Enhancing deep learning classification performance of tongue lesions in imbalanced data: mosaic-based soft labeling with curriculum learning.

BACKGROUND: Oral potentially malignant disorders (OPMDs) are associated with an increased risk of cancer of the oral cavity including the tongue. The early detection of oral cavity cancers and OPMDs is critical for reducing cancer-specific morbidity and mortality. Recently, there have been studies to apply the rapidly advancing technology of deep learning for diagnosing oral cavity cancer and OPMDs. However, several challenging issues such as class imbalance must be resolved to effectively train a deep learning model for medical imaging classification tasks. The aim of this study is to evaluate a new technique of artificial intelligence to improve the classification performance in an imbalanced tongue lesion dataset. METHODS: A total of 1,810 tongue images were used for the classification. The class-imbalanced dataset consisted of 372 instances of cancer, 141 instances of OPMDs, and 1,297 instances of noncancerous lesions. The EfficientNet model was used as the feature extraction model for classification. Mosaic data augmentation, soft labeling, and curriculum learning (CL) were employed to improve the classification performance of the convolutional neural network. RESULTS: Utilizing a mosaic-augmented dataset in conjunction with CL, the final model achieved an accuracy rate of 0.9444, surpassing conventional oversampling and weight balancing methods. The relative precision improvement rate for the minority class OPMD was 21.2%, while the relative [Formula: see text] score improvement rate of OPMD was 4.9%. CONCLUSIONS: The present study demonstrates that the integration of mosaic-based soft labeling and curriculum learning improves the classification performance of tongue lesions compared to previous methods, establishing a foundation for future research on effectively learning from imbalanced data.

Target registration errors in navigation-assisted mandibular surgery according to the tracking methods and the type of markers: experiments using human dry mandibular bone.

OBJECTIVES: This study was conducted to evaluate the accuracy of navigation process according to the type of tracking methods and registration markers. The target registration errors (TREs) were measured at seven anatomical landmarks of the mandible. METHODS: Four different experiments were performed to obtain the TREs using two tracking methods, the optical tracker (Polaris) and the electromagnetic (EM) tracker (Aurora), and two types of registration markers, invasive and noninvasive markers. All comparisons of TREs were statistically analyzed using SPSS and Python-based statistical package. RESULTS: The average TRE values obtained from the four experiments were as follows: (1) 0.85 mm (± 0.07) using invasive marker and Aurora, (2) 1.06 mm (± 0.12) using invasive marker and Polaris, (3) 1.43 mm (± 0.15) using noninvasive marker and Aurora, and (4) 1.57 mm (± 0.23) using noninvasive marker and Polaris. Comparisons between all the experimental results revealed statistically significant differences except for the type of tracking system. Although the comparison between the modality of the tracking system showed no significant differences, the EM-based approach consistently demonstrated better performances than the optical type in all comparisons. CONCLUSIONS: This study demonstrates that irrespective of the tracking modality, using invasive marker is a better choice in terms of accuracy. When using noninvasive marker, it is important to consider the increased TREs. In this study, the noninvasive marker caused a maximum increment of TREs of 0.81 mm compared with the invasive marker. Furthermore, using an EM-based tracker with invasive marker may result in the best accuracy for navigation.

Differences in mGluR5 Availability Depending on the Level of Social Avoidance in Drug-Naïve Young Patients with Major Depressive Disorder.

Background: Previous research has shown that metabotropic glutamate receptor-5 (mGluR5) signaling is significantly involved in social avoidance. We investigated the relationship between levels of social avoidance and mGluR5 availability in drug-naïve young patients with major depressive disorder (MDD). Methods: Twenty non-smoking patients and eighteen matched non-smoking healthy controls underwent [11C]ABP688 positron emission tomography (PET) and magnetic resonance imaging scans. The binding potential (BPND) of [11C]ABP688 was obtained using the simplified reference tissue model. Patients' level of social avoidance was assessed using the Social Avoidance and Distress Scale (SADS). For [11C]ABP688 BPND, the region-of-interest (ROI)-based between-group comparisons and correlations with SADS scores were investigated. The frontal cortices were chosen as a priori ROIs based on previous PET investigations in MDD, and on literature underscoring the importance of the frontal cortex in social avoidance. Results: Independent samples t-tests revealed no significant differences in [11C]ABP688 BPND in the frontal cortices between the MDD patient group as a whole and healthy controls. One-way analysis of variance with post-hoc tests revealed significantly lower BPND in the bilateral superior frontal cortex (SFC) and left middle frontal cortex (MFC) in MDD patients with low levels of social avoidance (L-SADS) than in healthy controls. The L-SADS patients also had significantly lower BPND in the medial part of the right SFC than both MDD patients with high levels of social avoidance (H-SADS) and healthy controls. The L-SADS patients also showed significantly lower BPND in the orbital parts of the SFC, MFC, and inferior frontal cortex than H-SADS patients. No significant group differences were found between H-SADS patients and healthy controls. The ROI-based correlation analysis revealed significant positive correlations between social avoidance levels and frontal [11C]ABP688 BPND in the entire patients. Conclusion: Our exploratory study shows significant differences in frontal mGluR5 availability depending on the level of social avoidance in drug-naïve non-smoking MDD patients, suggesting that social avoidance should be considered as one of the clinical factors involved in mGluR5 signaling changes in depression.

Errors according to the number of registered markers used in navigation-assisted surgery of the mandible.

BACKGROUND: The aim of this study was to evaluate the accuracy of navigation according to the number of markers in terms of target registration errors (TREs) at each anatomical location during the registration process of the navigation system for the mandible. METHODS: The TREs were measured in five different experiments, varying only in the number of registration reference markers, which ranged from three to seven. To measure the TREs according to the number of registration reference markers, two experimental navigation devices were used: 1) Cbyon navigation surgery equipment 2) Polaris optical tracker. Both experiments were conducted to obtain the TREs at the anatomical locations of the mandible according to the number of registration markers during the navigation process. Statistical analysis was performed using the SPSS 23.0 software. RESULTS: At all anatomical locations, errors were 2 mm or less. Further, significant differences in the target errors measured by the Cbyon system were found according to the number of registration markers. Significant differences in the target errors measured by the Polaris optical tracker were found according to the registration markers at the posterior border only. In both groups, the target errors did not decrease as the number of registration markers increased. CONCLUSIONS: This study demonstrates that an increase in the number of registration markers is not associated with a decrease in the TRE, and that a specific number of registration markers could reduce the TREs at each anatomical site. It is important to determine the minimum number of image registration markers at which the smallest TRE would be observed for different surgical sites.

A fusion PET-MRI system with a high-resolution research tomograph-PET and ultra-high field 7.0 T-MRI for the molecular-genetic imaging of the brain.

We have developed a positron emission tomography (PET) and magnetic resonance imaging (MRI) fusion system for the molecular-genetic imaging (MGI) of the in vivo human brain using two high-end imaging devices: the HRRT-PET, a high-resolution research tomograph dedicated to brain imaging on the molecular level, and the 7.0 T-MRI, an ultra-high field version used for morphological imaging. HRRT-PET delivers high-resolution molecular imaging with a resolution down to 2.5 mm full width at half maximum (FWHM), which allows us to observe the brain's molecular changes using the specific reporter genes and probes. On the other front, the 7.0 T-MRI, with submillimeter resolution images of the cortical areas down to 250 mum, allows us to visualize the fine details of the brainstem areas as well as the many cortical and subcortical areas. The new PET-MRI fusion imaging system will provide many answers to the questions on neurological diseases as well as cognitive neurosciences. Some examples of the answers are the quantitative visualization of neuronal functions by clear molecular and genetic bases, as well as diagnoses of many neurological diseases such as Parkinson's and Alzheimer's. The salient point of molecular-genetic imaging and diagnosis is the fact that they precede the morphological manifestations, and hence, the early and specific diagnosis of certain diseases, such as cancers.