Measuring the prediction difficulty of individual cases in a dataset using machine learning.

Different levels of prediction difficulty are one of the key factors that researchers encounter when applying machine learning to data. Although previous studies have introduced various metrics for assessing the prediction difficulty of individual cases, these metrics require specific dataset preconditions. In this paper, we propose three novel metrics for measuring the prediction difficulty of individual cases using fully-connected feedforward neural networks. The first metric is based on the complexity of the neural network needed to make a correct prediction. The second metric employs a pair of neural networks: one makes a prediction for a given case, and the other predicts whether the prediction made by the first model is likely to be correct. The third metric assesses the variability of the neural network's predictions. We investigated these metrics using a variety of datasets, visualized their values, and compared them to fifteen existing metrics from the literature. The results demonstrate that the proposed case difficulty metrics were better able to differentiate various levels of difficulty than most of the existing metrics and show constant effectiveness across diverse datasets. We expect our metrics will provide researchers with a new perspective on understanding their datasets and applying machine learning in various fields.

Anatomical study of the adductor canal: three-dimensional micro-computed tomography, histological, and immunofluorescence findings relevant to neural blockade.

BACKGROUND: A precise anatomical understanding of the adductor canal (AC) and its neural components is essential for discerning the action mechanism of the AC block. We therefore aimed to clarify the detailed anatomy of the AC using micro-computed tomography (micro-CT), histological evaluation, and immunofluorescence (IF) assays. METHODS: Gross dissections of 39 thighs provided morphometric data relevant to injection landmarks. Serial sectional images of the AC were defined using micro-CT and ultrasonography. The fascial and neural structures of the AC proper were histologically evaluated using Masson's trichrome and Verhoeff-Van Gieson staining, and double IF staining using choline acetyltransferase (ChAT) and neurofilament 200 antibodies. RESULTS: The posteromedial branch insertion of the nerve to vastus medialis (NVM) into the lateral border of the AC proper was lower (14.5 ± 2.4 cm [mean ± SD] above the base of the patella) than the origin of the proximal AC. The AC consists of a thin subsartorial fascia in the proximal region and a thick aponeurosis-like vastoadductor membrane in the distal region. In the proximal AC, the posteromedial branch of the NVM (pmNVM) consistently contained both sensory and motor fibers, and more ChAT-positive fibers were observed than in the saphenous nerve (27.5 ± 11.2 / 104 vs. 4.2 ± 2.6 / 104 [counts/µm2], P < 0.001). CONCLUSIONS: Anatomical differences in fascial structures between the proximal and distal AC and a mixed neural component of the neighboring pmNVM have been visualized using micro-CT images, histological evaluation, and IF assays.

Retrodural space of Okada in the posterior ligamentous complex region: clinical and anatomical findings relevant to lumbar interlaminar epidural injection.

BACKGROUND: The retrodural space of Okada is a potential space posterior to the ligamentum flavum that allows communication with the bilateral facet joints. However, the actual anatomy of this space has not been clearly visualized to date. We sought to investigate the characteristics of patients showing contrast spreading to the facet joint space during epidural injection and to clarify the anatomical structures of the retrodural space and adjacent ligamentous tissues in cadaveric specimens. METHODS: Fluoroscopic images of patients who underwent fluoroscopy-guided lumbar interlaminar epidural injection were assessed for contrast flow to the facet joints. Patient demographics, preprocedural imaging study findings, and epidural approaches were analyzed. The anatomical study included the sectional dissection, micro-CT imaging, and histological evaluation of lumbar spine specimens from 16 embalmed cadavers. RESULTS: Fluoroscopic images of 605 epidural injections were analyzed. Among them, 36 with inadvertent spread into the facet joints (5.9%) were identified. Multivariate analysis revealed that facet joint pathologies were significantly associated with inadvertent spread into the facet joints (OR 4.382; 95% CI 1.160 to 16.558; p=0.029). Micro-CT and histological findings consistently showed a retrodural space between the ligamentum flavum and interspinous ligament. Various anatomical communication routes in the posterior ligamentous complex leading to this space were observed in specimens with degenerative and pathological changes. CONCLUSION: Degenerative and pathological facet joint changes were associated with a higher incidence of spread into the retrodural space during epidural injection. Our findings confirm anatomical evidence for a false loss of resistance before the needle enters the epidural space.

Effect of a Cognitive Function and Social Skills-Based Digital Exercise Therapy Using IoT on Motor Coordination in Children with Intellectual and Developmental Disability.

This study aimed to determine the effects of a virtual reality exercise program based on cognitive function and social skills on motor coordination in children with intellectual and developmental disabilities (IDD). Thirty-five children with IDD were randomly assigned to either the cognitive function and social skills-based virtual reality exercise system (CS-VR) group or the conventional virtual reality exercise system (C-VR) group. Before and after the intervention, each participant was tested for motor coordination (extended horizontal jump, hop, stationary dribble, overarm throw) and exercise performance (standing long jump, YMCA step test). Compared with the C-VR group, the CS-VR group showed significant improvements in motor coordination in terms of extended horizontal jump, hop, and overarm throw (p < 0.01, p < 0.05, and p < 0.01, respectively). In addition, compared with the C-VR group, the CS-VR group showed a significant increase in standing long jump (p < 0.01), although no significant between-group variation was found in stationary dribble and recovery heart rate (RHR) as part of the YMCA step test (p > 0.05, and p > 0.05, respectively). These results suggest that for the development of motor skills in children with IDD, it is essential to develop an exercise program that reflects the levels of cognitive functions and social skills of these children.

An anatomical and radiological study of the tectorial membrane and its clinical implications.

The radiological image of an intact tectorial membrane (TM) became an important favorable prognostic factor for craniovertebral instability. This study visualized the fascial layers of the TM and adjacent connective tissues with clinical significance by micro-CT and histological analysis. The TM firmly attached to the bony surface of the clivus, traversed the atlantoaxial joint posteriorly, and was inserted to the body of the axis showing wide distribution on the craniovertebral junction. The supradental space between the clivus, dens of the axis, anterior atlantooccipital membrane, and the TM contained profound venous networks within the adipose tissues. At the body of the axis, the compact TM layer is gradually divided into multiple layers and the deeper TM layers reached the axis while the superficial layer continued to the posterior longitudinal ligament of the lower vertebrae. The consistent presence of the fat pad and venous plexus in the supradental space and firm stabilization of the TM on the craniovertebral junction was demonstrated by high-resolution radiologic images and histological analysis. The evaluation of the TM integrity is a promising diagnostic factor for traumatic craniovertebral dislocation.

Development of an ICT-Based Exergame Program for Children with Developmental Disabilities.

The purpose of this study was to develop an information and communications technology (ICT)-based exergame for children with developmental disabilities (DD) and to examine its impacts on physical fitness and fundamental motor skills (FMS). The ICT-based exergame consisted of visual and auditory demonstrations of diverse locomotor movement and object manipulation activities by the virtual characters. A total of 52 children with DD participated in the present study. The participants were divided into twenty-seven children in the experimental group and 25 children with DD in the control group. The experimental group participants engaged in the exergame program for 12 weeks. All participants' muscle strength (i.e., standing long jump) and four fundamental motor skills, such as the horizontal jump, hop, overhand throw, and dribble, were assessed during the pre- and post-test process. There were significant impacts on physical fitness and FMS (p < 0.001) between the groups. Specifically, the results of three FMS (hop, overhand throw, and dribble) and standing long jump significantly improved in children with DD except for the horizontal jump skill. The results of this study is evidence that the ICT-based exergame program for children with DD may be utilized to improve physical fitness and FMS in children with DD.

The Impact of a Rater Training Program on the TGMD-3 Scoring Accuracy of Pre-Service Adapted Physical Education Teachers.

This study aimed to examine the impact of a TGMD-3 rater training program on pre-service adapted physical education (APE) teachers' ability to score fundamental motor skills for children with developmental disabilities (DD). The training program consisted of a description of the TGMD-3 instrument and DD, as well as content focused on skill performance and correct scoring through systematic analysis of 13 FMS via the instrument. Three experts established the standard score through the TGMD-3 performance evaluation of a child with DD. Thirty-nine pre-service APE teachers in the experimental group and 30 pre-service teachers in the control group completed the pre- and post-test process in this study. There were significant impacts on the pre-service teachers' ability to score the locomotor, ball skill subtests, and total skill scores (p < 0.05) compared to the control group. Specifically, the results of the three locomotor skills (running, horizontal jump, & slide) and three ball skills (two-hand strike, dribble, & kick) significantly improved in scoring accuracy. The results of this study can evidence that a TGMD-3 rater training program for pre-service APE teachers enhances the scoring accuracy of FMS among children with DD.

The pathway of injectate spread during thoracic intertransverse process (ITP) block: Micro-computed tomography findings and anatomical evaluations.

STUDY OBJECTIVE: To investigate the mechanism of action of the thoracic intertransverse process (ITP) block. DESIGN: Three-dimensional micro-computed tomography (3D micro-CT) study and cadaveric evaluation. SETTING: A translational research unit for anatomy and analgesia in a university hospital. PATIENTS: Twelve embalmed and three non-embalmed human cadavers were used in this study. MEASUREMENTS: Micro-CT images of the mid-thoracic paravertebral space and its adjacent ligamentous tissues were acquired and 3D images were reconstructed. Manual dissection and histologic examination of these structures complemented the images. To confirm our findings, the dye-spreading pattern after ultrasound-guided ITP injection of 20 mL dye solution at T4-T5 was evaluated. MAIN RESULTS: Micro-CT and histologic findings showed that the costotransverse foramen (the medial slit of the superior costotransverse ligament) and the costotransverse space (between the rib and the transverse process) were potential pathways to the thoracic paravertebral space during ITP block. Single-level ITP injection with a dye solution resulted in a multilevel segmental paravertebral spread in cadaveric evaluation. CONCLUSIONS: The space posterior to the superior costotransverse ligament, the target area for ITP blocks, has potential anatomical pathways to the thoracic paravertebral space. The costotransverse foramen and the costotransverse space provided the anatomical conduit for the anterior and intersegmental paravertebral spread of the ITP block.

Micro-computed tomography with contrast enhancement: An excellent technique for soft tissue examination in humans.

Manual dissection and histologic examination are commonly used to investigate human structures, but there are limitations in the damage caused to delicate structures or the provision of limited information. Micro-computed tomography (microCT) enables a three-dimensional volume-rendered observation of the sample without destruction and deformation, but it can only visualize hard tissues in general. Therefore, contrast-enhancing agents are needed to help in visualizing soft tissue. This study aimed to introduce microCT with phosphotungstic acid preparation (PTA-microCT) by applying the method to different types of human tissue. Specimens from human cadavers were used to examine the orbicularis retaining ligament (ORL), nasolabial fold (NLF), and the calcaneal tunnel of the sole. Using PTA-microCT, relevant information of human structures was identified. In the ORL study, tiny and delicate ligamentous fibers were visualized in detail with multidirectional continuity. In the NLF study, complex structural formation consisting of various types of soft tissue were investigated comprehensively. In the calcaneal tunnel study, the space surrounded by diverse features and its inner vulnerable structures were examined without damage. Consequently, we successfully applied the PTA-microCT technique to the analysis of specific human soft tissue structures that are challenging to analyze by conventional methods.

Anatomy of the thoracic paravertebral space: 3D micro-CT findings and their clinical implications for nerve blockade.

BACKGROUND: A precise anatomical understanding of the thoracic paravertebral space (TPVS) is essential to understanding how an injection outside this space can result in paravertebral spread. Therefore, we aimed to clarify the three-dimensional (3D) structures of the TPVS and adjacent tissues using micro-CT, and investigate the potential routes for nerve blockade in this area. METHODS: Eleven embalmed cadavers were used in this study. Micro-CT images of the TPVS were acquired after phosphotungstic acid preparation at the mid-thoracic region. The TPVS was examined meticulously based on its 3D topography. RESULTS: Micro-CT images clearly showed the serial topography of the TPVS and its adjacent spaces. First, the TPVS was a very narrow space with the posterior intercostal vessels very close to the pleura. Second, the superior costotransverse ligament (SCTL) incompletely formed the posterior wall of the TPVS between the internal intercostal membrane and vertebral body. Third, the retro-SCTL space broadly communicated with the TPVS via slits, costotransverse space, intervertebral foramen, and erector spinae compartment. Fourth, the costotransverse space was intersegmentally connected to the adjacent retro-SCTL space. CONCLUSIONS: A non-destructive, multi-sectional approach using 3D micro-CT more comprehensively demonstrated the real topography of the intricate TPVS than previous cadaver studies. The posterior boundary and connectivity of the TPVS provides an anatomical rationale for the notion that paravertebral spread can be achieved with an injection outside this space.

Analysis of the positional relationship of the long thoracic nerve considering clinical treatment.

INTRODUCTION: The long thoracic nerve (LTN) has a risk of being damaged during chest surgery and should be considered when performing anesthesia such as a serratus anterior plane block (SAPB). We analyzed the relationship between landmarks-the fourth to ninth intercostal space (ICS) at the midaxillary line (MAL), through which the distal part of the LTN passes-and the LTN. MATERIALS AND METHODS: We used 25 specimens from 17 embalmed Korean cadavers. The MAL, level of rib and ICS, and regions 5 cm anteroposterior to the MAL (aMAL/pMAL) were established to measure the position of the LTN crossing the MAL, pathway of the LTN, and entering points of the LTN to the SA. RESULTS: The LTN crossed the MAL in 76% of the specimens. The LTN crossed the MAL within the fifth to sixth rib level in 70.8%. Seventy-six percent of the branches entered the SA within the fourth to sixth ICS. The fifth rib and ICS were the most frequent regions aMAL; however, several branches were found pMAL. The LTN entered the SA in 92.6% of the specimens within 3 cm anterior and 1 cm posterior to the MAL. CONCLUSIONS: We set the danger zone as 4 cm near the MAL at the fourth to sixth ICS for thoracotomy. In addition, we proposed the fifth ICS in aMAL at the superficial plane as the alternative injection point for SAPB when blocking the LTN, and the fifth ICS in pMAL at the deep plane to prevent blocking the LTN.

Deep-Asymmetry: Asymmetry Matrix Image for Deep Learning Method in Pre-Screening Depression.

To have an objective depression diagnosis, numerous studies based on machine learning and deep learning using electroencephalogram (EEG) have been conducted. Most studies depend on one-dimensional raw data and required fine feature extraction. To solve this problem, in the EEG visualization research field, short-time Fourier transform (STFT), wavelet, and coherence commonly used as method s for transferring EEG data to 2D images. However, we devised a new way from the concept that EEG's asymmetry was considered one of the major biomarkers of depression. This study proposes a deep-asymmetry methodology that converts the EEG's asymmetry feature into a matrix image and uses it as input to a convolutional neural network. The asymmetry matrix image in the alpha band achieved 98.85% accuracy and outperformed most of the methods presented in previous studies. This study indicates that the proposed method can be an effective tool for pre-screening major depressive disorder patients.

Heights and spatial relationships of the facial muscles acting on the nasolabial fold by dissection and three-dimensional microcomputed tomography.

The aim of this study was to clarify the heights and spatial relationships of the facial muscles acting on the nasolabial fold (NLF) by dissection and three-dimensional microcomputed tomography for use in aesthetic treatments. This study used 56 specimens from 34 embalmed adult Korean. A reference line (RF) was set to imitate the NLF after removing the skin, from the superior point of the alar facial crease to the lateral point of the orbicularis oris muscle at the level of the corner of the mouth. The heights and spatial relationships of the facial muscles along the RF could be categorized into five main patterns. The dominant pattern was that the levator labii superioris alaeque nasi muscle (LLSAN), levator labii superioris muscle (LLS), zygomaticus minor muscle (Zmi), and zygomaticus major muscle (Zmj) were on the medial third, medial half, middle third, and lateral third of the RF, respectively. In micro-CT imaging, beneath the skin of the medial half of the NLF, the LLSAN and Zmi fibers inserted into the dermis of the NLF and adjacent to the NLF. Beneath the skin of the middle third of the NLF, the Zmi fibers were found before the muscle inserted into the dermis of the NLF and adjacent to the NLF. Beneath the skin of the lateral third of the NLF, the lateral margin of the orbicularis oris muscle and some Zmj fibers were found at the location of the NLF. The present study utilized dissections and micro-CT to reveal the general pattern and variations of heights and spatial relationships of the facial muscles passing beneath the NLF. These findings will be useful for understanding which muscles affect specific parts of NLFs with various contours, for reducing the NLF in aesthetic treatments, and for reconstructing the NLF in cases of facial paralysis.

Effect of Cyclic Compressive Forces on New Bone Formation during the Distraction Period in Mandibular Distraction Osteogenesis Using a Microactuator-Generated Distractor.

BACKGROUND: The traditional method of distraction required a long period until the new bone was fully fused. This study aimed to evaluate the effect of cyclic and repetitive compressive force on new bone formation by applying a newly designed microactuator-generated distractor compared with the traditional distraction protocol. METHODS: The distraction devices were applied to the right and left mandibles of eight mature beagles, four of which were allocated to the experimental group and the other four to the control group. After a 5-day latency period, in the experimental group, 1.5-mm distraction and 0.5-mm compression were alternately applied every 12 hours; and in the control group, 0.5-mm distraction was repeated every 12 hours to perform a total of 7-mm mandibular lengthening. After an 8-week consolidation period, microstructure analysis using micro-computed tomography and histologic evaluations were performed. RESULTS: Bone volume fraction and trabecular number were significantly higher in the experimental group. Trabecular thickness did not show a significant difference, and trabecular separation was significantly higher in the control group. The measurement of new bone was significantly higher in the experimental group, but the difference in mineral apposition rate between the experimental and control groups was not statistically significant. CONCLUSIONS: This study has revealed that applying cyclic compressive and tensile forces during the distraction period induces better new bone formation than repetitive distraction alone. In addition, the new distraction device produced reliable distraction results during the experiment.

Longitudinal muscular column in the prostatic urethral wall: Its form, shape, and possible function based on mathematical simulation in ejaculation.

BACKGROUND: The shape and function of the longitudinal muscular column (LMC) of the prostate have not been established in detail. The present study was undertaken to elucidate the roles of the LMC of the posterior wall of the prostatic urethra (PSU) in the emission phase of ejaculation by investigating the form and muscular arrangement of the LMC. METHODS: Prostates and urinary bladders were obtained from 14 Korean adult cadavers. Nine specimens were histologically analyzed using hematoxylin and eosin, Masson's trichrome, and Verhoeff-van Gieson staining. Two specimens were scanned using microcomputed tomography (micro-CT), and all scanned images were reconstructed into a three-dimensional model. RESULTS: At the proximal level of the prostate, the ejaculatory ducts (EDs) and prostatic utricle (PU) together were surrounded by circular smooth-muscle fibers. However, at the seminal colliculus (SC) where the EDs and PU opened, they were mainly surrounded by an abundance of longitudinal fibers. The longitudinal fibers posterior to the EDs and PU formed a distinctive LMC in the posterior urethral wall. In histologic sections and micro-CT images, the LMC extended distally from the level of the SC to the level of the membranous urethra (MBU). We simulated a potential mechanism of LMC using a mathematical model of its movements. CONCLUSIONS: Comprehensive analyses based on in-depth assessment of histologic characteristics and micro-CT images demonstrated extension of the LMC from the level of the SC to the level of the MBU, enabling a better understanding of ejaculation physiology involving the LMC. These results suggest that the LMC in the posterior wall of the PSU is a critical component of ejaculation by facilitating the ejection of seminal vesicle fluid into the PSU via well-coordinated contractions.

Intramuscular innervation patterns of the splenius capitis and splenius cervicis and their clinical implications for botulinum toxin injections.

INTRODUCTION: The purpose of this study was to clearly characterize the intramuscular nerve distributions of the splenius capitis and splenius cervicis muscles that are both responsible for the onset of a chronic tension type headache and to use this information to identify the effective botulinum toxin (BoNT) injection sites. MATERIALS AND METHODS: Ten splenius capitis and splenius cervicis specimens were subjected to Sihler's staining to reveal intramuscular nerve arborization patterns and determined the optimal location for BoNT injection. RESULTS: Nerve distribution patterns in the splenius capitis were identified as nerve pathways that travel down toward the origin point and others that travel up toward the insertion point. This neuromuscular innervation from the central (50%) point was distributed evenly in these two directions. The neural pathways of splenius cervicis traveled vertically from the insertion point to the origin point. If the length from the muscle origin point to the insertion point is normalized to 100%, motor neurons innervate the muscle from around the 30% to the 70% point. CONCLUSIONS: The safest and most-effective BoNT injection sites for the splenius capitis and splenius cervicis were found at around the 50% point and the 30% to the 70% point, respectively.

The Nasolabial Fold: A Micro-Computed Tomography Study.

BACKGROUND: The nasolabial fold is known to be a challenging midface feature for aesthetic physicians. However, the steric conformation of the structures related to the nasolabial fold has remained undefined because the composition and topography of this region are highly intricate. Therefore, this study aimed to clarify the three-dimensional structures of the nasolabial fold using micro-computed tomography and verify their detailed composition by means of histologic observation. METHODS: Twenty-four specimens were collected from the area beside the alae nasi to the area above the oral angle of 12 cadavers (mean age, 80.3 years) bilaterally. Twelve specimens were evaluated by means of phosphotungstic acid contrast staining, and the rest were evaluated by means of histologic staining. All specimens were divided into three regions and analyzed comprehensively. RESULTS: The medial region of the nasolabial fold had dense irregular connective tissue intermingled with muscle fibers; the lateral region of the nasolabial fold had numerous fibrous septa with abundant adipose tissue. The levator labii alaeque nasi and the zygomaticus minor were attached to the medial part of the nasolabial fold, and the fascial septa were intermittently tethered to the dermis, lateral to the nasolabial fold. The extension of the adipose tissue within the fascial septa was limited by the lateral border of the muscle attachment. CONCLUSIONS: Dimensional and distributional alterations of the adipose tissues with senescence could render the nasolabial fold deeper by increasing the depth of the subcutaneous layer, lateral to the fold. Thus, to ameliorate the fold, the adipose tissue, lateral to the fold, or the muscle traction, medial to the fold, should be altered.

Stacking Ensemble Technique for Classifying Breast Cancer.

OBJECTIVES: Breast cancer is the second most common cancer among Korean women. Because breast cancer is strongly associated with negative emotional and physical changes, early detection and treatment of breast cancer are very important. As a supporting tool for classifying breast cancer, we tried to identify the best meta-learner model in a stacking ensemble when the same machine learning models for the base learner and meta-learner are used. METHODS: We used machine learning models, such as the gradient boosted model, distributed random forest, generalized linear model, and deep neural network in a stacking ensemble. These models were used to construct a base learner, and each of them was used as a meta-learner again. Then, we compared the performance of machine learning models in the meta-learner to determine the best meta-learner model in the stacking ensemble. RESULTS: Experimental results showed that using the GBM as a meta-learner led to higher accuracy than that achieved with any other model for breast cancer data and using the GLM as a meta learner led to low root-mean-squared error for both sets of breast cancer data. CONCLUSIONS: We compared the performance of every meta-learner model in a stacking ensemble as a supporting tool for classifying breast cancer. The study showed that using specific models as a metalearner resulted in better performance than single classifiers, and using GBM and GLM as a meta-learner is appropriate as a supporting tool for classifying breast cancer data.