectively measuring the quantity and quality of women's movements becomes increasingly important. This includes studying the impact of epidural analgesia on maternal mobility, walking patterns, and associated obstetrical outcomes. In this paper, the potential use of wearable inertial sensors for gait analysis in the first stage of labor is discussed.

Factors Influencing Duration of Breastfeeding: Insights from a Prospective Study of Maternal Health Literacy and Obstetric Practices.

Numerous factors concerning early breastfeeding abandonment have been described, including health literacy (HL). This study's objective was to analyze factors related to early breastfeeding abandonment (<6 months). This prospective multicentric study examined the duration of breastfeeding at 6 months postpartum and was conducted in four different regions of Spain from January 2021 to January 2023. A total of 275 women participated in this study, which focused on maternal HL and obstetric practices. A decrease in the breastfeeding rate was observed from hospital discharge (n = 224, 81.5%) to the sixth month postpartum (n = 117, 42.5%). A Cox regression analysis revealed that inadequate HL levels, lack of mobilization during labour, and induced labour were significantly associated with early breastfeeding cessation (p = 0.022, p = 0.019, and p = 0.010, respectively). The results highlight that women with adequate HL had a 32% lower risk of early breastfeeding abandonment. In comparison, mobilization during labour and induction of labour were linked to a 32.4% reduction and a 53.8% increase in this risk, respectively. These findings emphasize the importance of considering obstetric and HL factors when addressing the breastfeeding duration, indicating opportunities for educational and perinatal care interventions.

Mining co-location patterns of manufacturing firms using Q statistic and additive color mixing.

The agglomeration effect significantly influences firms' site selection. Manufacturing firms often exhibit intricate spatial co-location patterns that are indicative of agglomerations due to their reliance on material input and product output across various subdivisions of manufacture. In this study, we present an analytical approach employing the Q statistic and additive color mixing visualization to assess co-location patterns of manufacturing firms. We identified frequent pairs and triplets of manufacturing divisions, mapping them to reveal distinct categories: labor-intensive clusters, upstream/downstream industrial chains, and technology-spillover clusters. These agglomeration categories concentrate in different regions of the city. Policy implications are proposed to promote the upgrade of labor-intensive divisions, enhance the operational efficiency of upstream/downstream industrial chains, and reinforce the spillover effects of technology-intensive divisions.

Experiences of women with physical disabilities in labor and delivery assistance.

OBJECTIVE: To understand the meaning attributed by women with physical disabilities to the health care received and expected during labor and delivery. METHODS: Qualitative study, based on Social Network Theory, conducted through an online workshop in April 2022, with the participation of six women with physical disabilities. Data, collected through the focus group technique, underwent thematic content analysis with the assistance of the IRaMuTeQ tool. RESULTS: Three thematic categories emerged: Challenges experienced during pregnancy; The experience within the maternity ward; and, The importance of social networks. The assistance provided by healthcare professionals sometimes differed between what was expected and what was received by women with physical disabilities during labor and delivery. FINAL CONSIDERATIONS: Experiences were predominantly negative, resulting from inappropriate professional conduct due to ableist attitudes. Support from members of social networks is crucial for preventing stressors.

Anaphylaxis-induced premature uterine contractions: a case report and literature review.

BACKGROUND: Preterm labor is caused by multiple etiologies, including intra-amniotic infection and/or intra-amniotic inflammation, vascular disorders, cervical disease, decidual senescence, and breakdown of maternal-fetal tolerance. Accumulating evidence in vivo and in vitro has shown that an allergic reaction, including anaphylaxis, can induce preterm uterine contractions. This report describes a case of a pregnant woman who developed anaphylaxis and regular uterine contractions after the ingestion of a strawberry-coated biscuit. We also review the mechanism of allergic reaction (hypersensitivity)-induced preterm labor. Case presentation A 31-year-old woman (gravida 1, para 0) at 30+2 weeks of gestation was admitted to the labor and delivery unit with regular uterine contractions and anaphylactic symptoms after she ingested a strawberry-coated biscuit as a snack. The uterine contractions resolved after the treatment of anaphylaxis by administering antihistamines and epinephrine. The patient subsequently delivered at 39+3 weeks of gestation. The amniotic fluid profile showed no infection or inflammation. A postpartum skin-prick test confirmed a positive type 1 hypersensitivity reaction to the strawberry-coated biscuit. CONCLUSIONS: We report a case of anaphylaxis-induced uterine contractility in which uterine contractions subsided after the treatment of anaphylaxis. The absence of intra-amniotic infection and/or intra-amniotic inflammation and the cause of the anaphylaxis were confirmed. Our findings indicate that maternal allergic reactions may be one of the mechanisms of preterm labor.

Evaluation of Leukocyte Chemotaxis Induced by Human Fetal Membranes in an In Vitro Model.

Leukocyte infiltration into the maternal-fetal interface is a consequence of the robust inflammation in the gestational tissues during term labor and preterm labor with or without infection. During pregnancy, the fetal membranes act as a physical barrier that isolates the fetus into the amniotic cavity, keeping it in an optimal environment for its development. In addition, the fetal membranes possess immunological competencies such as the secretion of cytokines and chemokines in response to different stimuli. Clinical and experimental evidence indicates that these tissues are involved in the extensive chemotaxis of immune cells in normal or pathological conditions.Few studies have evaluated the chemotactic capacities of the fetal membranes considering that this tissue is composed of two adjacent tissues, the amnion and the chorion, which have different characteristics. Although these tissues function as a unit, their response is complex since there is an interaction between them, where each tissue contributes differently. The protocol described here allows us to evaluate the in vitro chemotactic capacities of fetal membranes in response to various applied stimuli, considering the contribution of each of their components (amnion and choriodecidua) using a Boyden chamber assay and phenotyping the chemo-attracted leukocytes by flow cytometry.

The Rate of Cesarean Section and Newborn Apgar score In Two Types Of Physiological Delivery and Facilitated Delivery in Mothers with First Pregnancy.

INTRODUCTION: Labor pain is often severe and its lack of relief can have a bad effect on the mother's physiological condition. Accurate measurement and appropriate treatment of pain is an important problem. There are several choices for the control of labor pain, however, each method has its own risks and benefits regarding its efficiency and availability; therefore, the aim of this study was to compare the rate of cesarean section and newborn Apgar in two types of physiological delivery and facilitated delivery among mothers with first delivery. MATERIALS AND METHODS: This case-control study was conducted on mothers aged 18-35 who went to the hospital to give birth. In the present study, 8 sessions (90 minutes each session) were held to prepare mothers from 20 to 37 weeks of pregnancy. Candidate mothers for physiological labor were hospitalized without intervention and in the active phase of labor, and their labor stages were planned physiologically. The pain intensity of mothers was asked using a scale (VAS) during labor every half hour (according to the pain scale from 0 to 10). The average score of pain during the first and second stages, type of delivery, amount of postpartum bleeding, length of the stage of delivery, infection, and fever after delivery, Apgar score of the baby, perineal status, mother's satisfaction and the rate of hospitalization of the baby in NICU were recorded. Furthermore, the level of satisfaction with childbirth was evaluated with the help of Mackey's standard satisfaction questionnaire. RESULTS: The average age of the physiological delivery and facilitated delivery groups was 26.37 ± 5.23 years and 26.58 ± 5.79 years, respectively. Physiological delivery significantly required less conversion to cesarean section. The most common etiology was caesarean section in the physiologic labor group, and a drop in NST was reported in the facilitated group. The analysis of the results demonstrated no significant difference between the etiology of cesarean section and the study groups. The Apgar score in the physiological group was significantly better than the facilitated group. There is no significant relationship between the study group and the amount of bleeding during delivery. Furthermore, there was no significant relationship between the study group and the incidence of postpartum infection. The rate of NICU admission in the facilitated group was found to be significantly higher than physiological delivery. CONCLUSION: The results of this study revealed that the rate of caesarean section, the rate of need for NICU, newborns' Apgar score, and the rate of mothers' satisfaction in physiological delivery group were significantly different from the other groups, but the rate of wound infection and the amount of postpartum bleeding in the two groups showed a significant difference.

Multichannel mapping of in vivo rat uterine myometrium exhibits both high and low frequency electrical activity in non-pregnancy.

The uterus exhibits intermittent electrophysiological activity in vivo. Although most active during labor, the non-pregnant uterus can exhibit activity of comparable magnitude to the early stages of labor. In this study, two types of flexible electrodes were utilized to measure the electrical activity of uterine smooth muscle in vivo in anesthetized, non-pregnant rats. Flexible printed circuit electrodes were placed on the serosal surface of the uterine horn of six anesthetized rats. Electrical activity was recorded for a duration of 20-30 min. Activity contained two components: high frequency activity (bursts) and an underlying low frequency 'slow wave' which occurred concurrently. These components had dominant frequencies of 6.82 ± 0.63 Hz for the burst frequency and 0.032 ± 0.0055 Hz for the slow wave frequency. There was a mean burst occurrence rate of 0.76 ± 0.23 bursts per minute and mean burst duration of 20.1 ± 6.5 s. The use of multiple high-resolution electrodes enabled 2D mapping of the initiation and propagation of activity along the uterine horn. This in vivo approach has the potential to provide the organ level detail to help interpret non-invasive body surface recordings.

Research on Human-Robot Collaboration Method for Parallel Robots Oriented to Segment Docking.

In the field of aerospace, large and heavy cabin segments present a significant challenge in assembling space engines. The substantial inertial force of cabin segments' mass often leads to unexpected motion during docking, resulting in segment collisions, making it challenging to ensure the accuracy and quality of engine segment docking. While traditional manual docking leverages workers' expertise, the intensity of the labor and low productivity are impractical for real-world applications. Human-robot collaboration can effectively integrate the advantages of humans and robots. Parallel robots, known for their high precision and load-bearing capacity, are extensively used in precision assembly under heavy load conditions. Therefore, human-parallel-robot collaboration is an excellent solution for such problems. In this paper, a framework is proposed that is easy to realize in production, using human-parallel-robot collaboration technology for cabin segment docking. A fractional-order variable damping admittance control and an inverse dynamics robust controller are proposed to enhance the robot's compliance, responsiveness, and trajectory tracking accuracy during collaborative assembly. This allows operators to dynamically adjust the robot's motion in real-time, counterbalancing inertial forces and preventing collisions between segments. Segment docking assembly experiments are performed using the Stewart platform in this study. The results show that the proposed method allows the robot to swiftly respond to interaction forces, maintaining compliance and stable motion accuracy even under unknown interaction forces.

Lightweight model-based sheep face recognition via face image recording channel.

The accurate identification of individual sheep is a crucial prerequisite for establishing digital sheep farms and precision livestock farming. Currently, deep learning technology provides an efficient and non-contact method for sheep identity recognition. In particular, convolutional neural networks can be used to learn features of sheep faces to determine their corresponding identities. However, the existing sheep face recognition models face problems such as large model size, and high computational costs, making it difficult to meet the requirements of practical applications. In response to these issues, we introduce a lightweight sheep face recognition model called YOLOv7-Sheep Face Recognition (YOLOv7-SFR). Considering the labor-intensive nature associated with manually capturing sheep face images, we developed a face image recording channel to streamline the process and improve efficiency. This study collected facial images of 50 Small-tailed Han sheep through a recording channel. The experimental sheep ranged in age from 1 to 3 yr, with an average weight of 63.1 kg. Employing data augmentation methods further enhanced the original images, resulting in a total of 22,000 sheep face images. Ultimately, a sheep face dataset was established. To achieve lightweight improvement and improve the performance of the recognition model, a variety of improvement strategies were adopted. Specifically, we introduced the shuffle attention module into the backbone and fused the Dyhead module with the model's detection head. By combining multiple attention mechanisms, we improved the model's ability to learn target features. Additionally, the traditional convolutions in the backbone and neck were replaced with depthwise separable convolutions. Finally, leveraging knowledge distillation, we enhanced its performance further by employing You Only Look Once version 7 (YOLOv7) as the teacher model and YOLOv7-SFR as the student model. The training results indicate that our proposed approach achieved the best performance on the sheep face dataset, with a mean average precision@0.5 of 96.9%. The model size and average recognition time were 11.3 MB and 3.6 ms, respectively. Compared to YOLOv7-tiny, YOLOv7-SFR showed a 2.1% improvement in mean average precision@0.5, along with a 5.8% reduction in model size and a 42.9% reduction in average recognition time. The research results are expected to drive the practical applications of sheep face recognition technology.

Accurate identification of individual sheep is a crucial prerequisite for establishing digital sheep farms and precision livestock farming. In this study, we developed a lightweight sheep face recognition model, YOLOv7-SFR. Utilizing a face image recording channel, we efficiently collected facial images from 50 experimental sheep, resulting in a comprehensive sheep face dataset. Training results demonstrated that YOLOv7-SFR surpassed state-of-the-art lightweight sheep face recognition models, achieving a mean average precision@0.5 of 96.9%. Notably, the model size and average recognition time of YOLOv7-SFR were merely 11.3 MB and 3.6 ms, respectively. In summary, YOLOv7-SFR strikes an optimal balance between performance, model size, and recognition speed, offering promising practical applications for sheep face recognition technology. This study employs deep learning for sheep face recognition tasks, ensuring the welfare of sheep in the realm of digital agriculture and automation practices.