Toward Precise Localization of Abnormal Brain Activity: 1D CNN on Single Voxel fMRI Time-Series.

To explore the spatial patterns of the amplitude of low-frequency fluctuation (ALFF) in patients with hepatic encephalopathy (HE) of varying severity and to correlate these fluctuations with clinical markers of HE. This study was approved by the local ethics committee, and written informed consent was obtained from all participants. Twenty-nine patients with HE (15 with overt and 14 with minimal HE) and 17 healthy control subjects underwent resting-state functional magnetic resonance (MR) imaging. The ALFF, an index reflecting the amplitudes of spontaneous brain activity, was compared among patients with overt HE, patients with minimal HE, and control subjects with analysis of variance tests and t tests between each pair. Pearson correlation analysis was performed between the ALFF and the venous blood ammonia level and Child-Pugh score of all patients with HE. Compared with control subjects, patients with overt and minimal HE showed decreased ALFF mainly in regions within the default-mode network (DMN) and increased ALFF in the cerebellum and middle temporal gyrus. Compared with patients with minimal HE, those with overt HE showed decreased ALFF in DMN regions and increased ALFF in the posterior insular cortex (P<.05, corrected for multiple comparisons). Both the venous blood ammonia levels and Child-Pugh scores of individual patients with HE showed negative correlation with ALFF within some DMN regions, whereas they showed positive correlation with ALFF in the posterior insular cortex (P<.05, corrected for multiple comparisons). Patients with HE have diffuse abnormalities in intrinsic brain activity. The levels of decreased ALFF in the DMN and increased ALFF in the posterior insular cortex are dependent on the severity of HE, suggesting continuous impairment of the DMN and a compensatory role of the insula during the progression of HE. Resting-state functional MR imaging with ALFF analysis may be a noninvasive modality with which to detect the progression of HE.

Objective To investigate the characteristics of hypomanic symptoms (HS) to the brain neural activity in the resting-state of the patients with major depressive disorder (MDD). Methods Twenty-three patients from 54 patients with MDD were defined as MDD with HS because their HCL-32 scores were higher than or equal to 12. The rest 31 patients were defined as MDD patients without HS because of the lower scores. After checking the match of the factors such as gender and age et al, there were 19 MDD patients with HS, 19 patients without HS and 19 healthy controls in the final analysis. We analyzed the impact of HS to the regional brain function of the patients with MDD by amplitude of low-frequency fluctuation (ALFF). Results The patients with MDD had abnormal resting-state brain activity in a diffuse brain areas compared to healthy controls by the ALFF analysis. Compared to the patients without HS, the ALFF of the depressed MDD patients with HS were significantly higher in the bilateral orbitofrontal cortex (OFC; left OFC: MNI(x,y,z): - 34,56,- 10; t=5.26; right OFC: MNI(x,y,z): 20, 66, - 2; t=493) and superior frontal cortex (SFC; left SFC: MNI(x,y,z): -4, 66, 8; t=4.12; right SFC: MNI(x,y,z): 14, 56, 18, t=3.97, all P< 0.05); while lower in bilateral thalamus and left fusiform gyrus (P<0.05). Correlation analysis of ALFF at each voxel in the whole brain against the HCL-32 in the patients with MDD revealed significantly positive correlation in OFC (r=0.363, 0.426, 0.405, all P<0.05), negative correlation in left thalamus and left fusiform gyrus (r=- 0.352, P<0.05; r=- 0.468, P<0.01). Conclusions The study suggested that there were differences between the two subgroups patients with MDD defined by the HCL-32. Compared to the MDD without HS patients, MDD with HS patients showed increase ALFF in bilateral frontal cortex (especially the OFC), and decrease ALFF in left thalamus and left fusiform gyrus. These features may be regarded as the biomarker of the MDD patients who were positive to HCL-32 screening. Key words: Major depressive disorder; Hypomanic symptoms; Resting state; Functional magnetic resonance imaging; Amplitude of low-frequency fluctuation

Objective To explore the changes of rest-state functional MRI (fMRI) features in patients with Parkinson's disease. Methods Twelve patients with Parkinson's disease, collected in our hospital from September 2013 to March 2014, and 12 volunteers (controls) were recruited. The fMRI based on blood oxygen level dependent (BOLD) was performed at resting state; the differences of amplitude of low frequency fluctuation (ALFF) and regional homogeneity (ReHo) were statistically analyzed; and the correlations of values of ALFF and ReHo with uniformed Parkinsonism rating scale (UPDRS) scores and Hoehn-Yahr (H-Y) grading were analyzed. Results PD patients exhibited increased ALFF values in the limbic lobe, the right parahippocampal gyrus and the right middle temporal pole, and reduced ALFF values in the left thalamus as compared with the controls. PD patients exhibited increased ReHo values in the superior medial frontal gyrus, the left middle frontal gyrus, brodmann area (BA) 9 and medial frontal gyrus, and reduced ReHo values in the left rolandic, the left superior temporal gyrus and insula lobe as compared with the controls. ALFF values were positively correlated with the UPDRS III scores in the right posterior cerebellum lobe and cerebellar tonsil. ALFF values and H-Y grading were negatively correlated on the left superior frontal gyrus. Negative correlation was noted between the ReHo values and UPDRS III scores in the right superior temporal gyrus and insular lobe. Positive correlation was noted between the ReHo values and H-Y grading in medial frontal gyrus, limbic lobe, anterior cingulate, medial orbitofrontal area, middle superior frontal gyrus, BA32, left superior frontal gyrus and BA10. Conclusion By resting-state fMRI, abnormal brain function in Parkinson's disease was noted; ReHo and ALFF showed abnormal changes in the connection of neuronal networks, and the compensatory response may lead to the increased brain function. Key words: Parkinson’s disease; Functional MR imaging; Regional homogeneity; Amplitude of low frequency fluctuation

Classification of non-small cell lung cancer using one-dimensional convolutional neural network

Objective To investigate the characteristic of regional amplitude of low frequency fluctuation (ALFF) in the brain of adolescents with online game addiction during the resting state. Methods The ALFF method was employed to explore the local features of spontaneous brain activity in adolescents with online game addiction (n=26) and healthy controls (n=23) during resting-state by Resting-State fMRI Data Analysis Toolkit. Results Compared with healthy controls, adolescents with online game addiction showed increased ALFF values in the anterior cingulate cortex (ACC), left middle cingulate cortex (MCC), bilateral caudate, right pallidum, bilateral parahippocampal gyrus, and right insula. It was noteworthy that the ALFF values of the right caudate (r=0.4437, P=0.0232) and right parahippocampal gyrus (r=0.5227, P=0.0062) were positively correlated with the duration of online game addiction. Conclusion The abnormal spontaneous neuronal activity properties during the resting state were revealed in the brain regions of adolescents with online game addiction. And the abnormalities may become more significant as the disease progresses. Key words: Online game addiction; Resting state; Amplitude of low frequency fluctuation

Anemia is one of the most prevalent nutritional and hematological disorders worldwide, characterized by low hemoglobin levels caused by iron deficiency, genetic factors, or chronic diseases. Diagnosis commonly relies on Complete Blood Count (CBC) interpretation, a manual process that is time-consuming and susceptible to human error. This study proposes a novel hybrid framework that integrates Extreme Gradient Boosting (XGBoost) and a One-Dimensional Convolutional Neural Network (1D-CNN) to enhance anemia classification. The methodological novelty lies in employing XGBoost as a feature-augmentation mechanism, where its class-probability outputs are fused with the original CBC features before being processed by the 1D-CNN, enabling richer representation learning compared to conventional single-model approaches. The model was trained and evaluated using a CBC dataset consisting of 364 samples covering four anemia classes (normocytic, microcytic, macrocytic, and normal), with performance assessed through an 80:20 stratified train–test split. Experimental results demonstrate that the proposed XGB–1DCNN model achieves a testing accuracy of 97.26%, precision of 98.68%, recall of 96.46%, and F1-score of 97.48%, outperforming the baseline 1D-CNN model (83.56%). These findings demonstrate that combining ensemble learning and deep learning significantly improves the model’s ability to capture complex nonlinear patterns in CBC data, offering a more reliable solution for AI-based early anemia diagnosis and clinical decision support.

BackgroundConvolutional neural networks (CNNs) have outperformed dermatologists in classifying pigmented skin lesions under artificial conditions. We investigated, for the first time, the performance of three-dimensional (3D) and two-dimensional (2D) CNNs and dermatologists in the early detection of melanoma in a real-world setting. MethodsIn this prospective study, 1690 melanocytic lesions in 143 patients with high-risk criteria for melanoma were evaluated by dermatologists, 2D-FotoFinder-ATBM and 3D-Vectra WB360 total body photography (TBP). Excision was based on the dermatologists' dichotomous decision, an elevated CNN risk score (study-specific malignancy cut-off: FotoFinder >0.5, Vectra >5.0) and/or the second dermatologist's assessment with CNN support. The diagnostic accuracy of the 2D and 3D CNN classification was compared with that of the dermatologists and the augmented intelligence based on histopathology and dermatologists’ assessment. Secondary end-points included reproducibility of risk scores and naevus counts per patient by medical staff (gold standard) compared to automated 3D and 2D TBP CNN counts. ResultsThe sensitivity, specificity, and receiver operating characteristics area under the curve (ROC-AUC) for risk-score-assessments compared to histopathology of 3D-CNN with 95% confidence intervals (CI) were 90.0%, 64.6% and 0.92 (CI 0.85–1.00), respectively. While dermatologists and augmented intelligence achieved the same sensitivity (90%) and comparable classification ROC-AUC (0.91 [CI 0.80–1.00], 0.88 [CI 0.77–1.00]) with 3D-CNN, their specificity was superior (92.3% and 86.2%, respectively). The 2D-CNN (sensitivity: 70%, specificity: 40%, ROC-AUC: 0.68 [CI 0.46–0.90]) was outperformed by 3D CNN and dermatologists. The 3D-CNN showed a higher correlation coefficient for repeated measurements of 246 lesions (R = 0.89) than the 2D-CNN (R = 0.79). The mean naevus count per patient varied significantly (gold standard: 210 lesions; 3D-CNN: 469; 2D-CNN: 1324; p < 0.0001). ConclusionsOur study emphasises the importance of validating the classification of CNNs in real life. The novel 3D-CNN device outperformed the 2D-CNN and achieved comparable sensitivity with dermatologists. The low specificity of CNNs and the lack of automated counting of TBP nevi currently limit the use of augmented intelligence in clinical practice.

Following artificial intelligence implementation in computer vision fields, especially deep learning, many Computer-Aided Diagnosis (CAD) tools are proposed to help to detect lung cancer by the scoring system or by identifying the characteristics of nodules. However, lung cancer is a clinical diagnosis which does not provide detailed information needed by radiologists and clinician to prevent unnecessary invasive diagnostic procedures compared to lung nodule texture detection and classification. Hence, to answer this problem, this research explores the steps needed to implement 3D CNN on raw thorax CT scan datasets and usage of RetinaNet 3D + Inception 3D with transfer learning. The 3D CNN CAD tools can improve the speed, performance, and ability to detect lung nodule texture instead of malignancy status done by previous studies. This study implements 3D CNN on Moscow private datasets acquired from NVIDIA Asia Pacific. The proposed method of data conversion can minimize information loss from raw data to 3D CNN input data. After 100 epoch training, the researchers conclude that the best-proposed model (3D CNN with transfer learning of pretrained LIDC public datasets weight) provides 22.86% of mean average precision (mAP) detection capability and 70.36% of Area Under the Curve (AUC) in Moscow private dataset lung texture detection tasks. It outperforms non-transfer learning 3D CNN model (trained from scratch) and 3D CNN with transfer learning of pre-trained ImageNet weight.

We present a hardware-friendly deep learning architecture with one-dimensional convolutional neural networks (1D CNN) for fast analyzing fluorescence lifetime imaging (FLIM) data. A 1D CNN shows unparalleled advantages; they are more straightforward, quicker to train, and faster than high dimensional CNNs. 1D CNNs can be easily applied to multi-exponential fluorescence decay models. Compared with traditional least-square methods, superior performances of 1D CNNs on fluorescence lifetime image reconstruction have been validated using simulated data. We also employ the proposed 1D CNN to analyze two-photon FLIM images of functionalized gold nanoprobes in Hek293 and human prostate cancer cells. The results further demonstrate that 1D CNNs are fast and can accurately extract lifetime parameters from fluorescence signals. Our study shows that 1D CNNs have great potential in various real-time FLIM applications.

Frequency-specific alterations of the frontal-cerebellar circuit in first-episode, drug-naive somatization disorder

Neuroimaging studies have revealed that low-back pain (LBP) alters spatiotemporal dynamics of the blood oxygen level-dependent signal in response to persistent noxious stimulus. This study aimed to investigate changes in spontaneous neural activity of various brain regions in acute LBP using resting-state functional magnetic resonance imaging and amplitude of low-frequency fluctuation (ALFF). Twelve healthy subjects underwent two separate resting-state functional magnetic resonance imaging scans at health status as baseline and after intramuscular injection of hypertonic saline (0.5 mL, 5%) into the back muscles to induce acute LBP. Compared with baseline, acute LBP showed decreased ALFF in the right posterior cingulate cortex/precuneus and left primary somatosensory cortex (S1) but increased ALFF in the right medial prefrontal cortex, right middle temporal gyrus, bilateral inferior temporal gyrus, bilateral insula, right anterior cingulate cortex, and left cerebellum. In addition, significant negative correlations were observed between visual analog scale scores and ALFF of the bilateral medial prefrontal cortex, left inferior frontal gyrus, left S1, right anterior cingulate cortex, and left middle temporal gyrus. These findings suggest that abnormally spontaneous neural activity involving some brain regions are responsible for sensory, affective, and cognitive functions, which may be implicated in the underlying pathophysiology of acute LBP.

To analyze the spontaneous brain activity patterns in patients with end-stage renal disease (ESRD) undergoing peritoneal dialysis (PD) by using resting-state functional magnetic resonance (MR) imaging with an amplitude of low-frequency fluctuations (ALFF) algorithm. This study received institutional review board approval, and all subjects gave informed consent. Forty-four patients with ESRD, 24 of whom were undergoing PD (PD group; eight women; mean age, 34 years ± 8) and 20 who were not undergoing PD or hemodialysis (nondialysis group; six women; mean age, 37 years ± 9) and 24 healthy control subjects (eight women; mean age, 32 years ± 9 years) were included. All subjects underwent neuropsychologic tests, and patients with ESRD underwent laboratory testing. ALFF values were compared among the three groups. The relationship between ALFF values and clinical markers was investigated by using multiple regression analysis. Patients in both the PD and nondialysis groups showed lower ALFF values in default mode network regions than did healthy control subjects (P < .01, false discovery rate corrected). Patients in the PD group showed lower ALFF values than did those in the nondialysis group in the left superior parietal lobe (1.51 ± 0.21 vs 2.01 ± 0.40), left inferior parietal lobe (0.99 ± 0.16 vs 1.13 ± 0.22) and left precuneus (1.45 ± 0.39 vs 1.77 ± 0.41) (P < .01, corrected with simulation software). In patients in the PD group, neuropsychologic test scores correlated with ALFF values of the middle temporal gyrus and the parietal and occipital lobe, serum urea and creatinine levels negatively correlated with ALFF in some default mode network regions, and hemoglobin positively correlated with ALFF in the bilateral precuneus, precentral, and supplementary motor areas (P < .01 corrected). Patients with ESRD who were undergoing PD showed more severe spontaneous brain activity abnormalities that correlate with cognitive impairments than did patients who were not undergoing dialysis. Elevated serum urea, creatinine, and lowered hemoglobin levels affect spontaneous brain activity in patients with ESRD.

Millions of people have been affected by coronavirus disease 2019 (COVID-19), which has caused millions of deaths around the world. Artificial intelligence (AI) plays an increasing role in all areas of patient care, including prognostics. This paper proposes a novel predictive model based on one dimensional convolutional neural networks (1D CNN) to use clinical variables in predicting the survival outcome of COVID-19 patients. We have considered two scenarios for survival analysis, 1) uni-variate analysis using the Log-rank test and Kaplan-Meier estimator and 2) combining all clinical variables ([Formula: see text]=44) for predicting the short-term from long-term survival. We considered the random forest (RF) model as a baseline model, comparing to our proposed 1D CNN in predicting survival groups. Our experiments using the univariate analysis show that nine clinical variables are significantly associated with the survival outcome with corrected p < 0.05. Our approach of 1D CNN shows a significant improvement in performance metrics compared to the RF and the state-of-the-art techniques (i.e., 1D CNN) in predicting the survival group of patients with COVID-19. Our model has been tested using clinical variables, where the performance is found promising. The 1D CNN model could be a useful tool for detecting the risk of mortality and developing treatment plans in a timely manner. The findings indicate that using both Heparin and Exnox for treatment is typically the most useful factor in predicting a patient's chances of survival from COVID-19. Moreover, our predictive model shows that the combination of AI and clinical data can be applied to point-of-care services through fast-learning healthcare systems.

This study investigated the characteristic changes of brain function in children with Tourette syndrome (TS) based on resting-state functional MRI (rs-fMRI) and their correlation with the Yale Global Tic Severity Scale (YGTSS) scores. A total of 19 children diagnosed as TS at the Department of Pediatrics in China-Japan Friendship Hospital from October 2021 to April 2023 were enrolled prospectively as the TS group, and 20 healthy children matched for age and gender were field recruited from the community as the control group. All participants underwent head rs-fMRI scans, and regional homogeneity (ReHo) and amplitude of low-frequency fluctuations (ALFF) values were calculated. YGTSS scores were assessed for the TS group. Pearson correlation analysis and Spearman correlation analysis were used to evaluate the correlation between ReHo, ALFF and YGTSS scores. Receiver operating characteristic (ROC) curves were plotted and area under the curve (AUC) was calculated to evaluate the efficacy of brain function parameters related to YGTSS score in diagnosing TS. The TS group included 19 patients, and aged (8.5±2.6) years, including 13 males (68.4%). The control group included 20 children, and aged (5.5±1.8) years, including 7 males (35.0%). In the TS group, the ReHo values in the left lentiform nucleus gray matter, left occipital lobe fusiform gyrus, right occipital lobe medial gyrus, and right frontal lobe paracentral lobule white matter were lower than those in the control group. The ReHo values in the pons and right cerebellar tonsil region, as well as the ALFF values in the right lingual gyrus, were higher than those in the control group (all P<0.01). The ReHo values in the right frontal lobe paracentral lobule, right occipital lobe medial gyrus, and left occipital lobe fusiform gyrus were positively correlated with vocal frequency (r=0.466, 0.576, 0.491) and intensity (r=0.498, 0.593, 0.609); the ReHo values in the right occipital lobe medial gyruss (r=0.615) and left occipital lobe fusiform gyrus (r=0.661) were positively correlated with the number of vocal tics (all P<0.05). The AUC (95%CI) for discriminating TS based on the ReHo values in the right frontal lobe paracentral lobule, right occipital lobe medial gyrus, and left occipital lobe fusiform gyrus were 0.890 (0.774-1.000), 0.863 (0.745-0.982), and 0.858 (0.760-0.988), respectively, with no statistically significant differences in the efficacy of these 3 ReHo values for discriminating TS (all P>0.05). Children with TS have characteristic brain function changes, and there is a correlation between the characteristic brain regions of MRI brain function changes and vocal tic symptoms.

Early Prediction of Drug-Resistant Epilepsy using Clinical and EEG Features Based on Convolutional Neural Network