Deep learning for risk stratification of thymoma pathological subtypes based on preoperative CT images.

OBJECTIVES: This study aims to develop an innovative, deep model for thymoma risk stratification using preoperative CT images. Current algorithms predominantly focus on radiomic features or 2D deep features and require manual tumor segmentation by radiologists, limiting their practical applicability. METHODS: The deep model was trained and tested on a dataset comprising CT images from 147 patients (82 female; mean age, 54 years ± 10) who underwent surgical resection and received subsequent pathological confirmation. The eligible participants were divided into a training cohort (117 patients) and a testing cohort (30 patients) based on the CT scan time. The model consists of two stages: 3D tumor segmentation and risk stratification. The radiomic model and deep model (2D) were constructed for comparative analysis. Model performance was evaluated through dice coefficient, area under the curve (AUC), and accuracy. RESULTS: In both the training and testing cohorts, the deep model demonstrated better performance in differentiating thymoma risk, boasting AUCs of 0.998 and 0.893 respectively. This was compared to the radiomic model (AUCs of 0.773 and 0.769) and deep model (2D) (AUCs of 0.981 and 0.760). Notably, the deep model was capable of simultaneously identifying lesions, segmenting the region of interest (ROI), and differentiating the risk of thymoma on arterial phase CT images. Its diagnostic prowess outperformed that of the baseline model. CONCLUSIONS: The deep model has the potential to serve as an innovative decision-making tool, assisting on clinical prognosis evaluation and the discernment of suitable treatments for different thymoma pathological subtypes. KEY POINTS: • This study incorporated both tumor segmentation and risk stratification. • The deep model, using clinical and 3D deep features, effectively predicted thymoma risk. • The deep model improved AUCs by 16.1pt and 17.5pt compared to radiomic model and deep model (2D) respectively.

Deep Learning Model for Predicting Proliferative Hepatocellular Carcinoma Using Dynamic Contrast-Enhanced MRI: Implications for Early Recurrence Prediction Following Radical Resection.

RATIONALE AND OBJECTIVES: The proliferative nature of hepatocellular carcinoma (HCC) is closely related to early recurrence following radical resection. This study develops and validates a deep learning (DL) prediction model to distinguish between proliferative and non-proliferative HCCs using dynamic contrast-enhanced MRI (DCE-MRI), aiming to refine preoperative assessments and optimize treatment strategies by assessing early recurrence risk. MATERIALS AND METHODS: In this retrospective study, 355 HCC patients from two Chinese medical centers (April 2018-February 2023) who underwent radical resection were included. Patient data were collected from medical records, imaging databases, and pathology reports. The cohort was divided into a training set (n = 251), an internal test set (n = 62), and external test sets (n = 42). A DL model was developed using DCE-MRI images of primary tumors. Clinical and radiological models were generated from their respective features, and fusion strategies were employed for combined model development. The discriminative abilities of the clinical, radiological, DL, and combined models were extensively analyzed. The performances of these models were evaluated against pathological diagnoses, with independent and fusion DL-based models validated for clinical utility in predicting early recurrence. RESULTS: The DL model, using DCE-MRI, outperformed clinical and radiological feature-based models in predicting proliferative HCC. The area under the curve (AUC) for the DL model was 0.98, 0.89, and 0.83 in the training, internal validation, and external validation sets, respectively. The AUCs for the combined DL and clinical feature models were 0.99, 0.86, and 0.83 in these sets, while the AUCs for the combined DL, clinical, and radiological model were 0.99, 0.87, and 0.8, respectively. Among models predicting early recurrence, the DL plus clinical features model showed superior performance. CONCLUSION: The DL-based DCE-MRI model demonstrated robust performance in predicting proliferative HCC and stratifying patient risk for early postoperative recurrence. As a non-invasive tool, it shows promise in enhancing decision-making for individualized HCC management strategies.

Development and Validation of Multi-Omics Thymoma Risk Classification Model Based on Transfer Learning.

The paper aims to develop prediction model that integrates clinical, radiomics, and deep features using transfer learning to stratifying between high and low risk of thymoma. Our study enrolled 150 patients with thymoma (76 low-risk and 74 high-risk) who underwent surgical resection and pathologically confirmed in Shengjing Hospital of China Medical University from January 2018 to December 2020. The training cohort consisted of 120 patients (80%) and the test cohort consisted of 30 patients (20%). The 2590 radiomics and 192 deep features from non-enhanced, arterial, and venous phase CT images were extracted and ANOVA, Pearson correlation coefficient, PCA, and LASSO were used to select the most significant features. A fusion model that integrated clinical, radiomics, and deep features was developed with SVM classifiers to predict the risk level of thymoma, and accuracy, sensitivity, specificity, ROC curves, and AUC were applied to evaluate the classification model. In both the training and test cohorts, the fusion model demonstrated better performance in stratifying high and low risk of thymoma. It had AUCs of 0.99 and 0.95, and an accuracy of 0.93 and 0.83, respectively. This was compared to the clinical model (AUCs of 0.70 and 0.51, accuracy of 0.68 and 0.47), the radiomics model (AUCs of 0.97 and 0.82, accuracy of 0.93 and 0.80), and the deep model (AUCs of 0.94 and 0.85, accuracy of 0.88 and 0.80). The fusion model integrating clinical, radiomics and deep features based on transfer learning was efficient for noninvasively stratifying high risk and low risk of thymoma. The models could help to determine surgery strategy for thymoma cancer.

Alterations of voxel-wise spontaneous activity and corresponding brain functional networks in multiple system atrophy patients with mild cognitive impairment.

Emerging evidence has indicated that cognitive impairment is an underrecognized feature of multiple system atrophy (MSA). Mild cognitive impairment (MCI) is related to a high risk of dementia. However, the mechanism underlying MCI in MSA remains controversial. In this study, we conducted the amplitude of low-frequency fluctuation (ALFF) and seed-based functional connectivity (FC) analyses to detect the characteristics of local neural activity and corresponding network alterations in MSA patients with MCI (MSA-MCI). We enrolled 80 probable MSA patients classified as cognitively normal (MSA-NC, n = 36) and MSA-MCI (n = 44) and 40 healthy controls. Compared with MSA-NC, MSA-MCI exhibited decreased ALFF in the right dorsal lateral prefrontal cortex (RDLPFC) and increased ALFF in the right cerebellar lobule IX and lobule IV-V. In the secondary FC analyses, decreased FC in the left inferior parietal lobe (IPL) was observed when we set the RDLPFC as the seed region. Decreased FC in the bilateral cuneus, left precuneus, and left IPL and increased FC in the right middle temporal gyrus were shown when we set the right cerebellar lobule IX as the seed region. Furthermore, FC of DLPFC-IPL and cerebello-cerebral circuit, as well as ALFF alterations, were significantly correlated with Montreal Cognitive Assessment scores in MSA patients. We also employed whole-brain voxel-based morphometry analysis, but no gray matter atrophy was detected between the patient subgroups. Our findings indicate that altered spontaneous activity in the DLPFC and the cerebellum and disrupted DLPFC-IPL, cerebello-cerebral networks are possible biomarkers of early cognitive decline in MSA patients.

Multimodal striatal neuromarkers in distinguishing parkinsonian variant of multiple system atrophy from idiopathic Parkinson's disease.

AIMS: To develop an automatic method of classification for parkinsonian variant of multiple system atrophy (MSA-P) and Idiopathic Parkinson's disease (IPD) in early to moderately advanced stages based on multimodal striatal alterations and identify the striatal neuromarkers for distinction. METHODS: 77 IPD and 75 MSA-P patients underwent 3.0 T multimodal MRI comprising susceptibility-weighted imaging, resting-state functional magnetic resonance imaging, T1-weighted imaging, and diffusion tensor imaging. Iron-radiomic features, volumes, functional and diffusion scalars of bilateral 10 striatal subregions were calculated and provided to the support vector machine for classification RESULTS: A combination of iron-radiomic features, function, diffusion, and volumetric measures optimally distinguished IPD and MSA-P in the testing dataset (accuracy 0.911 and area under the receiver operating characteristic curves [AUC] 0.927). The diagnostic performance further improved when incorporating clinical variables into the multimodal model (accuracy 0.934 and AUC 0.953). The most crucial factor for classification was the functional activity of the left dorsolateral putamen. CONCLUSION: The machine learning algorithm applied to multimodal striatal dysfunction depicted dorsal striatum and supervening prefrontal lobe and cerebellar dysfunction through the frontostriatal and cerebello-striatal connections and facilitated accurate classification between IPD and MSA-P. The dorsolateral putamen was the most valuable neuromarker for the classification.

Identifying and validating subtypes of Parkinson's disease based on multimodal MRI data via hierarchical clustering analysis.

Objective: We wished to explore Parkinson's disease (PD) subtypes by clustering analysis based on the multimodal magnetic resonance imaging (MRI) indices amplitude of low-frequency fluctuation (ALFF) and gray matter volume (GMV). Then, we analyzed the differences between PD subtypes. Methods: Eighty-six PD patients and 44 healthy controls (HCs) were recruited. We extracted ALFF and GMV according to the Anatomical Automatic Labeling (AAL) partition using Data Processing and Analysis for Brain Imaging (DPABI) software. The Ward linkage method was used for hierarchical clustering analysis. DPABI was employed to compare differences in ALFF and GMV between groups. Results: Two subtypes of PD were identified. The "diffuse malignant subtype" was characterized by reduced ALFF in the visual-related cortex and extensive reduction of GMV with severe impairment in motor function and cognitive function. The "mild subtype" was characterized by increased ALFF in the frontal lobe, temporal lobe, and sensorimotor cortex, and a slight decrease in GMV with mild impairment of motor function and cognitive function. Conclusion: Hierarchical clustering analysis based on multimodal MRI indices could be employed to identify two PD subtypes. These two PD subtypes showed different neurodegenerative patterns upon imaging.

Use of machine learning method on automatic classification of motor subtype of Parkinson's disease based on multilevel indices of rs-fMRI.

OBJECTIVE: This study aimed to develop an automatic classifier to distinguish different motor subtypes of Parkinson's disease (PD) based on multilevel indices of resting-state functional magnetic resonance imaging (rs-fMRI). METHODS: Ninety-six PD patients, which included thirty-nine postural instability and gait difficulty (PIGD) subtype and fifty-seven tremor-dominant (TD) subtype, were enrolled and allocated to training and validation datasets with a ratio of 7:3. A total of five types of index, consisting of mean regional homogeneity (mReHo), mean amplitude of low-frequency fluctuation (mALFF), degree of centrality (DC), voxel-mirrored homotopic connectivity (VMHC), and functional connectivity (FC), were extracted. The features were then selected using a two-sample t-test, the least absolute shrinkage and selection operator (LASSO), and Spearman's rank correlation coefficient. Finally, support vector machine (SVM) models based on the separate index and multilevel indices were built, and the performance of models was assessed via the area under the receiver operating characteristic curve (AUC). Feature importance was evaluated using Shapley additive explanation (SHAP) values. RESULTS: The optimal SVM model was obtained based on multilevel rs-fMRI indices, with an AUC of 0.934 in the training dataset and an AUC of 0.917 in the validation dataset. The AUCs of the models based on the separate index were ranged from 0.783 to 0.858 for the training dataset and from 0.713 to 0.792 for the validation dataset. SHAP analysis revealed that functional activity and connectivity in frontal lobe and cerebellum were important features for differentiating PD subtypes. CONCLUSIONS: Our findings demonstrated multilevel rs-fMRI indices could provide more comprehensive information on brain functionalteration. Furthermore, the machine learning method based on multilevel rs-fMRI indices might be served as an alternative approach for automatically classifying clinical subtypes in PD at the individual level.

Morphological Changes in Cortical and Subcortical Structures in Multiple System Atrophy Patients With Mild Cognitive Impairment.

OBJECTIVE: This study aimed to investigate the morphometric alterations in the cortical and subcortical structures in multiple system atrophy (MSA) patients with mild cognitive impairment (MCI), and to explore the association with cognitive deficits. METHODS: A total of 45 MSA patients (25 MSA-only, 20 MSA-MCI) and 29 healthy controls were recruited. FreeSurfer software was used to analyze cortical thickness, and voxel-based morphometry was used to analyze the gray matter volumes. Cortical thickness and gray matter volume changes were correlated with cognitive scores. RESULTS: Compared to healthy controls, both MSA subgroups exhibited widespread morphology alterations of brain structures in the fronto-temporal regions. Direct comparison of MSA-MCI and MSA-only patients showed volume reduction in the left superior and middle temporal gyrus, while cortical thinning was found in the left middle and inferior temporal gyrus in MSA-MCI patients. Cortical thinning in the left middle temporal gyrus correlated with cognitive assessment and disease duration. CONCLUSION: Structural changes in the brain occur in MSA-MCI patients. The alteration of brain structure in the left temporal regions might be a biomarker of cognitive decline in MSA-MCI patients.

Alterations in Degree Centrality and Functional Connectivity in Parkinson's Disease Patients With Freezing of Gait: A Resting-State Functional Magnetic Resonance Imaging Study.

OBJECTIVE: Freezing of gait (FOG) is a common disabling motor symptom in Parkinson's disease (PD), but the potential pathogenic mechanisms are still unclear. METHODS: A total of 22 patients with PD with FOG (PD-FOG), 28 patients with PD without FOG (PD-nFOG), and 33 healthy controls (HCs) were recruited in this study. Degree centrality (DC)-a graph theory-based measurement of global connectivity at the voxel level by measuring the number of instantaneous functional connections between one region and the rest of the brain-can map brain hubs with high sensitivity, specificity, and reproducibility. DC was used to explore alterations in the centrality of PD-FOG correlated with brain node levels. PD-FOG cognitive network dysfunction was further revealed via a seed-based functional connectivity (FC) analysis. In addition, correlation analyses were carried out between clinical symptoms and acquired connectivity measurement. RESULTS: Compared to the PD-nFOG group, the PD-FOG group showed remarkably increased DC values in the right middle frontal gyrus (RMFG). There were no significant differences in other gray matter regions. Importantly, the clinical severity of FOG was related to the mean DC values in the RMFG. This brain region served as a seed in secondary seed-based FC analysis, and we further found FC changes in the right precuneus, right inferior frontal gyrus, right superior frontal gyrus (SFG), and cerebellum. CONCLUSION: Increased RMFG activity and FC network alterations in the middle frontal cortex with the precuneus, inferior, and SFG, and the cerebellum may have great potential in brain dysfunction in PD with FOG.

Altered resting-state voxel-level whole-brain functional connectivity in multiple system atrophy patients with cognitive impairment.

OBJECTIVE: There is increasing evidence of cognitive impairment (CI) frequently occurring in patients with multiple system atrophy (MSA); however, the neurobiological mechanisms underlying CI in patients with MSA remain unclear. METHODS: We enrolled 61 patients with probable MSA and 33 healthy controls (HC). We used degree centrality (DC) analysis to assess changes in the centrality level of MSA-CI related brain nodes. We conducted a secondary seed-based functional connectivity (FC) analysis to investigate dysfunctions in cognitive networks related to MSA. Further, we analysed the correlation between clinical symptoms and acquired connectivity measures. RESULTS: Compared with HC, patients with MSA-CI and those with MSA with normal cognition (MSA-NCI) exhibited lower DC values in the left calcarine and right postcentral regions and higher DC values in the bilateral caudate and left precuneus. There were significant differences in the DC values in the right middle prefrontal gyrus between the MSA-CI and MSA-NCI groups. The mean DC values in the right middle prefrontal gyrus (RMPFG) were correlated with clinical cognitive severity. Consequently, we used this brain region as a seed in secondary seed-based FC analysis and observed FC changes within the right precuneus, inferior parietal lobe, and right insula. CONCLUSIONS: Decreased middle prefrontal cortex activity and its altered functional connectivity with the precuneus, inferior parietal lobe, and insula are possible biomarkers of cognitive dysfunction in patients with MSA-CI. SIGNIFICANCE: Cognitive impairment in MSA is associated with alterations in the dorsolateral prefrontal cortex network.