RESUMEN

BACKGROUND: Falls are the leading cause of injuries in older adults. Environmental objects (such as furniture, walls, and handrails) may act as hazards or facilitators to balance maintenance and safe landing. There is lack of objective evidence on how older adults interact with objects during falls. We addressed this gap by characterizing body part contacts with objects other than the floor during real-life falls in long-term care. METHODS: We analyzed videos of 1759 falls experienced by 584 residents to characterize the prevalence of contacts with objects before, during, and after fall initiation. Using generalized estimating equations, we compared the prevalence of falls with versus without contact to objects after fall initiation. Using linear mixed models, we tested for differences across body parts in the probability of contacting objects after fall initiation. RESULTS: In nearly one-third of falls, interactions with objects (e.g., trips over objects, loss of support with objects) or with other people (e.g., being pushed by another person) had a primary role in causing imbalance and initiating the fall. After fall initiation, participants contacted objects in 60% of falls, with intentional hand contacts to objects via reach-to-grasp or bracing being the most common type of interaction (Probability ± SE = 0.32 ± 0.01), followed by unintentional impacts to the torso (0.21 ± 0.01) and head (0.16 ± 0.01). Intentional hand contact to an object was more common during forward than backward falls (p < 0.001), while head and torso contacts to objects were more common during backward and sideways falls than forward falls (multiple p values ≤ 0.003). The hand most often contacted chairs, wheelchairs or couches, followed by tables or counters, walls, other people, walkers, and handrails. The head, torso, and shoulder most often contacted a wall. CONCLUSIONS: Most falls in long-term care involved contacts with objects other than the ground, indicating that complex environments often accompany falls in long-term care. Higher probabilities of intentional hand contacts in forward falls, versus unintentional head and torso impacts in backward and sideways falls may reflect the influence of being able to visualize and adjust one's falling patterns to nearby objects.

Asunto(s)

RESUMEN

Video integrity is a crucial aspect of forensic science that guarantees the reliability and validity of visual evidence used in court proceedings. In an era where digital alteration tools are readily available, ensuring that video recordings remain unmodified is essential to upholding the integrity of the legal system. Our technique offers a new, simple way to check the integrity of video data. Our approach makes use of the BLAKE2b hash function, the blockchain, and the Edwards Curve Digital Signature Algorithm. Video segments are pre-recorded video clips for which signatures are generated and kept in chronological blocks. To provide an additional degree of protection, the signature from the prior block is kept in the present block. These signatures are validated at the moment of validation. According to experimental data, our method performs faster and more securely than state-of-the-art approaches. With negligible extra storage requirements, our approach can detect every kind of counterfeit on any video file, by anybody, at any time. Our security analysis further demonstrates that our approach is resistant to a wide range of attacks, such as side channel, collision, key substitution, and chosen message assaults.

RESUMEN

The advent of computer vision technology and increased usage of video cameras in clinical settings have facilitated advancements in movement disorder analysis. This review investigated these advancements in terms of providing practical, low-cost solutions for the diagnosis and analysis of movement disorders, such as Parkinson's disease, ataxia, dyskinesia, and Tourette syndrome. Traditional diagnostic methods for movement disorders are typically reliant on the subjective assessment of motor symptoms, which poses inherent challenges. Furthermore, early symptoms are often overlooked, and overlapping symptoms across diseases can complicate early diagnosis. Consequently, deep learning has been used for the objective video-based analysis of movement disorders. This study systematically reviewed the latest advancements in automatic two-dimensional & three-dimensional video analysis using deep learning for movement disorders. We comprehensively analyzed the literature published until September 2023 by searching the Web of Science, PubMed, Scopus, and Embase databases. We identified 68 relevant studies and extracted information on their objectives, datasets, modalities, and methodologies. The study aimed to identify, catalogue, and present the most significant advancements, offering a consolidated knowledge base on the role of video analysis and deep learning in movement disorder analysis. First, the objectives, including specific PD symptom quantification, ataxia assessment, cerebral palsy assessment, gait disorder analysis, tremor assessment, tic detection (in the context of Tourette syndrome), dystonia assessment, and abnormal movement recognition were discussed. Thereafter, the datasets used in the study were examined. Subsequently, video modalities and deep learning methodologies related to the topic were investigated. Finally, the challenges and opportunities in terms of datasets, interpretability, evaluation methods, and home/remote monitoring were discussed.

Asunto(s)

RESUMEN

Instrumented mouthguards (iMGs) are widely applied to measure head acceleration event (HAE) exposure in sports. Despite laboratory validation, on-field factors including potential sensor skull-decoupling and spurious recordings limit data accuracy. Video analysis can provide complementary information to verify sensor data but lacks quantitative kinematics reference information and suffers from subjectivity. The purpose of this study was to develop a rigorous multi-stage screening procedure, combining iMG and video as independent measurements, aimed at improving the quality of on-field HAE exposure measurements. We deployed iMGs and gathered video recordings in a complete university men's ice hockey varsity season. We developed a four-stage process that involves independent video and sensor data collection (Stage I), general screening (Stage II), cross verification (Stage III), and coupling verification (Stage IV). Stage I yielded 24,596 iMG acceleration events (AEs) and 17,098 potential video HAEs from all games. Approximately 2.5% of iMG AEs were categorized as cross-verified and coupled iMG HAEs after Stage IV, and less than 1/5 of confirmed or probable video HAEs were cross-verified with iMG data during stage III. From Stage I to IV, we observed lower peak kinematics (median peak linear acceleration from 36.0 to 10.9 g; median peak angular acceleration from 3922 to 942 rad/s2) and reduced high-frequency signals, indicative of potential reduction in kinematic noise. Our study proposes a rigorous process for on-field data screening and provides quantitative evidence of data quality improvements using this process. Ensuring data quality is critical in further investigation of potential brain injury risk using HAE exposure data.

Asunto(s)

RESUMEN

The sensitivity of DNA analysis has progressed to the point that trace levels of DNA, originating from only a few cells, can generate informative profiles. This means that virtually any item or surface can be sampled with a reasonable chance of obtaining a DNA profile. As the presence of DNA does not suggest how it was deposited, questions are often raised as to how the DNA came to be at a particular location and the activity that led to its deposition. Therefore, understanding different modes of DNA deposition, reflective of realistic forensic casework situations, is critical for proper evaluation of DNA results in court. This study aimed to follow the movements of DNA to and from individuals and common household surfaces in a residential premises, while socially interacting. This took place over an hour and involved four participants, with known shedder status, designated as visitors (a male and a female) and hosts (a male and a female), who engaged in the activity of playing a board game while being served food. During the study, the participants were instructed to use the toilet on a single occasion to assess the transfer of DNA to new and unused underwear that was provided. All contacts made by the participants in the dining room and kitchen were video recorded to follow the movements of DNA. Samples were collected based on the history of contact, which included hands, fingernails and penile swabs. Direct contacts resulted in detectable transfer (LR > 1) in 87 % (87/100) of the non-intimate samples and clothing. For surfaces touched by multiple participants, DNA from the person who made the last contact was not always detectable. The duration and number of contacts did not significantly affect the detection of the person contacting the item. On the other hand, presence of background DNA and participant's shedder status appear to play an important role. Further, unknown contributors were detected in the majority of samples. Finally, indirect transfer was observed on a number of occasions including co-habiting partners of guests who were not present at the study location. The results of this study may assist with decision making for exhibit selection or targeting areas for sampling within the home environment. Our findings can also be used in conjunction with previous literature to develop activity-level evaluations in such situations where the source of the DNA is conceded, but the mode of deposition is disputed.

Asunto(s)

RESUMEN

For patients at risk of developing either lung cancer or colorectal cancer, the identification of suspect lesions in endoscopic video is an important procedure. The physician performs an endoscopic exam by navigating an endoscope through the organ of interest, be it the lungs or intestinal tract, and performs a visual inspection of the endoscopic video stream to identify lesions. Unfortunately, this entails a tedious, error-prone search over a lengthy video sequence. We propose a deep learning architecture that enables the real-time detection and segmentation of lesion regions from endoscopic video, with our experiments focused on autofluorescence bronchoscopy (AFB) for the lungs and colonoscopy for the intestinal tract. Our architecture, dubbed ESFPNet, draws on a pretrained Mix Transformer (MiT) encoder and a decoder structure that incorporates a new Efficient Stage-Wise Feature Pyramid (ESFP) to promote accurate lesion segmentation. In comparison to existing deep learning models, the ESFPNet model gave superior lesion segmentation performance for an AFB dataset. It also produced superior segmentation results for three widely used public colonoscopy databases and nearly the best results for two other public colonoscopy databases. In addition, the lightweight ESFPNet architecture requires fewer model parameters and less computation than other competing models, enabling the real-time analysis of input video frames. Overall, these studies point to the combined superior analysis performance and architectural efficiency of the ESFPNet for endoscopic video analysis. Lastly, additional experiments with the public colonoscopy databases demonstrate the learning ability and generalizability of ESFPNet, implying that the model could be effective for region segmentation in other domains.

RESUMEN

Background: The literature on the running kinematics of youth distance runners is limited. Purpose: We sought to describe 2-dimensional (2D) video analysis of running kinematics in healthy adolescent distance runners, which has not been previously described. Methods: We conducted an observational study of healthy, competitive runners between the ages of 14 and 18 years, prospectively recruited through local running clubs and our hospital's outreach between August 2019 and July 2023. Participants ran on a treadmill at a self-selected speed with markers attached to the thorax, pelvis, and lower extremities. A high-definition video camera recorded the runners in the sagittal and frontal planes. Kinematic measurements were completed using Dartfish software and reported as means and standard deviations. Results: Of the 53 participants (51% boys, mean age: 16.0 ± 1.4 years) included in the 2D running analysis, 91% ran with a rearfoot strike pattern, with a mean foot inclination angle of 10.2° ± 6.2°. Knee flexion angle at initial contact was 13.2° ± 3.8°, tibia inclination angle was 8.5° ± 3.2°, and peak knee flexion was 44.5° ± 3.6°. Cadence was 168.7° ± 8.6°. Contralateral pelvic drop was 6.0° ± 2.2° and peak rearfoot eversion was 11.8° ± 3.6°. Conclusions: This study is the first to describe running kinematics as captured by 2D video in healthy adolescent runners and to identify kinematic variables that may differ from those of adult runners. Further research is required to determine if adult recommendations are applicable to adolescent populations.

RESUMEN

Information about the concordance between dynamic emotional experiences and objective signals is practically useful. Previous studies have shown that valence dynamics can be estimated by recording electrical activity from the muscles in the brows and cheeks. However, whether facial actions based on video data and analyzed without electrodes can be used for sensing emotion dynamics remains unknown. We investigated this issue by recording video of participants' faces and obtaining dynamic valence and arousal ratings while they observed emotional films. Action units (AUs) 04 (i.e., brow lowering) and 12 (i.e., lip-corner pulling), detected through an automated analysis of the video data, were negatively and positively correlated with dynamic ratings of subjective valence, respectively. Several other AUs were also correlated with dynamic valence or arousal ratings. Random forest regression modeling, interpreted using the SHapley Additive exPlanation tool, revealed non-linear associations between the AUs and dynamic ratings of valence or arousal. These results suggest that an automated analysis of facial expression video data can be used to estimate dynamic emotional states, which could be applied in various fields including mental health diagnosis, security monitoring, and education.

Asunto(s)

RESUMEN

OBJECTIVES: To determine whether spotters with medical training and experience in managing concussion have higher inter-rater reliability and accuracy than non-medical personnel when identifying video signs associated with concussion in Australian football. DESIGN: Retrospective cohort study. METHODS: Video clips were collected of all impacts potentially resulting in concussion during 2012 and 2013 Australian Football League (AFL) seasons. Raters were divided into medical doctors and a non-medical group comprising allied health practitioners (physiotherapists) and non-medical/non-allied health personnel (performance analysts). Raters assessed 102 randomly selected videos for signs of concussion. The inter-rater reliability was calculated. Sensitivity, specificity, positive and negative predictive values were calculated by comparing the rater responses to the consensus opinion from two highly experienced clinicians with expertise in concussion. RESULTS: No statistically significant difference in inter-rater reliability was observed between the medical doctors and the non-medical group. Both groups demonstrated good to excellent agreement for slow to get up, clutching at head/face and facial injury. Both groups displayed intra-class coefficient >0.55 for no protective action-floppy, loss of responsiveness, and motor incoordination, and displayed lowest agreement for no protective action-tonic posturing, impact seizure and blank/vacant look. No statistically significant difference was found between the groups for sensitivity, specificity, positive and negative predictive values for correctly classifying video signs compared to the expert consensus opinion. CONCLUSIONS: After completing sufficient standardised training and testing, medical and non-medical personnel demonstrate comparable reliability in identifying video signs of concussion in professional Australian football and may be suitable for the role of video spotter.

RESUMEN

OBJECTIVES: To undertake a systematic analysis of 17 medical attention and time-loss lateral ankle ligament sprain (LALS) events from televised Australian professional netball games during the 2020-2023 seasons. DESIGN: Case series. METHODS: Three analysts independently assessed the video footage and then convened to review and discuss each case until a consensus was reached. RESULTS: When in possession (7 cases) a player was commonly performing an agility-based manoeuvre to break free from an opponent and reposition themselves to be a passing option (5/7 cases). When out of possession (10 cases) a player was either attempting to intercept a pass (6 cases) or marking an opponent (4 cases). Players tended to land on the anterior one-third of the plantar surface of the foot - forefoot or shoe tip (7 cases). Players often landed on either the ground (7 cases) or the opponent's shoe then the ground (8 cases). In 9 cases the ankle-foot was considered to be in a neutral alignment in the frontal plane at landing. At the estimated index frame the players' weight tended to be all on the foot on the injured side (11 cases) or favouring the foot on the injured side (5 cases). Inversion and adduction was a common injury mechanism. Plantar-flexion was rarely involved. CONCLUSION: Landing on the anterior one-third of the plantar surface of the foot and subsequent weight transference onto the injured limb side was more important than ankle-foot inversion at initial ground contact. Exercises involving external perturbations that challenge the control of frontal and transverse plane ankle-foot motion and improve proprioception, neuromuscular control, and dynamic balance are warranted.

Asunto(s)

RESUMEN

BACKGROUND: Laparoscopic pancreatoduodenectomy (LPD) is one of the most challenging operations and has a long learning curve. Artificial intelligence (AI) automated surgical phase recognition in intraoperative videos has many potential applications in surgical education, helping shorten the learning curve, but no study has made this breakthrough in LPD. Herein, we aimed to build AI models to recognize the surgical phase in LPD and explore the performance characteristics of AI models. METHODS: Among 69 LPD videos from a single surgical team, we used 42 in the building group to establish the models and used the remaining 27 videos in the analysis group to assess the models' performance characteristics. We annotated 13 surgical phases of LPD, including 4 key phases and 9 necessary phases. Two minimal invasive pancreatic surgeons annotated all the videos. We built two AI models for the key phase and necessary phase recognition, based on convolutional neural networks. The overall performance of the AI models was determined mainly by mean average precision (mAP). RESULTS: Overall mAPs of the AI models in the test set of the building group were 89.7% and 84.7% for key phases and necessary phases, respectively. In the 27-video analysis group, overall mAPs were 86.8% and 71.2%, with maximum mAPs of 98.1% and 93.9%. We found commonalities between the error of model recognition and the differences of surgeon annotation, and the AI model exhibited bad performance in cases with anatomic variation or lesion involvement with adjacent organs. CONCLUSIONS: AI automated surgical phase recognition can be achieved in LPD, with outstanding performance in selective cases. This breakthrough may be the first step toward AI- and video-based surgical education in more complex surgeries.

Asunto(s)

RESUMEN

An animal's locomotor rate is an important indicator of its motility. In studies of the nematode C. elegans, assays of the frequency of body bending waves have often been used to discern the effects of mutations, drugs, or aging. Traditional manual methods for measuring locomotor frequency are low in throughput and subject to human error. Most current automated methods depend on image segmentation, which requires high image quality and is prone to errors. Here, we describe an algorithm for automated estimation of C. elegans locomotor frequency using image invariants, i.e., shape-based parameters that are independent of object translation, rotation, and scaling. For each video frame, the method calculates a combination of 8 Hu's moment invariants and a set of Maximally Stable Extremal Regions (MSER) invariants. The algorithm then calculates the locomotor frequency by computing the autocorrelation of the time sequence of the invariant ensemble. Results of our method show excellent agreement with manual or segmentation-based results over a wide range of frequencies. We show that compared to a segmentation-based method that analyzes a worm's shape and a method based on video covariance, our technique is more robust to low image quality and background noise. We demonstrate the system's capabilities by testing the effects of serotonin and serotonin pathway mutations on C. elegans locomotor frequency.

RESUMEN

Species categorical authentication of accelerants has traditionally relied on fire debris analysis. To explore a novel method for identifying the accelerants species, four commonly used accelerants for arson were loaded onto different substrates and ignited at different locations. The entire combustion process was recorded and flame characteristics were analyzed. The results showed that the probability density function (PDF) of flame apex angle counts within a certain period after ignition can be used to distinguish accelerant species, and this method is not affected by accelerant loading amount, ignition location, and substrate, demonstrating strong stability and universality, while the temporal variation of flame area and the value obtained by dividing half of the flame width by the flame height (tangent of flame cone angle) can effectively differentiate gasoline and diesel. The utilization of flame characteristics for identifying accelerants species holds significant implications for arson investigation.

RESUMEN

The tackle contest is the most common and most injurious match contact event in rugby and is an indicator of performance. Tackle Ready is World Rugby's tackle technique education program. Limited research has characterized the tackle contest in women's rugby. The purpose of this study is to: (1) identify the match situational characteristics, ball-carrier and tackler technical actions demonstrated in elite women's Rugby Union and (2) to determine the extent to which Tackle Ready recommended tackle techniques were exhibited. Technical characteristics for 1500 tackle events in the 2022-2023 Women's Six Nations Championship were visually assessed according to a predefined coding framework and the Tackle Ready program. Tackles lacked full completion (0.2%) of the 22 coded Tackle Ready techniques with 47% of the recommended techniques demonstrated in each tackle on average (range 15%-98%). A high proportion of tackles involved two defenders (48%), approaching ball-carriers from the side (38%) or oblique angles (39%), in an upright position (30%), and with initial contact made with the arm (51%). Incorrect pre-contact head positioning and head placement upon contact accounted for 50% and 15% of tackles, respectively, and there was a mean of 14 (95% CI 11-18) head and neck contacts to a tackler and 18 (95% CI 14-22) head and neck contacts to a ball-carrier per game. Targeted interventions to encourage adoption of recommended techniques are needed to reduce tackle-related injury risk in women's rugby. This study provides valuable context for future discussion across law enforcement, coach education and gender-specific tackle coaching in the women's game.

Asunto(s)

RESUMEN

PURPOSE: To describe the injury mechanism and situational patterns of severe (absence >28 days) hamstring muscle injuries in professional male and female football (soccer) players. METHODS: The data for males were sourced from Serie A clubs participating in both national and international competitions from 2018 to 2021. For the female cohort, hamstring injuries were identified during matches of the top national/international competitions from 2017 to 2023. Video footage was obtained, and three raters categorised injury mechanisms and situational patterns. Injuries were also examined according to the month, minute and location. RESULTS: A total of 129 severe hamstring injuries were identified, with 64 occurring in females and 65 in males. Video analysis was possible for 29 (45%) female cases and 61 (94%) male cases. Female injuries had longer lay-off times (97.8 ± 77.1 days) than males (39.6 ± 20.9 days). Females had a higher proportion of indirect contact injuries (34%) than males (13%) and a lower proportion of non-contact injuries (66% vs. 87%). Four situational patterns were identified: running was the most common for both sexes, representing 59% of female injuries and 41% of male injuries. Over-stretching injuries were split across open and CKC scenarios but collectively explained nearly half (48%) of male injuries but only one in five (21%) female injuries. Kicking injuries had a higher proportion in females (17%) than males (10%). Injuries were more common in the second half for females and the first half for males. CONCLUSION: Females had a higher proportion of indirect contact, running and kicking injuries and a lower proportion of non-contact and stretch-type injuries than males. Understanding injury patterns can inform tailored prevention programs, considering sex-specific differences. LEVEL OF EVIDENCE: Level IV.

RESUMEN

This study aims to examine and describe the characteristics of potential injury situations during a men's professional international tournament quantified using the FIFA Football Language Medical Coding. A prospective study was conducted during the 64 matches of the FIFA World Cup Qatar 2022™, during which five analysts recorded potential injury situations from video analysis. "Potential injuries" were recorded when players stayed down > 5 s and/or requested medical attention. Characteristics were further recorded for variables such as opponent's action and body location. In total, 720 potential injury situations were recorded of which 139 required medical assessment. The actions which resulted in most potential injuries were running while receiving a pass (74; 10%), while passing the ball (59; 8%), and while progressing with the ball (48; 7%). Duels and ball progression led to a potential injury in 3.0% and 2.1% of all similar actions in total. Both aerial duels and ball progression led to an potential injury that required medical assessment on 0.4% of occasions. Most potential injuries involved the head (149; 21%), foot (120; 17%), or lower leg (110; 15%) with most medical assessments of the head (35; 25%), lower leg (17; 12%), and knee (15; 11%) with a median duration of 47 seconds (IQR 28-61). This study provides a detailed overview of match circumstances that may have a higher injury risk. Although some variables within the coding system need improvement to increase reliability, its use will allow a more detailed comparison of differences between high-risk player actions leading to injury and those that do not, which can improve future prevention strategies.

RESUMEN

Background: The side hop test (SHT) measures the number of jumps performed over 30 s. Although this measure has demonstrated its value in clinical practice, the temporal parameters of the SHT allow for a deeper analysis of the execution strategy. The aim of this study is to assess the reliability and construct validity of contact time parameters during the SHT recorded by a video analysis system in anterior cruciate ligament reconstructed (ACLR) patients. Methods: We investigated the reliability (intra-rater, standard error of measurement (SEM), and minimum detectable change (MDC)), discriminant validity (operated (OP) versus non-operated (NOP) side), and convergent validity (relationship with strength and psychological readiness) of SHT contact time parameters, number of valid hops and limb symmetry index (LSI) in 38 ACLR patients. Contact time parameters are presented as mean, standard deviation (SD), and coefficient of variation (CV) of contact time. Results: Intra-tester reliability was good to excellent for all contact time parameters. For discriminant validity, the mean and SD contact times of the OP leg were significantly longer than those of the NOP leg, although the difference was smaller than the SEM and MDC values. The number of valid jumps and CV contact time parameters were not significantly different. Isokinetic quadriceps strength (60°/s) was strongly correlated with mean contact time for both legs. However, psychological readiness was not correlated with any of the contact time parameters. Conclusions: Temporal parameters of the SHT measured on video analysis are valid and reliable parameters to assess the performance strategy of the SHT. The results should be interpreted with caution regarding the SEM and MDC values. Further studies are needed to measure criterion validity, inter-rater reliability, and responsiveness.

RESUMEN

BACKGROUND: A growing body of evidence suggests that children of mothers with eating disorders (EDs) have a greater risk of early feeding problems. Recognizing and reacting adequately to the infant's signals during feeding is crucial for the child's development of internal and external regulatory mechanisms of food intake. Parental EDs might affect this ability. Therefore, we investigated the quality of mother-infant interactions during feeding using video recording and a structured coding system. METHODS: The data of this pilot study was collected in a prospective cohort study investigating the influence of maternal EDs on child outcomes. Twenty women with ED history and 31 control women were videotaped while feeding their infant during a main meal at ten months postpartum. The mother-infant interactions were evaluated by two raters using the Chatoor Feeding Scale. We assessed birth outcomes, the mother's ED and depression status, breastfeeding practices, infant feeding problems and infant temperament by maternal self-report. RESULTS: Mothers with and without ED history scored very similar on the Feeding Scale, however mothers from the control group experienced more struggle for control with their infants during feeding (p = 0.046) and made more negative comments about the infant's food intake (p = 0.010). Mothers with ED history were more concerned about infant feeding at three months postpartum and reported significantly more problems with solid foods in their children. Birth outcomes were comparable between groups, except for lower weight-for-length birth percentiles in children of women with ED history. CONCLUSION: Whilst examined mothers with ED history are more concerned about feeding their children, ED psychopathology does not affect the quality of mother-infant interaction during feeding at the transition to autonomous eating at ten months of age.

Asunto(s)

RESUMEN

Eating behavior is a key factor for nutritional intake and plays a significant role in the development of eating disorders and obesity. The standard methods to detect eating behavior events (i.e., bites and chews) from video recordings rely on manual annotation, which lacks objective assessment and standardization. Yet, video recordings of eating episodes provide a non-invasive and scalable source for automation. Here, we present a rule-based system to count bites automatically from video recordings with 468 3D facial key points. We tested the performance against manual annotation in 164 videos from 15 participants. The system can count bites with 79% accuracy when annotation is available, and 71.4% when annotation is unavailable. The system showed consistent performance across varying food textures. Eating behavior researchers can use this automated and objective system to replace manual bite count annotation, provided the system's error is acceptable for the purpose of their study. Utilizing our approach enables real-time bite counting, thereby promoting interventions for healthy eating behaviors. Future studies in this area should explore rule-based systems and machine learning methods with 3D facial key points to extend the automated analysis to other eating events while providing accuracy, interpretability, generalizability, and low computational requirements.