XGboost Prediction Model Based on 3.0T Diffusion Kurtosis Imaging Improves the Diagnostic Accuracy of MRI BiRADS 4 Masses

Diagnostic Accuracy of Machine Learning-Assisted MRI for Mild Cognitive Impairment in Parkinson's Disease: A Systematic Review and Meta-Analysis

To evaluate the diagnostic accuracy of machine learning-assisted magnetic resonance imaging (MRI) in detecting cognitive impairment among Parkinson's disease (PD) patients through a systematic review and meta-analysis. We systematically searched for studies that applied machine learning algorithms to MRI data for diagnosing PD with mild cognitive impairment (PD-MCI). Data were extracted and synthesized to calculate pooled sensitivity, specificity, positive likelihood ratio (PLR) and negative diagnostic likelihood ratio (NLR), and diagnostic odds ratios (DOR). A bivariate random-effects model and summary receiver operating characteristic (SROC) curves were employed for statistical analysis. The quality of studies was assessed using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) instrument. The publication bias was investigated through Deeks' funnel plot. All statistical analyses were conducted using Stata 14.0. The pooled sensitivity and specificity for diagnosing PD-MCI using machine learning-assisted MRI were 0.82 (95% CI: 0.75-0.87) and 0.81 (95% CI: 0.73-0.87), respectively. The PLR was 4.28 (95% CI: 2.93-6.27), and the NLR was 0.23 (95% CI: 0.16-0.32), indicating a high diagnostic accuracy. The area under the curve (AUC) for the SROC was 0.85 (95% CI: 0.82-0.88). Quality assessment using the QUADAS-2 tool showed a predominantly low risk of bias among the studies, and the Deeks' funnel plot suggested no significant publication bias (p=0.30). In summary, the MRI combined with machine learning for diagnosing PD-MCI achieved high accuracy with the pooled sensitivity of 82% and specificity of 81%.

Exploring the Potential of Machine Learning Algorithms to Improve Diffusion Nuclear Magnetic Resonance Imaging Models Analysis

Purpose

This paper explores different machine learning (ML) algorithms for analyzing diffusion nuclear magnetic resonance imaging (dMRI) models when analytical fitting shows restrictions. It reviews various ML techniques for dMRI analysis and evaluates their performance on different b-values range datasets, comparing them with analytical methods.

Materials and Methods

After standard fitting for reference, four sets of diffusion-weighted nuclear magnetic resonance images were used to train/test various ML algorithms for prediction of diffusion coefficient (D), pseudo-diffusion coefficient (D*), perfusion fraction (f), and kurtosis (K). ML classification algorithms, including extra-tree classifier (ETC), logistic regression, C-support vector, extra-gradient boost, and multilayer perceptron (MLP), were used to determine the existence of diffusion parameters (D, D*, f, and K) within single voxels. Regression algorithms, including linear regression, polynomial regression, ridge, lasso, random forest (RF), elastic-net, and support-vector machines, were used to estimate the value of the diffusion parameters. Performance was evaluated using accuracy (ACC), area under the curve (AUC) tests, and cross-validation root mean square error (RMSECV). Computational timing was also assessed.

Results

ETC and MLP were the best classifiers, with 94.1% and 91.7%, respectively, for the ACC test and 98.7% and 96.3% for the AUC test. For parameter estimation, RF algorithm yielded the most accurate results The RMSECV percentages were: 8.39% for D, 3.57% for D*, 4.52% for f, and 3.53% for K. After the training phase, the ML methods demonstrated a substantial decrease in computational time, being approximately 232 times faster than the conventional methods.

Conclusions

The findings suggest that ML algorithms can enhance the efficiency of dMRI model analysis and offer new perspectives on the microstructural and functional organization of biological tissues. This paper also discusses the limitations and future directions of ML-based dMRI analysis.

3.0 T multi-parametric MRI reveals metabolic and microstructural abnormalities in the posterior visual pathways in patients with thyroid eye disease

Introduction

We aim to explore the microstructural and metabolic changes in visual pathways in patients with thyroid eye disease (TED) using 3T multi-parametric MRI.

Methods

Thirty-four TED patients (inactive group = 20; active group = 14; acute group = 18; chronic group = 16) and 12 healthy controls (HC) were recruited from November 2020 to July 2021. Proton magnetic resonance spectroscopy (1H-MRS), glutamate chemical exchange saturation transfer (GluCEST) and diffusion kurtosis imaging (DKI) were performed on 3.0T MR scanner. Data analysis and group comparisons were performed after MR data processing.

Results

As compare to HC group, the levels of total choline (tCh) in optic radiation (OR) in active group ([1.404 ± 0.560] vs. [1.022 ± 0.260]; p < 0.05), together with tCh ([1.415 ± 0.507] vs. [1.022 ± 0.260]; p < 0.05) in OR in acute group were significantly increased. Glutamine (Gln) levels were higher in OR in the chronic group than those in HCs and were positively correlated with the levels of thyroglobulin antibody (TgAb), thyroid peroxidase antibody (TPOAb), free triiodothyronine (FT3) and FT4 in chronic group. Glutamate (Glu) levels by 1H-MRS did not show significant differences between any two groups. Interestingly, MTRasym (3.0 ppm) was higher in OL in inactive group, active group, acute group and chronic group than those in HCs, and was positively correlated with Glu levels in OL in 1H-MRS. Fractional anisotropy (FA) values from DKI in OR in acute group were significantly lower than those in HCs.

Discussion

Our initial study demonstrate that GluCEST performs better than 1H-MRS to monitor Glu alterations in visual pathway in TED patients. Changes of brain glutamine levels in TED patients are closely related to their associated hormones alterations, indicating that disease injury status could be reflected through non-invasive metabolites detection by brain 1H-MRS. FA is the most sensitive DKI index to reveal the visual pathway impairment in TED patients. Altogether, our study revealed that 3T multiparametric MR techniques are useful to demonstrate metabolic and microstructural alterations in visual pathways in TED patients. We found that damage to visual pathways occurs in mild TED cases, which not only offers a new approach to the diagnosis of dysthyroid optic neuropathy, but also demonstrates neuropathy in TED is a gradual and continuous spatio-emporal progression.

Prediction of postoperative recovery in patients with acoustic neuroma using machine learning and SMOTE-ENN techniques

Acoustic neuroma is a common benign tumor that is frequently associated with postoperative complications such as facial nerve dysfunction, which greatly affects the physical and mental health of patients. In this paper, clinical data of patients with acoustic neuroma treated with microsurgery by the same operator at Xiangya Hospital of Central South University from June 2018 to March 2020 are used as the study object. Machine learning and SMOTE-ENN techniques are used to accurately predict postoperative facial nerve function recovery, thus filling a gap in auxiliary diagnosis within the field of facial nerve treatment in acoustic neuroma. First, raw clinical data are processed and dependent variables are identified based on clinical context and data characteristics. Secondly, data balancing is corrected using the SMOTE-ENN technique. Finally, XGBoost is selected to construct a prediction model for patients' postoperative recovery, and is also compared with a total of four machine learning models, LR, SVM, CART, and RF. We find that XGBoost can most accurately predict the postoperative facial nerve function recovery, with a prediction accuracy of 90.0% and an AUC value of 0.90. CART, RF, and XGBoost can further select the more important preoperative indicators and provide therapeutic assistance to physicians, thereby improving the patient's postoperative recovery. The results show that machine learning and SMOTE-ENN techniques can handle complex clinical data and achieve accurate predictions.