Heterogeneous multi-modal medical imaging data need to be properly handled in classification. Currently, generating models using multi-modal imaging data has become a common practice and greatly benefits the brain disorder diagnosis, which also holds considerable clinical potential. Although the majority of classification studies focus on using features from single modality, there is substantial evidence suggesting that classification based on multi-modal features is on upward trend. Hence, effective integration of heterogeneous data is in urgent demand. Here, we proposed a multi-kernel SVM for schizophrenia classification with nested 10-fold cross validation, which could integrate multi-modal data using the subspace similarity of the decomposed components in each MRI modality. To validate the effectiveness of the proposed method, we performed experiments on two independent datasets with three different modalities to classify schizophrenia patients and healthy controls. Specifically, multi-modal fusion method was first applied on preprocessed fMRI, DTI and sMRI data to generate components that could be used for classification. Then multi-kernel SVM models were trained on the selected component features using subspace similarity measures, and were tested on independent validation data across sites. The results on both datasets demonstrated that our method achieved accuracies of 87.6% and 79.9% separately on two datasets when combining all three modalities, which outperformed alternative methods and might provide potential biomarkers for cross-site classification and co-varying components among different modalities.
The cognitive deficits of schizophrenia are largely resistant to current treatment, and are thus a life-long burden to patients. The MATRICS consensus cognitive battery (MCCB) provides a reliable and valid assessment of cognition across a comprehensive set of cognitive domains for schizophrenia. In resting-state fMRI, functional connectivity associated with MCCB has not yet been examined. In this paper, the interrelationships between MCCB and the abnormalities seen in two types of functional measures from resting-state fMRI—fractional amplitude of low frequency fluctuations (fALFF) and functional network connectivity (FNC) maps were investigated in data from 47 schizophrenia patients and 50 age-matched healthy controls. First, the fALFF maps were generated and decomposed by independent component analysis (ICA), and then the component showing the highest correlation with MCCB composite scores was selected. Second, the whole brain was separated into functional networks by group ICA, and the FNC maps were calculated. The FNC strengths with most significant correlations with MCCB were displayed and spatially overlapped with the fALFF component of interest. It demonstrated increased cognitive performance associated with higher fALFF values (intensity of regional spontaneous brain activity) in prefrontal regions, inferior parietal lobe (IPL) but lower ALFF values in thalamus, striatum, and superior temporal gyrus (STG). Interestingly, the FNC showing significant correlations with MCCB were in well agreement with the activated regions with highest z-values in fALFF component. Our results support the view that functional deficits in distributed cortico-striato-thalamic circuits and inferior parietal lobe may account for several aspects of cognitive impairment in schizophrenia.
Proc. SPIE. 7626, Medical Imaging 2010: Biomedical Applications in Molecular, Structural, and Functional Imaging
KEYWORDS: Image fusion, Independent component analysis, Data modeling, Neuroimaging, Functional magnetic resonance imaging, Simulation of CCA and DLA aggregates, Current controlled current source, Data fusion, Brain imaging, Brain
On joint analysis of multi-task brain imaging data sets, a variety of multivariate methods have shown their
strengths and been applied to achieve different purposes based on their respective assumptions. In this paper, we provide
a comprehensive review on optimization assumptions of six data fusion models, including 1) four blind methods: joint
independent component analysis (jICA), multimodal canonical correlation analysis (mCCA), CCA on blind source
separation (sCCA) and partial least squares (PLS); 2) two semi-blind methods: parallel ICA and coefficient-constrained
ICA (CC-ICA). We also propose a novel model for joint blind source separation (BSS) of two datasets using a
combination of sCCA and jICA, i.e., 'CCA+ICA', which, compared with other joint BSS methods, can achieve higher
decomposition accuracy as well as the correct automatic source link. Applications of the proposed model to real multitask
fMRI data are compared to joint ICA and mCCA; CCA+ICA further shows its advantages in capturing both shared
and distinct information, differentiating groups, and interpreting duration of illness in schizophrenia patients, hence
promising applicability to a wide variety of medical imaging problems.
The Infrared thermal imaging systems has developments advance rapidly during the development of the research and the
manufacture technical. And its applied field has going deep into the astronautics, industry, agriculture, medical, traffic
and other fields from the national defense and military appliance. Especially in the application of the military, it has
come into being a specialty IR System Engineering field. But in many important applications, the lens calibre of the IR
thermal imaging systems often be made very large to advance the SNR of the systems. This increased the weight and the
research cost of the whole system very much. Many research indicated that the main factor to affect the image quality of
the IR systems is the fixed pattern noise (FPN) or spatial non-uniformity under the actual technical and manufacture
level. If we using the effective dynamic self-adaptive non-uniformity correction algorithms for the IR system, and use
the image enhancement technology simultaneity. We can advance the imaging quality greatly. With this plan, the
correction image we got with large F number can receive the level that uncorrected image with 1 or 2 smaller F number.
It means the lens calibre of the system will be reduced effectively. And the weight, the cubage and the research cost of
the system will be reduced greatly. It will have most important value in the applied of the actual engineering.
A recently developed scene-based nonuniformity correction algorithm for focal plane array (FPA) sensors named Crossing
Path Scene-Based Algorithm (CPSBA) is present. The goal of this thesis is to design and evaluate scene-based nonuniformity
correction algorithms that are able to suppress fixed pattern noise without need for external hardware such as temperature
reference equipment. In particular, algorithms should be able to accurately estimate motion between images and use this
knowledge to improve performance. The algorithms have been tested by using real image data from existing infrared
imaging systems with good results.
Nonuniformity correction (NUC) is a critical task for achieving higher performances in modern infrared imaging systems. The striping fixed pattern noise produced by the scanning-type infrared imaging system can hardly be removed clearly by many scene-based non-uniformity correction methods, which can work effectively for staring focal plane arrays (FPA). We proposed an improved nonuniformity algorithm that corrects the aggregate nonuniformity by two steps for the infrared line scanners (IRLS). The novel contribution in our approach is the integration of local constant statistics (LCS) constraint and neural networks. First, the nonuniformity due to the readout electronics is corrected by treating every row of pixels as one channel and normalizing the channel outputs so that each channel produces pixels with the same mean and standard deviation as median value of the local channels statistics. Second, for IRLS every row is generated by pushbrooming one detector on line sensors, we presume each detector has one neuron with a weight and an offset as correction parameters, which can update column by column recursively at Least Mean Square sense. A one-dimensional median filter is used to produce ideal output of linear neural network and some optimization strategies are added to increase the robustness of learning process. Applications to both simulated and real infrared images demonstrated that this algorithm is self-adaptive and able to complete NUC by only one frames. If the nonuniformity is not so severe then only the first step can obtain a good correction result. Combination of two steps can achieve a higher correction level and remove stripe pattern noise clearly.
The Infrared Focal Plane Array (IRFPA) is a key part of modern infrared system and thermal imaging system. Imaging with IRFPA is future development direction of infrared and thermal imaging system. However the most difficult problem associated with the IRFPA is intrinsic spatial photo-response non-uniformity. Although two- point or multi-point correction algorithms may correct the non-uniformity of IRFPAs, they can be limited by pixel nonlinearities and instabilities. So adaptive non-uniformity correction techniques are needed. Many researchers develop the methods of real-time correction based the scene being viewed. In this paper, we introduced eight scene-based real-time correction methods, such as: the approach of the whole frame correction, Kalman filtering correction, adaptive filtering correction, trace along trajectory correction, based on the scene moving analysis correction, the wavelet transform is used for design the low-pass filter correction, both of the high-pass filter and the artificial neural network correction, based on the application of wavelet filter, video sequences and registration, orthogonal least squares correction, etc. All these non-uniformity correction algorithms are deeply explored, and non-uniformity correction simulation experiments are carried. Finally we compare these algorithms with two-point correction algorithm.
The performance of direct viewing low light level (LLL) imaging system is mainly determined by three factors: photons noise, MTF of optical system(OS) and human eyes characteristic. And the image detecting theory which denotes the optimal performance of imaging system has been a positive impetus for the development of the LLL imaging and night vision technique. The system minimum resolvable angle was traditionally used to estimate the image detecting performance which is mainly determined by photons noise at low target illuminance and by MTF at high target illuminance. This criterion can represent the system performance on the whole; however, assuming the signal to noise ratio (SNR)of the image and MTF of OS uncorrelative, is theoretically not complete, since the two factors interrelate actually. From the viewpoint of signal response, the MRC (minimum resolvable contrast) model of the ideal direct viewing LLL imaging system was deduced on the basis of human eyes characteristic. It is a more comprehensive evaluation method for imaging system performance, and can combine with the forecasting model of operating distance to analyze the general performance of night vision system. In conclusion, the relationship and the difference between the MRC model and the traditional detecting equation were investigated.