Cluster analysis of novel isometric strength measures produces a valid and evidence-based classification structure for wheelchair track racing

Mechanisms of Exercise Intolerance Across the Breast Cancer Continuum: A Pooled Analysis of Individual Patient Data

PURPOSE

The purpose of this study is to evaluate the prevalence of abnormal cardiopulmonary responses to exercise and pathophysiological mechanism(s) underpinning exercise intolerance across the continuum of breast cancer (BC) care from diagnosis to metastatic disease.

METHODS

Individual participant data from four randomized trials spanning the BC continuum ([1] prechemotherapy [n = 146], [2] immediately postchemotherapy [n = 48], [3] survivorship [n = 138], and [4] metastatic [n = 47]) were pooled and compared with women at high-risk of BC (BC risk; n = 64). Identical treadmill-based peak cardiopulmonary exercise testing protocols evaluated exercise intolerance (peak oxygen consumption; V̇O2peak) and other resting, submaximal, and peak cardiopulmonary responses. The prevalence of 12 abnormal exercise responses was evaluated. Graphical plots of exercise responses were used to identify oxygen delivery and/or uptake mechanisms contributing to exercise intolerance. Unsupervised, hierarchical cluster analysis was conducted to explore exercise response phenogroups.

RESULTS

Mean V̇O2peak was 2.78 ml O2.kg-1·min-1 (95% confidence interval [CI], -3.94, -1.62 mL O2.kg-1·min-1; P < 0.001) lower in the pooled BC cohort (52 ± 11 yr) than BC risk (55 ± 10 yr). Compared with BC risk, the pooled BC cohort had a 2.5-fold increased risk of any abnormal cardiopulmonary response (odds ratio, 2.5; 95% confidence interval, 1.2, 5.3; P = 0.014). Distinct exercise responses in BC reflected impaired oxygen delivery and uptake relative to control, although considerable inter-individual heterogeneity within cohorts was observed. In unsupervised, hierarchical cluster analysis, six phenogroups were identified with marked differences in cardiopulmonary response patterns and unique clinical characteristics.

CONCLUSIONS

Abnormal cardiopulmonary response to exercise is common in BC and is related to impairments in oxygen delivery and uptake. The identification of exercise response phenogroups could help improve cardiovascular risk stratification and guide investigation of targeted exercise interventions.

Cluster analysis of multiple impairment measures in evidence-based classification for para-alpine sit skiers

The International Paralympic Committee has been promoting the development of evidence-based classification to reduce the subjectivity in current decision-making systems. The current study aimed to evaluate the validity of the impairment and performance tests for para-alpine sit skiing classification, and whether cluster analysis of the measures would produce a valid classification structure. Thirty-eight para-alpine sit skiers with different disabilities completed seven tests. During these tests, isometric trunk strength, trunk muscle excitation, trunk range of movement (ROM), and simulated skiing performance (board tilt angle) were assessed. Correlations between the measures and the board tilt angle were calculated. To group athletes, K-means cluster analysis was performed according to how much the impairment measures affected the board tilting. There were significant correlations between all measures and the maximal board tilt angle (r = 0.35-0.81, p < 0.05). The cluster analysis revealed that the introduction of ROM and muscle excitation was an effective supplement to strength measures in improving the classification accuracy (53%-79%). It produced four clusters with strong structures (mean silhouette coefficient = 0.81) and large and significant inter-cluster differences in most measures and performance between clusters (p < 0.05). The cluster analysis produced classes comprising athletes with similar degrees of activity limitation. All tests reported can help establish a more transparent classification system for para-alpine sit skiers. This study also provides a reference for evidence-based classification systems in other Para sports.

The effects of trunk kinematics and EMG activity of wheelchair racing T54 athletes on wheelchair propulsion speeds

Background

The purpose of this study is to examine the impact of trunk kinematic characteristics and trunk muscle electromyography (EMG) activity on propulsion speeds in wheelchair racing T54 athletes.

Method

The Vicon infrared high-speed 3D motion capture system was utilized to acquire kinematic data of the shoulders, elbows, wrists, and trunk from twelve T54 athletes at four different speeds (5.55 m/s, 6.94 m/s, 8.33 m/s, and personal maximum speed). Additionally, the Trigno Wireless EMG system was employed to collect synchronous surface electromyography (EMG) data from the rectus abdominis and erector spinae muscles. The kinematics and EMG data of the trunk were compared across various wheelchair propulsion speeds while also examining the correlation coefficient between wheelchair propulsion speeds and: (1) the range of motion of upper limb joints as well as the trunk; (2) the maximum angular velocities of the upper limbs joints as well as the trunk; and (3) rectus abdominis and erector spinae EMG activity. Two multiple linear stepwise regression models were utilized to examine the impact of variables that had been identified as significant through correlation coefficient tests (1) and (2) on propulsion speed, respectively.

Results

There were significant differences in the range of motion (p<0.01) and angular velocity (p<0.01) of the athlete's trunk between different propulsion speeds. The range of motion (p<0.01, r = 0.725) and angular speed (p<0.01, r = 0.882) of the trunk showed a stronger correlation with propulsion speed than did upper limb joint movements. The multiple linear stepwise regression model revealed that the standardized β values of trunk motion range and angular velocity in athletes were greater than those of other independent variables in both models. In terms of the EMG variables, four of six variables from the rectus abdominis showed differences at different speeds (p<0.01), one of six variables from the erector spinae showed differences at different speeds (p<0.01). All six variables derived from the rectus abdominis exhibited a significant correlation with propulsion speed (p<0.05, r>0.3), while one variable derived from the erector spinae was found to be significantly correlated with propulsion speed (p<0.01, r = 0.551).

Conclusion

The movement of the trunk plays a pivotal role in determining the propulsion speed of wheelchair racing T54 athletes. Athletes are advised to utilize trunk movements to enhance their wheelchair's propulsion speed while also being mindful of the potential negative impact on sports performance resulting from excessive trunk elevation. The findings of this study indicate that it would be beneficial for wheelchair racing T54 athletes to incorporate trunk strength training into their overall strength training regimen, with a specific emphasis on enhancing the flexion and extension muscles of the trunk.

Isometric, Dynamic, and Manual Muscle Strength Measures and Their Association With Cycling Performance in Elite Paracyclists

OBJECTIVE

Paracycling classification aims to generate fair competition by discriminating between levels of activity limitation. This study investigated the relationship between lower limb manual muscle tests (MMT) with ratio-scaled measures of isometric and dynamic strength and of the ratio-scaled measures with cycling performance.

DESIGN

Fifty-six para cyclists (44 males, 12 females) with leg impairments performed isometric and dynamic strength tests: leg push and pull, and an all-out 20-sec sprint. The MMT results were obtained from the classification database ( n = 21) and race speeds from time trials ( n = 54).

RESULTS

Regression analyses showed significant associations of MMT with isometric push ( R2 = 0.49), dynamic push ( R2 = 0.35), and dynamic pull ( R2 = 0.28). Isometric strength was significantly correlated with dynamic push (ρ = 0.63) and pull (ρ = 0.54). The isometric and dynamic tests were significantly associated with sprint power and race speed ( R2 = 0.16-0.50).

CONCLUSIONS

The modified MMT and ratio-scaled measures were significantly associated. The significant relation of isometric and dynamic strength with sprint power and race speed maps the impact of lower limb impairments on paracycling performance. The MMT and the isometric and dynamic measures show potential for use in paracycling classification.

Stakeholders' Consensus to Guide the Minimum Impairment Criteria in Wheelchair Basketball

The International Paralympic Committee athlete classification code mandates sports to have defined minimum impairment criteria, describing the minimum level of an eligible impairment an athlete must possess, to be able to participate in that sport. The aim of this study was to establish stakeholders' consensus for the minimum impairment criteria in wheelchair basketball. From a pool of 48 expert stakeholders (identified via an international medical and scientific working group), 39 completed a 4-round Delphi survey. Questions were answered on the method of assessing each eligible impairment, and the level of impairment that should constitute the minimum impairment criteria. This study indicated where stakeholder consensus existed and noted that consensus was developed for impaired muscle power, impaired passive range of motion, leg length difference, hypertonia and ataxia. No consensus was found for limb deficiency and athetosis. Participants raised concerns with using subjective measurement scales for assessing certain impairments, whilst also calling for more quantitative research to be conducted into the level of impairment that should constitute the minimum impairment criteria. For these research findings to form practical minimum impairment criteria that are part of a wheelchair basketball classification system, it is required to examine their feasibility by conducting further research.

Assessment of muscle strength in para-athletes: A systematic review of observational studies

Accurate and reliable evaluation of muscle strength in para-athletes is essential for monitoring the effectiveness of strength training and/or rehabilitation programmes, and sport classification. Our aim is to synthesise evidence related to assessing muscle strength in para-athletes. Four databases were searched from January 1990 to July 2021 for observational studies focusing on strength assessment. Independent screening, data extraction, and quality assessment were performed in duplicate. A total of 1764 potential studies were identified. Thirty met the inclusion criteria and were included in the review. The mean age of participants was 30.7 years (standard deviation [SD]: 2.4). The majority were men (88%) participating in wheelchair sports, including basketball, rugby, and tennis (23/30: 76%). Overall quality varied, with more than half of the studies failing to identify strategies for dealing with confounding variables. Despite manual muscle testing being a standard component of para-sport classification systems, evidence examining strength characteristics in para-athletes is derived primarily from isometric and isokinetic testing. In studies that included comparative strength data, findings were mixed. Some studies found strength values were similar to or lower than able-bodied athletic controls. However, an important observation was that others reported higher shoulder strength in para-athletes taking part in wheelchair sports than both able-bodied and disabled non-athletes. Studies need to develop accessible, standardised strength testing methods that account for training influence and establish normative strength values in para-athletes. There is also a need for additional studies that include female para-athletes and para-athletes with greater functional impairments.

The Influence of Trunk Impairment Level on the Kinematic Characteristics of Alpine Sit-Skiing: A Case Study of Paralympic Medalists

This study aimed to examine the relationship between the trunk impairment level and the trunk kinematic characteristics during alpine sit-skiing from a classification perspective. Three Paralympic medalists in sitting classes (LW10-2, LW11, and LW12-2) participated in the present study. To simulate the racing conditions, giant slalom gates were set. To measure the kinematics of the skier and sit-ski during skiing, a motion capture method with inertial measurement units was used. The muscle activities of the trunk muscles were evaluated using electromyography. Chest lateral flexion, chest flexion, and hip flexion/extension angle during sit-skiing were reduced due to impairment. Additionally, the insufficient lateral flexion (angulation) caused a decrease in edging angle, and that the insufficient chest and hip flexion/extension caused a lower loading in the latter half of the turn through smaller vertical movement. Since edging angle and loading are key factors in ski control, the three joint motions could be measures of sport-specific activity limitation in sit-skiing classification. Between the LW10-2 and LW11 skiers, no distinct differences in trunk kinematics were found. Assuming the scaling factor of race time as a measure of skiing performance, one possible reason is that the difference in skiing performance the LW10-2 and LW11 skiers is considerably smaller relative to differences between the LW11 and LW12-2 skiers. There were no distinct differences among classes in the results of muscle activity, and therefore, this information appears to play a minimal role for classification.

Association between upper-limb isometric strength and handcycling performance in elite athletes

This study investigated the association among isometric upper-limb strength of handcyclists and sport-specific performance outcomes. At two international events, 62 athletes were tested on upper-limb strength, measured with an isometric-strength setup and with Manual Muscle Test (MMT). Horizontal force (Fz), effectiveness, rate of development, variability, and asymmetries were calculated for upper-limb pull and push. Performance measures were mean (POmean) and peak (POpeak) 20-s sprint power output and average time-trial velocity (TTvelocity). Regression models were conducted to investigate which pull and push strength variables associated strongest with performance measures. Additional regression analyses were conducted with an MMT sum score as predictor. Push and pull Fz showed the strongest associations with all outcomes. Combined push and pull Fz explained (p < .001) 80-81% of variance of POmean and POpeak. For TTvelocity, only push Fz was included in the model explaining 29% of the variance (p < .001). MMT models revealed weaker associations with sprint PO (R2 = .38-.40, p < .001) and TTvelocity (R2 = .18, p = 0.001). The findings confirmed the relevance of upper-limb strength on handcycling performance and the significance of ratio-scaled strength measures. Isometric strength outcomes are adequate sport-specific indicators of impairment in handcycling classification, but future research should corroborate this notion and its potential to discriminate between sports classes.

Development of evidence-based classification for para surfers with physical impairments: A narrative review

The sport of competitive para surfing is growing internationally without established classification procedures. A classification structure is essential for equitable sport competition and worldwide sport progression. This narrative review summarizes the existing knowledge on Paralympic classification and surfing biomechanics. Its primary purpose is to describe the development of an evidence-based para surfing classification structure that follows the International Paralympic Committee (IPC) Classification Code. Two databases-PubMed and Google Scholar-were searched for three themes: "Paralympic classification", "performance determining factors in surfing", and "impact of impairments on surf performance". The IPC Classification Code and IPC regulations were obtained from the IPC website and official publications. Seventy-six relevant articles were utilized to guide the design of this preliminary para surfing classification structure. A conceptual framework on athlete characteristics, adaptive sport characteristics, and use of equipment is presented to build the essential knowledge base for continual growth of para surfing. This classification structure and conceptual framework will support para surfing sport expansion and help pave the way for its inclusion in the Paralympic Games.

Establishing the reliability of instrumented trunk impairment assessment methods to enable evidence-based classification in Para swimming

This study examined the reliability of instrumented trunk assessment methods across two experiments to develop and improve evidence-based classification in Para swimming. Trunk coordination, range of motion (ROM), and strength were assessed in 38 non-disabled participants. Each test battery was completed on two occasions to determine inter-session reliability. Intra-session reliability was also determined in Experiment Two. Absolute agreement of two-way mixed intraclass correlation coefficients (ICC 3,1) was calculated to assess reliability. Standard errors of measurement (SEMs) were also reported to facilitate comparisons between different outcomes. Trunk coordination measures had low-to-moderate reliability (inter-session ICCs = 0.00-0.60; intra-session ICCs = 0.14-0.65) and variable SEMs (5-60%). Trunk ROM demonstrated moderate-to-excellent reliability (inter-session ICCs = 0.61-0.93; intra-session ICCs = 0.87-0.95) and good SEMs (<10%). Trunk strength measures demonstrated good-to-excellent reliability (ICCs = 0.87-0.98) and good SEMs (<10%). The strength values obtained for the load cell and hand-held dynamometer (HHD) were significantly different from each other with the HHD underestimating strength. Modifications provided in Experiment Two improved the reliability of strength and ROM assessments but did not improve coordination measures. Further research involving para swimmers is required to establish the validity of the methods.

The physical characteristics underpinning performance of wheelchair fencing athletes: A Delphi study of Paralympic coaches

Wheelchair fencing (WF) is a Paralympic sport which is practised by athletes with physical disabilities and is classified into three categories according to the degree of activity limitation the impairment causes in the sport. All Paralympic sports are requested to develop their own evidence-based classification system to enhance the confidence in the classification process; however, this is yet to be achieved in WF. Research within WF is scarce; therefore, the aim of this study was to reach expert consensus on the physical characteristics that underpin performance of athletes competing in the sport as this is known as one of the initial steps required to achieve an evidence-based classification system. Sixteen Paralympic WF coaches were invited to take part in a three-round Delphi study, with experts drawing consensus on qualities of speed, strength, power, flexibility and motor control of the trunk and fencing arm being associated with increased athletic success. The required qualities of the non-fencing arm led to diverging opinions across the expert panel. This study provides clear guidance of the physical qualities to be developed to maximize athletic performance while also providing the initial framework to guide future WF classification research.

Evaluating objective measures of impairment to trunk strength and control for cross-country sit skiing

In Paralympic cross-country sit skiing, athlete classification is performed by an expert panel, so it may be affected by subjectivity. An evidence-based classification is required, in which objective measures of impairment must be identified. The purposes of this study were: (i) to evaluate the reliability of 5 trunk strength measures and 18 trunk control measures developed for the purposes of classification; (ii) to rank the objective measures, according to the largest effects on performance. Using a new testing device, 14 elite sit-skiers performed two upright seated press tests and one simulated poling test to evaluate trunk strength. They were also subjected to unpredictable balance perturbations to measure trunk control. Tests were repeated on two separate days and test–retest reliability of trunk strength and trunk control measures was evaluated. A cluster analysis was run and correlation was evaluated, including all strength and control measures, to identify the measures that contributed most to clustering participants. Intraclass correlations coefficients (ICC) were 0.71 < ICC < 0.98 and 0.83 < ICC < 0.99 for upright seated press and perturbations, respectively. Cluster analysis identified three clusters with relevance for strength and balance control measures. For strength, in upright seated press peak anterior pushing force without backrest (effect size = 0.77) and ratio of peak anterior pushing force without and with backrest (effect size = 0.72) were significant. For balance control measures, trunk range of motion in forward (effect size = 0.81) and backward (effect size = 0.75) perturbations also contributed. High correlations (− 0.76 < r < − 0.53) were found between strength and control measures. The new testing device, protocol, and the cluster analysis show promising results in assessing impairment of trunk strength and control to empower an evidence-based classification.

Conceptual model of sport-specific classification for para-athletes with intellectual impairment

The present paper describes the conceptual basis of evidence-based classification of para-athletes with intellectual impairment (II). An extensive description of the theoretical and conceptual foundation of the system as currently conceived is provided, as are examples of its applications in the three sports included in the Paralympic programme for II-athletes in 2020 (i.e., athletics, swimming and table tennis). Evidence-based classification for II-athletes is driven by two central questions: i. How can intellectual impairment be substantiated in a valid and reliable way, and ii. Does intellectual impairment limit optimal sport proficiency? Evolution of the system and current best practice for addressing these questions are described, and suggestions for future research and development are provided. Challenges of understanding and assessing a complex (multifaceted and intersectional) impairment in the context of sport also are considered.

Optimising classification of proximal arm strength impairment in wheelchair rugby: A proof of concept study

This study examined the relationship between proximal arm strength and mobility performance in wheelchair rugby (WR) athletes and examined whether a valid structure for classifying proximal arm strength impairment could be determined. Fifty-seven trained WR athletes with strength impaired arms and no trunk function performed six upper body isometric strength tests and three 10 m sprints in their rugby wheelchair. All strength measures correlated with 2 m and 10 m sprint times (r ≥ -0.43; p ≤ 0.0005) and were entered into k-means cluster analyses with 4-clusters (to mirror the current International Wheelchair Rugby Federation [IWRF] system) and 3-clusters. The 3-cluster structure provided a more valid structure than both the 4-cluster and existing IWRF system, as evidenced by clearer differences in strength (Effect sizes [ES] ≥ 1.0) and performance (ES ≥ 1.1) between adjacent clusters and stronger mean silhouette coefficient (0.64). Subsequently, the 3-cluster structure for classifying proximal arm strength impairment would result in less overlap between athletes from adjacent classes and reduce the likelihood of athletes being disadvantaged due to their impairment. This study demonstrated that the current battery of isometric strength tests and cluster analyses could facilitate the evidence-based development of classifying proximal arm strength impairment in WR.

Cluster analysis of impairment measures to inform an evidence-based classification structure in RaceRunning, a new World Para Athletics event for athletes with hypertonia, ataxia or athetosis

RaceRunning enables athletes with limited or no walking ability to propel themselves independently using a three-wheeled frame that has a saddle, handle bars and a chest plate. For RaceRunning to be included as a para athletics event, an evidence-based classification system is required. This study assessed the impact of trunk control and lower limb impairment measures on RaceRunning performance and evaluated whether cluster analysis of these impairment measures produces a valid classification structure for RaceRunning. The Trunk Control Measurement Scale (TCMS), Selective Control Assessment of the Lower Extremity (SCALE), the Australian Spasticity Assessment Scale (ASAS), and knee extension were recorded for 26 RaceRunning athletes. Thirteen male and 13 female athletes aged 24 (SD = 7) years participated. All impairment measures were significantly correlated with performance (rho = 0.55-0.74). Using ASAS, SCALE, TCMS and knee extension as cluster variables in a two-step cluster analysis resulted in two clusters of athletes. Race speed and the impairment measures were significantly different between the clusters (p < 0.001). The findings of this study provide evidence for the utility of the selected impairment measures in an evidence-based classification system for RaceRunning athletes.

Measures of impairment applicable to the classification of Paralympic athletes competing in wheelchair sports: A systematic review of validity, reliability and associations with performance

A fundamental aspect of classification systems in Paralympic sport is having valid and reliable measures of impairment. However, minimal consensus exists for assessing impaired strength, coordination and range of motion. This review aimed to systematically identify measures of upper body strength, coordination and range of motion impairments that meet the requirements for use in evidence-based classification systems in wheelchair sports. Three electronic databases were searched from 2003 until 31 August 2019 for studies that assessed upper body function of participants and used a measurement tool that assessed strength, coordination or range of motion. The body of evidence for each identified measure was appraised using the Grading of Recommendations Assessment, Development and Evaluation framework. Twenty-three studies were included: ten measured strength and coordination, and six measured range of motion. There was "moderate" confidence in using isometric strength for assessing strength impairment. Tapping tasks for the assessment of coordination impairment received a "low" confidence rating. All other identified measures of coordination and range of motion impairment received a "very low" confidence rating. Several potential measures were identified for assessing upper body strength, coordination and range of motion impairments. Further research is warranted to investigate their use for classification in Paralympic wheelchair sports.

Inter-Rater Reliability, Concurrent Validity and Sensitivity of Current Methods to Assess Trunk Function in Boccia Player with Cerebral Palsy

Trunk function is a core factor to allocate Boccia players with cerebral palsy in BC1 and BC2 sport classes, according to the Boccia International Sports Federation (BISFed). However, the appropriateness of the current test to assess trunk function has never been studied to determine its reliability, validity and sensitivity to discriminate between different levels of impairment. Thirty-six players (BC1 = 13 and BC2 = 23) took part in this study. Trunk control was assessed through the BISFed trunk function scale (TFS) and a posturographic test battery consisting of two static and three dynamic tasks. The inter-rater reliability for the BISFed TFS was set at 94.44% of agreement. Moderate-to-high correlations were obtained between posturographic tasks (0.39 < r < 0.96; p < 0.05–0.01), while the BISFed TFS only correlated with two of the dynamic tasks and the overall dynamic score (−0.38 < r < −0.51; p < 0.05). The BISFed TFS was not able to discriminate between sport classes, whereas the static posturographic task did so (p = 0.004). Even though the current BISFed TFS presented good inter-rater reliability, it does not seem to have enough sensitivity to discriminate between BC1 and BC2. Although the static posturographic tasks were able to discriminate between sports classes, it seems necessary to develop new field tests assessing participants’ trunk stabilization abilities.

Evaluation of the bilateral function in para-athletes with spastic hemiplegia: A model-based clustering approach

OBJECTIVES

Spastic hemiplegia is one of the most common forms of cerebral palsy, in which one side of the body is affected to a greater extent than the other one. Hemiplegia severity (i.e. moderate vs mild forms) is currently used in some Para sports for classification purposes. This study evaluates the sensitivity of several tests of stability (e.g. one-legged stance test), dynamic balance (side-step test), coordination (rapid heel-toe placements), range of movement (backward stepping lunge), and lower limb power (the triple hop distance and the isometric peak force of the knee extensors) to discriminate between the impaired and unimpaired lower extremities' function in para-athletes with spastic hemiplegia.

METHODS

A sample of 87 international para-athletes with cerebral palsy took part in the study, and their bilateral performance was measured for the abovementioned tests. The tests' sensitivity to discriminate between impaired vs unimpaired legs was assessed using Boruta's method.

RESULTS

The triple hop distance, the magnitude of the mean velocity in the one-legged stance test and the time to perform the rapid heel-toe placement test are the most sensitive variables when performing random forest classifiers. In addition, the study confirms two optimal clusters by Gaussian finite mixture models to represent the athletes' performance.

CONCLUSIONS

Reference scores for the clusters are provided, demonstrating that coordination, balance, and power of the lower limbs are relevant variables for classifying para-athletes with spastic hemiplegia.

An Industrial Micro-Defect Diagnosis System via Intelligent Segmentation Region

In the field of machine vision defect detection for a micro workpiece, it is very important to make the neural network realize the integrity of the mask in analyte segmentation regions. In the process of the recognition of small workpieces, fatal defects are always contained in borderline areas that are difficult to demarcate. The non-maximum suppression (NMS) of intersection over union (IOU) will lose crucial texture information especially in the clutter and occlusion detection areas. In this paper, simple linear iterative clustering (SLIC) is used to augment the mask as well as calibrate the score of the mask. We propose an SLIC head of object instance segmentation in proposal regions (Mask R-CNN) containing a network block to learn the quality of the predict masks. It is found that parallel K-means in the limited region mechanism in the SLIC head improved the confidence of the mask score, in the context of our workpiece. A continuous fine-tune mechanism was utilized to continuously improve the model robustness in a large-scale production line. We established a detection system, which included an optical fiber locator, telecentric lens system, matrix stereoscopic light, a rotating platform, and a neural network with an SLIC head. The accuracy of defect detection is effectively improved for micro workpieces with clutter and borderline areas.