Ischemic and haemorrhagic stroke risk estimation using a machine-learning-based retinal image analysis.

Background: Stroke is the second leading cause of death worldwide, causing a considerable disease burden. Ischemic stroke is more frequent, but haemorrhagic stroke is responsible for more deaths. The clinical management and treatment are different, and it is advantageous to classify their risk as early as possible for disease prevention. Furthermore, retinal characteristics have been associated with stroke and can be used for stroke risk estimation. This study investigated machine learning approaches to retinal images for risk estimation and classification of ischemic and haemorrhagic stroke. Study design: A case-control study was conducted in the Shenzhen Traditional Chinese Medicine Hospital. According to the computerized tomography scan (CT) or magnetic resonance imaging (MRI) results, stroke patients were classified as either ischemic or hemorrhage stroke. In addition, a control group was formed using non-stroke patients from the hospital and healthy individuals from the community. Baseline demographic and medical information was collected from participants' hospital medical records. Retinal images of both eyes of each participant were taken within 2 weeks of admission. Classification models using a machine-learning approach were developed. A 10-fold cross-validation method was used to validate the results. Results: 711 patients were included, with 145 ischemic stroke patients, 86 haemorrhagic stroke patients, and 480 controls. Based on 10-fold cross-validation, the ischemic stroke risk estimation has a sensitivity and a specificity of 91.0% and 94.8%, respectively. The area under the ROC curve for ischemic stroke is 0.929 (95% CI 0.900 to 0.958). The haemorrhagic stroke risk estimation has a sensitivity and a specificity of 93.0% and 97.1%, respectively. The area under the ROC curve is 0.951 (95% CI 0.918 to 0.983). Conclusion: A fast and fully automatic method can be used for stroke subtype risk assessment and classification based on fundus photographs alone.

Risk Assessment of CHD Using Retinal Images with Machine Learning Approaches for People with Cardiometabolic Disorders.

BACKGROUND: Coronary heart disease (CHD) is the leading cause of death worldwide, constituting a growing health and social burden. People with cardiometabolic disorders are more likely to develop CHD. Retinal image analysis is a novel and noninvasive method to assess microvascular function. We aim to investigate whether retinal images can be used for CHD risk estimation for people with cardiometabolic disorders. METHODS: We have conducted a case-control study at Shenzhen Traditional Chinese Medicine Hospital, where 188 CHD patients and 128 controls with cardiometabolic disorders were recruited. Retinal images were captured within two weeks of admission. The retinal characteristics were estimated by the automatic retinal imaging analysis (ARIA) algorithm. Risk estimation models were established for CHD patients using machine learning approaches. We divided CHD patients into a diabetes group and a non-diabetes group for sensitivity analysis. A ten-fold cross-validation method was used to validate the results. RESULTS: The sensitivity and specificity were 81.3% and 88.3%, respectively, with an accuracy of 85.4% for CHD risk estimation. The risk estimation model for CHD with diabetes performed better than the model for CHD without diabetes. CONCLUSIONS: The ARIA algorithm can be used as a risk assessment tool for CHD for people with cardiometabolic disorders.

Prevalence of diabetic retinopathy within a national diabetic retinopathy screening service.

AIMS: Determine the prevalence and severity of diabetic retinopathy (DR) and risk factors in a large community based screening programme, in order to accurately estimate the future burden of this specific and debilitating complication of diabetes. METHODS: A cross-sectional analysis of 91,393 persons with diabetes, 5003 type 1 diabetes and 86,390 type 2 diabetes, at their first screening by the community based National Diabetic Retinopathy Screening Service for Wales from 2005 to 2009. Image capture used 2×45° digital images per eye following mydriasis, classified by qualified retinal graders with final grading based on the worst eye. RESULTS: The prevalence of any DR and sight-threatening DR in those with type 1 diabetes was 56.0% and 11.2%, respectively, and in type 2 diabetes was 30.3% and 2.9%, respectively. The presence of DR, non-sight-threatening and sight-threatening, was strongly associated with increasing duration of diabetes for either type 1 or type 2 diabetes and also associated with insulin therapy in those with type 2 diabetes. CONCLUSIONS: Prevalence of DR within the largest reported community-based, quality assured, DR screening programme, was higher in persons with type 1 diabetes; however, the major burden is represented by type 2 diabetes which is 94% of the screened population.

The application of nonlinear Dynamic Causal Modelling for fMRI in subjects at high genetic risk of schizophrenia.

Nonlinear Dynamic Causal Modelling (DCM) for fMRI provides computational modelling of gating mechanisms at the neuronal population level. It allows for estimations of connection strengths with nonlinear modulation within task-dependent networks. This paper presents an application of nonlinear DCM in subjects at high familial risk of schizophrenia performing the Hayling Sentence Completion Task (HSCT). We analysed scans of 19 healthy controls and 46 subjects at high familial risk of schizophrenia, which included 26 high risk subjects without psychotic symptoms and 20 subjects with psychotic symptoms. The activity-dependent network consists of the intra parietal cortex (IPS), inferior frontal gyrus (IFG), middle temporal gyrus (MTG), anterior cingulate cortex (ACC) and the mediodorsal (MD) thalamus. The connections between the MD thalamus and the IFG were gated by the MD thalamus. We used DCM to investigate altered connection strength of these connections. Bayesian Model Selection (BMS) at the group and family level was used to compare the optimal bilinear and nonlinear models. Bayesian Model Averaging (BMA) was used to assess the connection strengths with the gating from the MD thalamus and the IFG. The nonlinear models provided the better explanation of the data. Furthermore, the BMA analysis showed significantly lower connection strength of the thalamocortical connection with nonlinear modulation from the MD thalamus in high risk subjects with psychotic symptoms and those who subsequently developed schizophrenia. These findings demonstrate that nonlinear DCM provides a method to investigate altered connectivity at the level of neural circuits. The reduced connection strength with thalamic gating may be a neurobiomarker implicated in the development of psychotic symptoms. This study suggests that nonlinear DCM could lead to new insights into functional and effective dysconnection at the network level in subjects at high familial risk.

Impact of cannabis use on thalamic volume in people at familial high risk of schizophrenia.

BACKGROUND: No longitudinal study has yet examined the association between substance use and brain volume changes in a population at high risk of schizophrenia. AIMS: To examine the effects of cannabis on longitudinal thalamus and amygdala-hippocampal complex volumes within a population at high risk of schizophrenia. METHOD: Magnetic resonance imaging scans were obtained from individuals at high genetic risk of schizophrenia at the point of entry to the Edinburgh High-Risk Study (EHRS) and approximately 2 years later. Differential thalamic and amygdala-hippocampal complex volume change in high-risk individuals exposed (n = 25) and not exposed (n = 32) to cannabis in the intervening period was investigated using repeated-measures analysis of variance. RESULTS: Cannabis exposure was associated with bilateral thalamic volume loss. This effect was significant on the left (F = 4.47, P = 0.04) and highly significant on the right (F= 7.66, P= 0.008). These results remained significant when individuals using other illicit drugs were removed from the analysis. CONCLUSIONS: These are the first longitudinal data to demonstrate an association between thalamic volume loss and exposure to cannabis in currently unaffected people at familial high risk of developing schizophrenia. This observation may be important in understanding the link between cannabis exposure and the subsequent development of schizophrenia.

Correlations between fMRI activation and individual psychotic symptoms in un-medicated subjects at high genetic risk of schizophrenia.

BACKGROUND: It has been proposed that different types of psychopathology in schizophrenia may reflect distinguishable pathological processes. In the current study we aimed to address such associations in the absence of confounders such as medication and disease chronicity by examining specific relationships between fMRI activation and individual symptom severity scores in un-medicated subjects at high genetic risk of schizophrenia. METHODS: Associations were examined across two functional imaging paradigms: the Hayling sentence completion task, and an encoding/retrieval task, comprising encoding (at word classification) and retrieval (old word/new word judgement). Symptom severity was assessed using the positive and negative syndrome scale (PANSS). Items examined were hallucinations, delusions, and suspiciousness/persecution. RESULTS: Associations were seen in the anterior middle temporal gyrus in relation to hallucination scores during the sentence completion task, and in the medial temporal lobe in association with suspiciousness/persecution scores in the encoding/retrieval task. Cerebellar activation was associated with delusions and suspiciousness/persecution scores across both tasks with differing patterns of laterality. CONCLUSION: These results support a role for the lateral temporal cortex in hallucinations and medial temporal lobe in positive psychotic symptoms. They also highlight the potential role of the cerebellum in the formation of delusions. That the current results are seen in un-medicated high risk subjects indicates these associations are not specific to the established illness and are not related to medication effects.