Search

PURPOSE: To compare the evaluation metrics for deep learning methods that were developed using imbalanced imaging data in osteoarthritis studies. MATERIALS AND METHODS: This retrospective study utilized 2996 sagittal intermediate-weighted fat-suppressed knee MRIs with MRI Osteoarthritis Knee Score readings from 2467 participants in the Osteoarthritis Initiative study. We obtained probabilities of the presence of bone marrow lesions (BMLs) from MRIs in the testing dataset at the sub-region (15 sub-regions), compartment, and whole-knee levels based on the trained deep learning models. We compared different evaluation metrics (e.g., receiver operating characteristic (ROC) and precision-recall (PR) curves) in the testing dataset with various class ratios (presence of BMLs vs. absence of BMLs) at these three data levels to assess the models performance. RESULTS: In a subregion with an extremely high imbalance ratio, the model achieved a ROC-AUC of 0.84, a PR-AUC of 0.10, a sensitivity of 0, and a specificity of 1. CONCLUSION: The commonly used ROC curve is not sufficiently informative, especially in the case of imbalanced data. We provide the following practical suggestions based on our data analysis: 1) ROC-AUC is recommended for balanced data, 2) PR-AUC should be used for moderately imbalanced data (i.e., when the proportion of the minor class is above 5% and less than 50%), and 3) for severely imbalanced data (i.e., when the proportion of the minor class is below 5%), it is not practical to apply a deep learning model, even with the application of techniques addressing imbalanced data issues.

Proving clinical superiority of personalized care models in interventional and surgical pain management is challenging. The apparent difficulties may arise from the inability to standardize complex surgical procedures that often involve multiple steps. Ensuring the surgery is performed the same way every time is nearly impossible. Confounding factors, such as the variability of the patient population and selection bias regarding comorbidities and anatomical variations are also difficult to control for. Small sample sizes in study groups comparing iterations of a surgical protocol may amplify bias. It is essentially impossible to conceal the surgical treatment from the surgeon and the operating team. Restrictive inclusion and exclusion criteria may distort the study population to no longer reflect patients seen in daily practice. Hindsight bias is introduced by the inability to effectively blind patient group allocation, which affects clinical result interpretation, particularly if the outcome is already known to the investigators when the outcome analysis is performed (often a long time after the intervention). Randomization is equally problematic, as many patients want to avoid being randomly assigned to a study group, particularly if they perceive their surgeon to be unsure of which treatment will likely render the best clinical outcome for them. Ethical concerns may also exist if the study involves additional and unnecessary risks. Lastly, surgical trials are costly, especially if the tested interventions are complex and require long-term follow-up to assess their benefit. Traditional clinical testing of personalized surgical pain management treatments may be more challenging because individualized solutions tailored to each patient's pain generator can vary extensively. However, high-grade evidence is needed to prompt a protocol change and break with traditional image-based criteria for treatment. In this article, the authors review issues in surgical trials and offer practical solutions. [ABSTRACT FROM AUTHOR]

Aim: To evaluate the efficacy and safety of brachytherapy with double-strand 125I seeds and biliary drainage for malignant obstructive jaundice.Methods and Materials: 42 patients with obstructive jaundice because of extrahepatic cholangiocarcinoma were enrolled. 22 patients (group A) received a biliary stent with common drainage tube implantation, and 20 patients (group B) received a biliary stent with double-strand 125I seeds radiotherapy drainage tube placement. The length, location and pathological stage of biliary stricture were recorded in the two groups. Total bilirubin (TBIL), direct bilirubin (DBIL), IgA, IgG, IgM, alanine aminotransferase and white blood cell (WBC) count were measured before and after percutaneous transhepatic cholangial drainage (PTCD). Tumor diameter was measured before and three months after PTCD, and the difference were calculated. Stent patency time, survival time, and complications were recorded.Results: There was no significant difference in the length, location and pathological stage of biliary stenosis between the two groups. There was no significant difference in TBIL, DBIL, IgA, IgG, IgM, alanine aminotransferase and WBC count between the two groups before or after PTCD (P > 0.05). Three months after PTCD, tumors growth in group A and tumors shrinkage in group B. The difference in tumor size between the two groups before and after PTCD was statistically significant (P < 0.05). The average stent patency times in groups A and B were 3.55 ± 0.76 months and 8.76 ± 1.85 months, respectively (P < 0.05). The average survival times in groups A and B were 133.5 ± 27.8 days and 252.5 ± 114.5 days, respectively (P < 0.05). There was no statistically significant difference in the incidence of complications between the two groups (P > 0.05).Conclusion: Double-strand 125I seeds radiotherapy biliary drainage tubes can safely and effectively control tumors, prolong the patency of biliary stents, and prolong patient survival. [ABSTRACT FROM AUTHOR]

Glenoid component loosening remains a common complication following anatomic total shoulder arthroplasty (TSA); however, plain radiographs are unable to accurately detect early implant migration. The purpose of this study was to validate the accuracy of a method of postoperative, three‐dimensional (3D) computed tomography (CT) imaging with metal artifact reduction (MAR) to detect glenoid component migration following anatomic TSA. Tantalum bead markers were inserted into polyethylene glenoid components for implant detection on 3D CT. In‐vitro validation was performed using a glenoid component placed into a scapula sawbone and incrementally translated and rotated, with MAR 3D CT acquired at each test position. Accuracy was evaluated by root mean square error (RMSE). In‐vivo validation was performed on six patients who underwent anatomic TSA, with two postoperative CT scans acquired in each patient and marker‐based radiostereometric analysis (RSA) performed on the same days. Glenoid component migration was calculated relative to a scapular coordinate system for both MAR 3D CT and RSA. Accuracy was evaluated by RMSE and paired Student's t‐tests. The largest RMSE on in‐vitro testing was 0.24 mm in translation and 0.11° in rotation, and on in‐vivo testing was 0.47 mm in translation and 1.04° in rotation. There were no significant differences between MAR 3D CT and RSA measurement methods. MAR 3D CT imaging is capable of quantifying glenoid component migration with a high level of accuracy. MAR 3D CT imaging is advantageous over RSA because it is readily available clinically and can also be used to evaluate the implant‐bone interface. [ABSTRACT FROM AUTHOR]

Pyrochlore titanate (Y2Ti2O7) is a promising material for a wide range of applications in optoelectronics and photocatalysis due to its advantageous chemical, mechanical, and optical properties. To enhance its potential for such uses, however, a high-quality and scalable synthesis method is required. We here investigate the crystallization of sol–gel produced Y2Ti2O7 layers. We observe a transition of the amorphous pyrochlore phase at annealing temperatures below 700 °C. The transmittances of the Y2Ti2O7 thin layers annealed at 400 to 700 °C are approximately 92.3%. The refractive indices and packing densities of Y2Ti2O7 thin layers annealed at 400–700 °C/1 h vary from 1.931 to 1.954 and 0.835 to 0.846, respectively. The optical bandgap energies of Y2Ti2O7 thin layers annealed at 400–700 °C/1 h reduce from 4.356 to 4.319 eV because of the Moss–Burstein effect. These good electronic and optical properties make Y2Ti2O7 thin layers a promising host material for many potential applications. [ABSTRACT FROM AUTHOR]

The spacing between the median nerve and transverse carpal ligament (TCL) within the carpal tunnel can potentially affect the nerve morphology. This study aimed to quantify the spatial relationship between the median nerve and transverse carpal ligament in asymptomatic hands. Twelve subjects were recruited to image the carpal tunnel using robot-assisted ultrasound. The median nerve and TCL were segmented from each image and three-dimensionally reconstructed using kinematic information from the robot. The TCL-median nerve distance, nerve cross-sectional area, circularity, and position were measured along the entirety of the nerve length within the carpal tunnel. Results were averaged at every 5% of nerve length. At the nerve length percentages of 0% (distal), 25%, 50%, 75%, and 100% (proximal), the TCL-median nerve distance (±SD) was 0.7 ± 0.4, 0.7 ± 0.2, 0.5 ± 0.2, 0.5 ± 0.2, and 0.6 ± 0.3 mm, respectively. The corresponding nerve cross-sectional area was 9.4 ± 1.9, 10.6 ± 2.6, 11.2 ± 2.1, 11.2 ± 1.7, and 9.7 ± 1.9 mm2. A one-way analysis of variance showed no significant differences between the respective percentages of nerve length for TCL-median nerve distance (p = 0.219) and cross-sectional area (p = 0.869). Significant (p < 0.0001) but weak correlations were observed between the TCL-median nerve distance with cross-sectional area (r = -0.247) and circularity (r = -0.244). This study shows that the healthy median nerve morphology is consistent along the continuous nerve length within the carpal tunnel, supporting the use of 2D imaging in the evaluation of the healthy nerve., (Copyright © 2023 by ASME.)