Keywords: Diagnosis/Prediction, Epilepsy

Motivation: This study seeks to address the challenge of limited visibility of periventricular lesions in Tuberous Sclerosis Complex (TSC).

Goal(s): Develop FLAIR3, a deep neural network, for adaptive fusion of T2w and FLAIR images in TSC patients to improve lesion detection.

Approach: The study adopts a dual-stream U-Net network with a pre-fusion module and employs spatial and channel fusion weight for feature fusion. Gradient loss and segmentation annotations are utilized to generate fusion images with clear textures and improved contrast.

Results: The fused image, FLAIR3, demonstrates enhanced lesion contrast and outperforms T2w and FLAIR images in lesion segmentation.

Impact: The enhanced lesion visualization provided by FLAIR3 can aid doctors in accurately identifying and diagnosing cortical tubers, improving the overall epilepsy diagnosis and treatment in TSC patients. This work improves the accuracy of automatic tuber segmentation.

Keywords: Diagnosis/Prediction, Aging

Motivation: Enlarged perivascular space (PVS) has been brought into attention in aging populations. However, which cardiovascular risk factors contribute to enlarged PVS are not well understood.

Goal(s): This study aims to quantify PVS morphology and investigate which cardiovascular risk factors contribute the PVS deformity in aging populations.

Approach: We employed random forest to predict PVS morphological changes using 9 cardiovascular risk factors and computed the importance index for all predictive factors.

Results: Our findings highlighted the significant role of sleep quality, being the best predictor to PVS count, linearity, and diameter. Cardiovascular risk factors such as triglycerides best predicted PVS tortuosity.

Impact: Our study is the first to investigate which cardiovascular risk factors are predictive of atypical PVS morphology. Our discovery provides valuable insights into the mechanism underlying PVS deformity and their subsequent impact on glymphatic system and cerebral vascular diseases.

Keywords: Diagnosis/Prediction, Quantitative Susceptibility mapping

Motivation:  Rim lesions are important subset of chronic active MS lesions that show strong correlation to patient disability. Rim identification by experts is time consuming.

Goal(s): Develop tool for supporting the expert in Rim identification using 1 mm QSM.

Approach: We developed an automated deep learning-based network for PRL detection on thin-slice 1mm QSMp. We evaluated the improvement in performance compared with networks trained using 1mm QSM and 3mm QSMp. 

Results: Use of high-resolution positive susceptibility source maps improves detection of Rim in MS patients compared to 1mm QSM and 3mm QSMp. The network does not require a precise QSM lesion mask to operate.

Impact: Using the Deep learning for detecting rim on 1mm QSMp, enabling reducing workload for human in detecting rim.

Keywords: Diagnosis/Prediction, Multimodal, functional connectivity, structural connectivity, graph convolution network, Parkinson's disease

Motivation: Brain functional connectivity (FC) and structural connectivity (SC) have distinct neural mechanisms for Parkinson’s disease (PD). Furthermore, the interactions between SC and FC could reveal underlying mechanisms and enhance classification performance. 

Goal(s): We aim to utilize structure-function interactions for PD classification.

Approach: We propose a brain structure-function interaction model via graph convolution network to incorporate both modality-specific embeddings and structure-function interactions.

Results: Results on 72 PD patients and 69 normal controls demonstrate that our method outperforms other state-of-the-art methods. We identify strong structure-function couplings in the precentral gyrus, prefrontal, superior temporal, cingulate cortices, and cerebellum that are associated with PD.

Impact: We proposed a novel brain structure-function interaction network based on GCN to utilize modality-specific features and interactions of SC and FC for PD classification. Our method identified the coupling strengths between SC and FC associated with PD.

Keywords: Diagnosis/Prediction, Epilepsy, medication; depp learning

Motivation: Epilepsy is a complex neurological disorder with a high degree of heterogeneity. Selecting the appropriate antiseizure medication(ASM) is a time-consuming trial-and-error process that requires expert knowledge from neurologists.

Goal(s): Our goal was to utilise Deep Learning(DL) techniques with neuroimaging information to predict the treatment outcome of ASM.

Approach: We developed a DL model that utilises multi-modal information (MRI scans and clinical characteristics) to predict seizure outcomes of initial ASM for patients with newly diagnosed epilepsy.

Results: Our model achieved AUROC/AUPRC of 0.72/0.71 respectively in predicting treatment outcomes, demonstrating the potential of brain MRI scans as a biomarker for treatment response.

Impact: The model showed promise for development of decision-support systems that could help neurologists select the best ASM, potentially improving treatment outcomes. Clinical translation will require larger datasets and external validation, but this work implies that MRI contains additional prognostic information.

Keywords: Diagnosis/Prediction, Machine Learning/Artificial Intelligence, 4D Dynamic PET, PCA

Motivation: The reduction in PET scan duration not only improves the efficiency of the scanning process but also contributes to a more comfortable experience for patients.

Goal(s): Leveraging the temporal models in conjunction with previously predicted weights of PCs, we aim to reconstruct entire 4D dynamic PET frames using an inverse PCA method. 

Approach: A novel technique has been developed to generate pseudo-T1 images from noisy 4D PET data, as well as the reverse process, obtaining the initial components of 4D dynamic PET images from MRI data.

Results: The results endorsed that only 5 minutes observation is enough to predict whole 70 minute data.

Impact: We predicted later PET frames from noisy initial frames using a novel approach combining anatomical and statistical temporal PCs from MRI data. This method has clinical potential for insights into dynamic processes, radiation reduction, and identifying abnormalities in medical imaging.

Keywords: Diagnosis/Prediction, Machine Learning/Artificial Intelligence

Motivation: ATAs are essential indicators of collateral pathways in cerebral perfusion anomalies. However, the conventional grading systems for ATA suffer from subjectivity, which may subjectively leads to variability

Goal(s): We aim to standardize ATA grading by a deep learning fusion model that combines information from ASL and DWI

Approach: A deep learning fusion model was developed, which applies two 3D CNNs to extract respective feature map of each modality; this model combines the high-level feature maps to fuse the multi-sequence MRI information

Results: The fusion model shows significant improvements over a single modality model, achieving an AUC value of 0.895

Impact: The good ATA evaluation performance of the deep learning fusion model shows its clinical potential in assisting neuroradiologists in conducting the treatment and prognosis analysis for patients with ischemic stroke

Keywords: Diagnosis/Prediction, Data Analysis, Deep Learning

Motivation: Current deep learning methods for assessing knee osteoarthritis have limitations in learning long-range spatial information from magnetic resonance imaging (MRI).

Goal(s): This study aims to develop a new deep learning model for total knee replacement (TKR) prediction using MRI.

Approach: We proposed a novel transformer-based model, MR-Transformer, adapted from the ImageNet pre-trained vision transformer DeiT-Ti. The model can capture long-range spatial information from MR images with transformer architecture. We evaluated our model on TKR prediction using MR images with different tissue contrasts.

Results: The experimental results demonstrated an improved performance of MR-Transformer compared to conventional deep learning models.

Impact: Our proposed MR-Transformer enhances computer-aided diagnosis accuracy in total knee replacement prediction using MRI. It has the potential to provide rapid and quality diagnostic outcomes, assisting physicians in making timely and informed treatment decisions.

Keywords: Diagnosis/Prediction, Quantitative Imaging, Cardiac T1-rho mapping; automated analysis

Motivation: Contrast agent-free myocardial T1-rho (T1ρ) mapping has shown promise in myocardial injury quantification. However, the lack of analysis tools hinders its clinical use and induces increased workload and operator variability.

Goal(s): To explore the feasibility and benefits of clinically-integrated artificial intelligence-driven analysis of myocardial T1ρ mapping.

Approach: The automated process combines left ventricular wall segmentation, right ventricular insertion point detection and the creation of a 16-segment American Heart Association model for segmental T1ρ values analysis.

Results: Automated T1ρ mapping showcased strong agreement with manual processing, enhanced with time efficiency.

Impact: Artificial intelligence-driven analysis of myocardial T1-rho mapping exhibits strong agreement with manual processing, bolstered by time efficiency. This approach shows promise for the rapid and non-invasive assessment of heart disease without the need for contrast agents.