Investigation of out-of-distribution detection across various models and training methodologies.

Machine learning-based algorithms demonstrate impressive performance across numerous fields; however, they continue to suffer from certain limitations. Even sophisticated and precise algorithms often make erroneous predictions when implemented with datasets having different distributions compared to the training set. Out-of-distribution (OOD) detection, which distinguishes data with different distributions from that of the training set, is a critical research area necessary to overcome these limitations and create more reliable algorithms. The OOD issue, particularly concerning image data, has been extensively studied. However, recently developed OOD methods do not fulfill the expectation that OOD performance will increase as the accuracy of in-distribution classification improves. Our research presents a comprehensive study on OOD detection performance across multiple models and training methodologies to verify this phenomenon. Specifically, we explore various pre-trained models popular in the computer vision field with both old and new OOD detection methods. The experimental results highlight the performance disparity in existing OOD methods. Based on these observations, we introduce Trimmed Rank with Inverse softMax probability (TRIM), a remarkably simple yet effective method for model weights with newly developed training methods. The proposed method could serve as a potential tool for enhancing OOD detection performance owing to its promising results. The OOD performance of TRIM is highly compatible with the in-distribution accuracy model and may bridge the efforts on improving in-distribution accuracy to the ability to distinguish OOD data.

Bacteria detection and species identification at the single-cell level using super-resolution fluorescence imaging and AI analysis.

The skin microbiome is thought to play a critical role in maintaining skin health and protecting against infection. While most microorganisms that live on the skin are harmless or even beneficial, some can cause skin infections or other health problems, emphasizing the importance of diagnosis of the composition and diversity of the skin flora. However, conventional diagnostic methods for evaluation of the skin microbiome are not sensitive enough to detect bacteria at low concentrations and suffer from poor specificity, thus limiting early diagnosis of bacterial infections. In this study, we developed novel approaches for bacterial species detection and identification methods with single-cell sensitivity using super-resolution microscopy and AI-based image analysis: a protein quantification-based method and an AI-based bacterial image analysis method. We demonstrate that these methods can differentiate between common bacterial members of the skin flora, including Staphylococcus aureus and Staphylococcus epidermidis, and different ribotypes of Cutibacterium acnes, both in purified bacterial samples and in scaling skin samples. The advantages of these methods, including the lack of time-consuming amplification or purification steps and single-cell level detection sensitivity, allow early diagnosis of bacterial infections, even from bacterial samples at extremely low concentrations, thus showing promise as a next-generation platform for microbiome detection as single-cell diagnostics.

Onlay patch augmentation in rotator cuff repair for moderate to large tears in elderly patients: clinical and radiologic outcomes.

BACKGROUND: This study evaluated the clinical and radiologic outcomes of onlay patch augmentation in rotator cuff repair for moderate-to-large tears in elderly patients. METHODS: We reviewed 24 patients who underwent onlay augmentation with dermal allograft after arthroscopic rotator cuff repair from January 2017 to March 2020. Inclusion criteria were patients aged >65 years with tears >2.5 cm, who were followed for >12 months after surgery, and patients who could raise their arms above 90° preoperatively. American Shoulder and Elbow Surgeons (ASES) score, Constant-Murley score, pain visual analog scale (VAS), and VAS for satisfaction were used as clinical outcomes. For the evaluation of cuff integrity, magnetic resonance imaging scans were performed every 3 months after surgery. The results were compared before and after surgery in all patients and between the retear and intact groups. RESULTS: The average follow-up period was 16.38 months, and the mean age of patients was 71.05 years. All patients showed significant improvement in ASES score, Constant-Murley score, and pain VAS at the last evaluation. The average value of satisfaction VAS was 7.27/10. The retear rate was 25% (6/24) if Sugaya type 3 was categorized in the retear group, otherwise 16.7% (4/24), if Sugaya type 3 was categorized into the intact group. Irrespective of Sugaya type 3 being included in the retear group, there was no significant difference in outcome variables between the intact and retear groups during follow-up. CONCLUSIONS: In moderate-to-large rotator cuff tear in elderly patients, onlay patch augmentation improved clinical outcomes. Retear did not adversely affect clinical outcomes.

Recent development of computational cluster analysis methods for single-molecule localization microscopy images.

With the development of super-resolution imaging techniques, it is crucial to understand protein structure at the nanoscale in terms of clustering and organization in a cell. However, cluster analysis from single-molecule localization microscopy (SMLM) images remains challenging because the classical computational cluster analysis methods developed for conventional microscopy images do not apply to pointillism SMLM data, necessitating the development of distinct methods for cluster analysis from SMLM images. In this review, we discuss the development of computational cluster analysis methods for SMLM images by categorizing them into classical and machine-learning-based methods. Finally, we address possible future directions for machine learning-based cluster analysis methods for SMLM data.

Development of Deep-Learning-Based Single-Molecule Localization Image Analysis.

Recent developments in super-resolution fluorescence microscopic techniques (SRM) have allowed for nanoscale imaging that greatly facilitates our understanding of nanostructures. However, the performance of single-molecule localization microscopy (SMLM) is significantly restricted by the image analysis method, as the final super-resolution image is reconstructed from identified localizations through computational analysis. With recent advancements in deep learning, many researchers have employed deep learning-based algorithms to analyze SMLM image data. This review discusses recent developments in deep-learning-based SMLM image analysis, including the limitations of existing fitting algorithms and how the quality of SMLM images can be improved through deep learning. Finally, we address possible future applications of deep learning methods for SMLM imaging.

Unrecognized bony Bankart lesion accompanying a dislocated four-part proximal humerus fracture before surgery: a case report.

Proximal humerus fractures are the third most common fractures, totaling 4% to 5% of all fractures. Here, we present the case of a 39-yearold man with a dislocated four-part fracture of the proximal humerus with a huge bony Bankart lesion. Preoperatively, the bony Bankart lesion of the glenoid was not visualized on computed tomography scans or magnetic resonance imaging because the fracture of the proximal humerus was comminuted, displaced, and complex. It was planned for only the humerus fracture to be treated by open reduction and internal fixation using a locking plate. However, a fractured fragment remained under the scapula after reduction of the dislocated humeral head. This was mistaken for a dislocated bone fragment of the greater tuberosity and repositioning was attempted. After failure, visual confirmation showed that the bone fragment was a piece of the glenoid. After reduction and fixation of this glenoid part with suture anchors, we acquired a well-reduced fluoroscopic image. Given this case of complex proximal humerus fracture, a glenoid fracture such as a bony Bankart lesion should be considered preoperatively and intraoperatively in such cases.

Complications and Frequency of Surgical Treatment with AO-Type Hook Plate in Shoulder Trauma: A Retrospective Study.

We investigated the complications and frequency of hook plate fixation in patients with shoulder trauma. We reviewed 216 cases of hook plate fixation use at our hospital between January 2010 and May 2020. Finally, we included 76 cases of acute distal clavicle fracture (DCF) and 84 cases of acute acromioclavicular joint dislocation (ACD). We investigated all complications after hook plate use, bony union in the DCF group, and reduction loss in the ACD group. We defined painful shoulder stiffness (PSS) as aggravating resting pain with stiff shoulder, and pain on shoulder elevation (PSE) as continued shoulder pain on elevation without PSS before plate removal. PSS was managed with intra-articular steroid injections or manipulation with or without arthroscopic capsular release (ACR). PSS occurred in 36 and 33 cases of DCF and ACD, respectively. PSE occurred in 17 of 76 fractures and 13 of 84 dislocations. However, no iatrogenic rotator cuff injury was verified by magnetic resonance imaging in patients with PSS or PSE. Subacromial erosion in patients with hook plate fixation should be considered a sequela and not a complication because it is unavoidable in surgery with an AO-type hook plate. The most common complication was PSS, followed by PSE.

How to handle noisy labels for robust learning from uncertainty.

Most deep neural networks (DNNs) are trained with large amounts of noisy labels when they are applied. As DNNs have the high capacity to fit any noisy labels, it is known to be difficult to train DNNs robustly with noisy labels. These noisy labels cause the performance degradation of DNNs due to the memorization effect by over-fitting. Earlier state-of-the-art methods used small loss tricks to efficiently resolve the robust training problem with noisy labels. In this paper, relationship between the uncertainties and the clean labels is analyzed. We present novel training method to use not only small loss trick but also labels that are likely to be clean labels selected from uncertainty called "Uncertain Aware Co-Training (UACT)". Our robust learning techniques (UACT) avoid over-fitting the DNNs by extremely noisy labels. By making better use of the uncertainty acquired from the network itself, we achieve good generalization performance. We compare the proposed method to the current state-of-the-art algorithms for noisy versions of MNIST, CIFAR-10, CIFAR-100, T-ImageNet and News to demonstrate its excellence.

Can manipulation under anesthesia alone provide clinical outcomes similar to arthroscopic circumferential capsular release in primary frozen shoulder (FS)?: the necessity of arthroscopic capsular release in primary FS.

BACKGROUND: We evaluated the need for arthroscopic capsular release (ACR) in refractory primary frozen shoulder (FS) by comparing clinical outcomes of patients treated with ACR and manipulation under anesthesia (MUA). METHODS: We assessed patients with refractory primary FS, 57 patients (group A) who were treated with MUA and 22 patients (group B) who were treated with ACR. In group A, manipulation including a backside arm-curl maneuver was performed under interscalene brachial block. In group B, manipulation was performed only to release the inferior capsule before arthroscopic circumferential capsular release, which was carried out for the unreleased capsule after manipulation. Pain, range of shoulder motion, and American Shoulder and Elbow Surgeons score were recorded at 1 week, 3 months, 6 months, and 1 year after surgery. We compared outcome variables between treatment groups and between diabetics and non-diabetics and also evaluated the numbers of patients receiving additional intra-articular steroid injection. RESULTS: Outcome variables at 3 months after surgery and improvements in outcome variables did not differ between groups. Group A showed significantly better results than group B in the evaluation of pain and range of motion at 1 week. Diabetics showed comparable outcomes to non-diabetics for most variables. Eleven patients required additional steroid injections between 8 to 16 weeks after surgery: 12.2% in group A, 18.2% in group B. Additional injections were given three times more often in diabetics compared to non-diabetics. CONCLUSIONS: MUA alone can yield similar clinical outcomes to ACR in refractory FS.

Strut Support with Tricortical Iliac Allografts in Unstable Proximal Humerus Fractures: Surgical Indication and New Definition of Poor Medial Column Support.

BACKGROUND: The execution of fibular allograft augmentation in unstable proximal humerus fractures (PHFs) was technically demanding. In this study, the authors evaluated the clinical and radiographic outcomes after tricortical iliac allograft (TIA) augmentation in PHFs. METHODS: We retrospectively assessed 38 PHF patients treated with locking-plate fixation and TIA augmentation. Insertion of a TIA was indicated when an unstable PHF showed a large cavitary defect and poor medial column support after open reduction, regardless of the presence of medial cortical comminution in preoperative images. Radiographic imaging parameters (humeral head height, HHH; humeral neck-shaft angle, HNSA; head mediolateral offset, HMLO; and status of the union), Constant score, and range of motion were evaluated. Patients were grouped according to whether the medial column support after open reduction was poor or not (groups A and B, respectively); clinical outcomes were compared for all parameters. RESULTS: All fractures healed radiologically (average duration to complete union, 5.8 months). At final evaluation, the average Constant score was 73 points and the mean active forward flexion was 148°. Based on the Paavolainen assessment method, 33 patients had good results and 5 patients showed fair results. The mean loss of reduction was 1.32 mm in HHH and 5.02% in HMLO. None of the parameters evaluated showed a statistically significant difference between the two groups (poor and not poor medial column support). CONCLUSIONS: In unstable PHFs, TIA augmentation can provide good clinical and radiological results when there are poor medial column support and a large cavitary defect after open reduction.

Effect of Weight-Bearing in Conservative and Operative Management of Fractures of the Base of the Fifth Metatarsal Bone.

BACKGROUND: There is no established principle regarding weight-bearing in conservative and operative management of fifth metatarsal base fractures. METHODS: We reviewed 86 patients with acute fifth metatarsal base fractures. Conservatively treated late or early weight-bearing patients were assigned to Group A or C, respectively. Operatively treated late or early weight-bearing patients were assigned to Group B or D, respectively. Results were evaluated by clinical union, bone resorption, and the American Orthopaedic Foot and Ankle Society (AOFAS) and Visual Analogue Scale (VAS) scores. RESULTS: All 4 groups had bone union at a mean of 6.9 weeks (range, 5.1-15.0). There were no differences between the groups in the AOFAS and VAS scores. In the early weight-bearing groups, there were fewer cases of bone resorption, and the bone unions periods were earlier. CONCLUSIONS: Early weight-bearing may help this patient population. Moreover, conservative treatment could be an option in patients with underlying diseases.

Surgical results of intradural extramedullary tumors.

BACKGROUND: To report the treatment results of 12 patients who underwent a total excision of intradural extramedullary tumors. METHODS: Twelve cases of histopathologically confirmed intradural extramedullary tumors were treated surgically between February 2002 and March 2005. There were 8 males and 4 females with an average age of 42.6 years. The mean postoperative follow-up period was 24.2 months. The histopathological findings, locations of the tumors, and clinical results were analyzed. The neurological findings obtained during the preoperative stage and the postoperative follow-up were evaluated according to the Frankel classification. RESULTS: The histopathological results are as follows: 4 cases of a meningioma, 4 cases of a schwannoma, 2 cases of an epidermoid cyst, 1 case of an arachnoid cyst, and 1 case of an ependymoma. The locations of the tumors were as follows: 7 cases in the thoracic region, 4 cases in the lumbar region, and 1 case in the cervical region. At the final follow-up, a 2-grade and 1-grade improvement was observed in 1 and 7 cases, respectively. There were no changes in the Frankel grade in 4 cases. The preoperative neurological deficit improved within 8 postoperative weeks in most cases and within 1 postoperative year in all cases. Postoperatively, there were 2 cases of cerebrospinal fluid leakage and 2 cases of paresthesia. CONCLUSIONS: Intradural extramedullary tumors detected by MRI are mostly benign and good clinical results can be obtained when treated surgically. Therefore, more active surgical approaches by orthopedic surgeons are recommended to decrease morbidity.