Similarity classifier with generalized mean applied to medical data

A new approach based on fuzzy similarity was presented for the detection of erythemato-squamous diseases, diabetes, liver disorders, breast cancer and thyroid. The domain contained records of patients with known diagnoses. The results were very promising with all data sets and some conclusions can be drawn that a fuzzy similarity model can be used for the diagnosis of patients taking into consideration the error rate. A fuzzy similarity classifier was used to detect the six erythemato-squamous diseases when 34 features defining six disease indications were used as inputs. The results confirmed that the proposed model has potential in detecting erythemato-squamous diseases. The fuzzy similarity model achieved accuracy rates (over 97%) which were higher than that of the stand-alone neural network model or the ANFIS model suggested in [E.D. Ubeyli, I. Güler, Comput. Biol. Med. 35(5) (2005) 421-433]. With PIMA Indian diabetes, the detection model has an error rate of about 25% which is much better than the overall rate of 33% for diabetes. The model was also tested with other data sets: thyroid and two breast cancer data sets where the average detection accuracy was over 96% for all cases, which is quite good. Also, the liver disorder data set gave promising results.     

Automatic detection of erthemato-squamous diseases using adaptive neuro- fuzzy inference systems

A new approach based on adaptive neuro-fuzzy inference system (ANFIS) was presented for detection of erythemato-squamous diseases. The domain contained records of patients with known diagnosis. Given a training set of such records, the ANFIS classifiers learned how to differentiate a new case in the domain. The six ANFIS classifiers were used to detect the six erythemato-squamous diseases when 34 features defining six disease indications were used as inputs. To improve diagnostic accuracy, the seventh ANFIS classifier (combining ANFIS) was trained using the outputs of the six ANFIS classifiers as input data. The proposed ANFIS model combined the neural network adaptive capabilities and the fuzzy logic qualitative approach. Some conclusions concerning the impacts of features on the detection of erythemato-squamous diseases were obtained through analysis of the ANFIS. The performances of the ANFIS model were evaluated in terms of training performances and classification accuracies and the results confirmed that the proposed ANFIS model has some potential in detecting the erythemato-squamous diseases. The ANFIS model achieved accuracy rates which were higher than that of the stand-alone neural network model.     

IntelliHealth: A medical decision support application using a novel weighted multi-layer classifier ensemble framework

Accuracy plays a vital role in the medical field as it concerns with the life of an individual. Extensive research has been conducted on disease classification and prediction using machine learning techniques. However, there is no agreement on which classifier produces the best results. A specific classifier may be better than others for a specific dataset, but another classifier could perform better for some other dataset. Ensemble of classifiers has been proved to be an effective way to improve classification accuracy. In this research we present an ensemble framework with multi-layer classification using enhanced bagging and optimized weighting. The proposed model called “HM-BagMoov” overcomes the limitations of conventional performance bottlenecks by utilizing an ensemble of seven heterogeneous classifiers. The framework is evaluated on five different heart disease datasets, four breast cancer datasets, two diabetes datasets, two liver disease datasets and one hepatitis dataset obtained from public repositories. The analysis of the results show that ensemble framework achieved the highest accuracy, sensitivity and F-Measure when compared with individual classifiers for all the diseases. In addition to this, the ensemble framework also achieved the highest accuracy when compared with the state of the art techniques. An application named “IntelliHealth” is also developed based on proposed model that may be used by hospitals/doctors for diagnostic advice.     

A systematic review of the precision and accuracy of dose measurements in photon radiotherapy using polymer and Fricke MRI gel dosimetry

The purpose of this work is to undertake a critical appraisal of the evidence in the published literature concerning the basic parameters of accuracy and precision associated with the use of Fricke and polymer gels (in conjunction with MR imaging) as radiation dosimeters in photon radiotherapy, condensing and analysing the body of published information (to the end of April 2002). A systematic review was undertaken addressing specific issues of precision and accuracy asking defined questions of the published literature. Accuracy and precision in relation to gel dosimetry were defined. Information was obtained from published, peer-reviewed journals. A defined search strategy utilizing MeSH headings and keywords, with extensive use of cross-referencing, identified 115 references dealing with gel dosimetry. Exclusion criteria were used to select only data from publications which would give unequivocal evidence. For accuracy, results had to be compared with an ionization chamber as gold standard and all gel samples had to be manufactured in the same batch. For precision, in addition to gels being from the same batch, samples must all have been irradiated at the same time and scanned simultaneously (or within a short time frame). Many results were found demonstrating 'dose mapping' examples using gels. However, there were very few publications containing firm evidence of precision and accuracy. There was no evidence which fulfilled our criteria about accuracy or precision using Fricke gels. For polymer gels only one paper was found for accuracy (4% (Low et al 1999 Med. Phys. 26 1542-51)) and precision (1.7% (Baldock et al 1998 Phys. Med. Biol. 43695-702)); however, both were carried out at only one dose level. If the exclusion criteria were relaxed to include accuracy results comparing gel to a non gold standard dosimeter (e.g. TLD), results give a median accuracy of 10% (range 8-23.5%) for polymer gel (Cosgrove et al 2000 Phys. Med. Biol. 45 1195-210, De Deene et al 1998 Radiother: Oncol. 48 283-91, Farajollahi et al 2000 Br. J. Radiol. 72 1085-92, McJury et al 1999b Phys. Med. Biol. 44 2431-44, Murphy et al 2000b Phys. Med. Biol. 45 835-45, Oldham et al 2001 Med. Phys. 28 1436-45) and 5% for Fricke gel (Chan and Ayyangar 1995b Med. Phys. 22 1171-5). Evidence also points to accuracy worsening at lower dose levels for both gels. The precision data should be viewed with caution as repeated MR measurements were not performed with the same samples. The only precision data for Fricke gels was 1.5% (Johansson Back et al 1998 Phys. Med. Biol. 43 261-76), but for zero dose. In conclusion, despite the amount of published data, sparse research has been undertaken which provides clear evidence of the accuracy and precision for both gels. That which has been published has used higher doses than would be routine in radiotherapy. The basic radiation dosimeter qualities of accuracy and precision have yet to be fully quantified for polymer and Fricke gels at clinically relevant dose levels.     

Multiple template-based fluoroscopic tracking of lung tumor mass without implanted fiducial markers

Precise lung tumor localization in real time is particularly important for some motion management techniques, such as respiratory gating or beam tracking with a dynamic multi-leaf collimator, due to the reduced clinical tumor volume (CTV) to planning target volume (PTV) margin and/or the escalated dose. There might be large uncertainties in deriving tumor position from external respiratory surrogates. While tracking implanted fiducial markers has sufficient accuracy, this procedure may not be widely accepted due to the risk of pneumothorax. Previously, we have developed a technique to generate gating signals from fluoroscopic images without implanted fiducial markers using a template matching method (Berbeco et al 2005 Phys. Med. Biol. 50 4481-90, Cui et al 2007 Phys. Med. Biol. 52 741-55). In this paper, we present an extension of this method to multiple-template matching for directly tracking the lung tumor mass in fluoroscopy video. The basic idea is as follows: (i) during the patient setup session, a pair of orthogonal fluoroscopic image sequences are taken and processed off-line to generate a set of reference templates that correspond to different breathing phases and tumor positions; (ii) during treatment delivery, fluoroscopic images are continuously acquired and processed; (iii) the similarity between each reference template and the processed incoming image is calculated; (iv) the tumor position in the incoming image is then estimated by combining the tumor centroid coordinates in reference templates with proper weights based on the measured similarities. With different handling of image processing and similarity calculation, two such multiple-template tracking techniques have been developed: one based on motion-enhanced templates and Pearson's correlation score while the other based on eigen templates and mean-squared error. The developed techniques have been tested on six sequences of fluoroscopic images from six lung cancer patients against the reference tumor positions manually determined by a radiation oncologist. The tumor centroid coordinates automatically detected using both methods agree well with the manually marked reference locations. The eigenspace tracking method performs slightly better than the motion-enhanced method, with average localization errors less than 2 pixels (1 mm) and the error at a 95% confidence level of about 2-4 pixels (1-2 mm). This work demonstrates the feasibility of direct tracking of a lung tumor mass in fluoroscopic images without implanted fiducial markers using multiple reference templates.     

Clinical validation of the LKB model and parameter sets for predicting radiation-induced pneumonitis from breast cancer radiotherapy

The choice of the appropriate model and parameter set in determining the relation between the incidence of radiation pneumonitis and dose distribution in the lung is of great importance, especially in the case of breast radiotherapy where the observed incidence is fairly low. From our previous study based on 150 breast cancer patients, where the fits of dose-volume models to clinical data were estimated (Tsougos et al 2005 Evaluation of dose-response models and parameters predicting radiation induced pneumonitis using clinical data from breast cancer radiotherapy Phys. Med. Biol. 50 3535-54), one could get the impression that the relative seriality is significantly better than the LKB NTCP model. However, the estimation of the different NTCP models was based on their goodness-of-fit on clinical data, using various sets of published parameters from other groups, and this fact may provisionally justify the results. Hence, we sought to investigate further the LKB model, by applying different published parameter sets for the very same group of patients, in order to be able to compare the results. It was shown that, depending on the parameter set applied, the LKB model is able to predict the incidence of radiation pneumonitis with acceptable accuracy, especially when implemented on a sub-group of patients (120) receiving [see text]|EUD higher than 8 Gy. In conclusion, the goodness-of-fit of a certain radiobiological model on a given clinical case is closely related to the selection of the proper scoring criteria and parameter set as well as to the compatibility of the clinical case from which the data were derived.     

Comparison of Shallow and Deep Learning Methods on Classifying the Regional Pattern of Diffuse Lung Disease

This study aimed to compare shallow and deep learning of classifying the patterns of interstitial lung diseases (ILDs). Using high-resolution computed tomography images, two experienced radiologists marked 1200 regions of interest (ROIs), in which 600 ROIs were each acquired using a GE or Siemens scanner and each group of 600 ROIs consisted of 100 ROIs for subregions that included normal and five regional pulmonary disease patterns (ground-glass opacity, consolidation, reticular opacity, emphysema, and honeycombing). We employed the convolution neural network (CNN) with six learnable layers that consisted of four convolution layers and two fully connected layers. The classification results were compared with the results classified by a shallow learning of a support vector machine (SVM). The CNN classifier showed significantly better performance for accuracy compared with that of the SVM classifier by 6–9%. As the convolution layer increases, the classification accuracy of the CNN showed better performance from 81.27 to 95.12%. Especially in the cases showing pathological ambiguity such as between normal and emphysema cases or between honeycombing and reticular opacity cases, the increment of the convolution layer greatly drops the misclassification rate between each case. Conclusively, the CNN classifier showed significantly greater accuracy than the SVM classifier, and the results implied structural characteristics that are inherent to the specific ILD patterns.     

Fluoroscopic gating without implanted fiducial markers for lung cancer radiotherapy based on support vector machines

Various problems with the current state-of-the-art techniques for gated radiotherapy have prevented this new treatment modality from being widely implemented in clinical routine. These problems are caused mainly by applying various external respiratory surrogates. There might be large uncertainties in deriving the tumor position from external respiratory surrogates. While tracking implanted fiducial markers has sufficient accuracy, this procedure may not be widely accepted due to the risk of pneumothorax. Previously, we have developed a technique to generate gating signals from fluoroscopic images without implanted fiducial markers using template matching methods (Berbeco et al 2005 Phys. Med. Biol. 50 4481-90, Cui et al 2007b Phys. Med. Biol. 52 741-55). In this note, our main contribution is to provide a totally different new view of the gating problem by recasting it as a classification problem. Then, we solve this classification problem by a well-studied powerful classification method called a support vector machine (SVM). Note that the goal of an automated gating tool is to decide when to turn the beam ON or OFF. We treat ON and OFF as the two classes in our classification problem. We create our labeled training data during the patient setup session by utilizing the reference gating signal, manually determined by a radiation oncologist. We then pre-process these labeled training images and build our SVM prediction model. During treatment delivery, fluoroscopic images are continuously acquired, pre-processed and sent as an input to the SVM. Finally, our SVM model will output the predicted labels as gating signals. We test the proposed technique on five sequences of fluoroscopic images from five lung cancer patients against the reference gating signal as ground truth. We compare the performance of the SVM to our previous template matching method (Cui et al 2007b Phys. Med. Biol. 52 741-55). We find that the SVM is slightly more accurate on average (1-3%) than the template matching method, when delivering the target dose. And the average duty cycle is 4-6% longer. Given the very limited patient dataset, we cannot conclude that the SVM is more accurate and efficient than the template matching method. However, our preliminary results show that the SVM is a potentially precise and efficient algorithm for generating gating signals for radiotherapy. This work demonstrates that the gating problem can be considered as a classification problem and solved accordingly.     

Modeling of heart systolic murmurs based on multivariate matching pursuit for diagnosis of valvular disorders

Heart murmurs are pathological sounds produced by turbulent blood flow due to certain cardiac defects such as valves disorders. Detection of murmurs via auscultation is a task that depends on the proficiency of physician. There are many cases in which the accuracy of detection is questionable. The purpose of this study is development of a new mathematical model of systolic murmurs to extract their crucial features for identifying the heart diseases. A high resolution algorithm, multivariate matching pursuit, was used to model the murmurs by decomposing them into a series of parametric time–frequency atoms. Then, a novel model-based feature extraction method which uses the model parameters was performed to identify the cardiac sound signals. The proposed framework was applied to a database of 70 heart sound signals containing 35 normal and 35 abnormal samples. We achieved 92.5% accuracy in distinguishing subjects with valvular diseases using a MLP classifier, as compared to the matching pursuit-based features with an accuracy of 77.5%.     

一种新特征评价方法在红斑鳞状皮肤病诊断中的应用

针对改进F-score特征评价准则没有考虑特征测量量纲对特征重要性的影响,提出一种新的特征重要性评价准则D-score,避免不同特征测量量纲的影响,衡量样本特征在两类或多类之间的辨别能力。将D-score分别与前向顺序搜索、前向顺序浮动搜索两种搜索策略结合,以支持向量机的分类准确率评估所选特征子集的有效性,结合Filter和Wrapper特征选择方法的优势进行特征选择,得到两种混合特征选择方法。将该方法应用于红斑鳞状皮肤病诊断研究,并与基于改进F-score的混合特征选择方法进行了实验对比。十折交叉验证实验结果显示:在红斑鳞状皮肤病诊断研究中,D-score特征评价准则优于改进的F-score准则,基于D-score和前向顺序搜索策略的诊断准确率提高1.11%;D-score结合前向顺序浮动搜索策略的最低诊断准确率提高约3个百分点,平均诊断准确率提高约0.3个百分点,最高诊断准确率达到100%。前向顺序浮动搜索中,D-score准则选择的共有特征是改进F-score准则所选择共有特征的子集。所提出的D-score特征重要性评价准则是一种有效的特征区分能力度量准则,在红斑鳞状皮肤病的诊断中选择出了更有分类意义的特征,提高了诊断准确性。     

Large-Scale Optimization-Based Classification Models in Medicine and Biology

We present novel optimization-based classification models that are general purpose and suitable for developing predictive rules for large heterogeneous biological and medical data sets. Our predictive model simultaneously incorporates (1) the ability to classify any number of distinct groups; (2) the ability to incorporate heterogeneous types of attributes as input; (3) a high-dimensional data transformation that eliminates noise and errors in biological data; (4) the ability to incorporate constraints to limit the rate of misclassification, and a reserved-judgment region that provides a safeguard against over-training (which tends to lead to high misclassification rates from the resulting predictive rule); and (5) successive multi-stage classification capability to handle data points placed in the reserved-judgment region. To illustrate the power and flexibility of the classification model and solution engine, and its multi-group prediction capability, application of the predictive model to a broad class of biological and medical problems is described. Applications include: the differential diagnosis of the type of erythemato-squamous diseases; predicting presence/absence of heart disease; genomic analysis and prediction of aberrant CpG island meythlation in human cancer; discriminant analysis of motility and morphology data in human lung carcinoma; prediction of ultrasonic cell disruption for drug delivery; identification of tumor shape and volume in treatment of sarcoma; discriminant analysis of biomarkers for prediction of early atherosclerois; fingerprinting of native and angiogenic microvascular networks for early diagnosis of diabetes, aging, macular degeneracy and tumor metastasis; prediction of protein localization sites; and pattern recognition of satellite images in classification of soil types. In all these applications, the predictive model yields correct classification rates ranging from 80 to 100%. This provides motivation for pursuing its use as a medical diagnostic, monitoring and decision-making tool.     

Classification of brain-stem trigeminal evoked potentials in multiple sclerosis, minor head injuries and post-concussion syndrome pathologies by similarity measurements

In this study measurements obtained from brain-stem trigeminal evoked potentials (BTEP) are applied to the problem of diagnosing Multiple Sclerosis (MS) and Post-concussion syndrome (PCS). We present a simplistic model that depicts the BTEP waveform as the linear combination of a set of filters excited by a short stimulus. The relation between the BTEP latencies and the 1st to 4th harmonic components is shown. The performance of a fuzzy similarity measure based classifier is compared with that of human experts. The efficiency of the proposed classifier in conjunction with delay time and amplitude features is examined. Using this novel approach, a classification rate of 93.55% and 84.1% for MS and PCS pathologies, respectively, was achieved. This performance compares favorably to the classification rates of 84.28% for MS and 70.47% for PCS pathologies achieved by human experts.     

3D shape analysis to reduce false positives for lung nodule detection systems

Using images from the Image Database Consortium and Image Database Resource Initiative (LIDC–IDRI), we developed a methodology for classifying lung nodules. The proposed methodology uses image processing and pattern recognition techniques. To classify volumes of interest into nodules and non-nodules, we used shape measurements only, analyzing their shape using shape diagrams, proportion measurements, and a cylinder-based analysis. In addition, we use the support vector machine classifier. To test the proposed methodology, it was applied to 833 images from the LIDC–IDRI database, and cross-validation with k-fold, where \(k = 5\), was used to validate the results. The proposed methodology for the classification of nodules and non-nodules achieved a mean accuracy of 95.33 %. Lung cancer causes more deaths than any other cancer worldwide. Therefore, precocious detection allows for faster therapeutic intervention and a more favorable prognosis for the patient. Our proposed methodology contributes to the classification of lung nodules and should help in the diagnosis of lung cancer.     

肺癌肿瘤标志物CYFRA21—1IRMA及其临床意义

本文用免疫放射分析法（ＩＲＭＡ）对３２７例肺癌病人（肺小细胞肺癌１２例，膜癌１０１例，鳞癌１２９例，未定型肺癌７４例，转移性肺癌１１例）；２９７例良性疾病患者（结核病人２１７例，感染病人６２例和１８例支气管扩张病人）的血清ＣＶＦＲＡ２１０１进行了测定，结果表明，灵敏度６２．４％（２０４／３２７），特异性９４．５％（２８２／２９７），准确率７７．９％（４８６／６２４）。肿瘤患者从Ｉ期到Ⅳ期血清ＣＹＦ     

Joint two-view information for computerized detection of microcalcifications on mammograms

We are developing new techniques to improve the accuracy of computerized microcalcification detection by using the joint two-view information on craniocaudal (CC) and mediolateral-oblique (MLO) views. After cluster candidates were detected using a single-view detection technique, candidates on CC and MLO views were paired using their radial distances from the nipple. Candidate pairs were classified with a similarity classifier that used the joint information from both views. Each cluster candidate was also characterized by its single-view features. The outputs of the similarity classifier and the single-view classifier were fused and the cluster candidate was classified as a true microcalcification cluster or a false-positive (FP) using the fused two-view information. A data set of 116 pairs of mammograms containing microcalcification clusters and 203 pairs of normal images from the University of South Florida (USF) public database was used for training the two-view detection algorithm. The trained method was tested on an independent test set of 167 pairs of mammograms, which contained 71 normal pairs and 96 pairs with microcalcification clusters collected at the University of Michigan (UM). The similarity classifier had a very low FP rate for the test set at low and medium levels of sensitivity. However, the highest mammogram-based sensitivity that could be reached by the similarity classifier was 69%. The single-view classifier had a higher FP rate compared to the similarity classifier, but it could reach a maximum mammogram-based sensitivity of 93%. The fusion method combined the scores of these two classifiers so that the number of FPs was substantially reduced at relatively low and medium sensitivities, and a relatively high maximum sensitivity was maintained. For the malignant microcalcification clusters, at a mammogram-based sensitivity of 80%, the FP rates were 0.18 and 0.35 with the two-view fusion and single-view detection methods, respectively. When the training and test sets were switched, a similar improvement was obtained, except that both the fusion and single-view detection methods had superior test performances on the USF data set than those on the UM data set. Our results indicate that correspondence of cluster candidates on two different views provides valuable additional information for distinguishing FPs from true microcalcification clusters.     

基于3D ResUnet网络的肺结节分割

目的:将深度残差结构和U-Net网络结合形成新的网络ResUnet,并利用ResUnet深度学习网络结构对胸部CT影像进行图像分割以提取肺结节区域。方法:使用的CT影像数据来源于LUNA16数据集,首先对CT图像预处理提取出肺实质,然后对其截取立体图像块并进行数据增强来扩充样本数,形成相应的肺结节掩膜图像,最后将生成的样本输入到ResUnet模型中进行训练。结果:本研究模型最终的精度和召回率分别为35.02%和97.68%。结论:该模型能自动学习肺结节特征,为后续的肺癌自动诊断提供可靠基础,减少临床诊断的成本并节省医生诊断的时间。 【关键词】肺结节;分割;深度残差结构;召回率;ResUnet     

On a PCA-based lung motion model

Respiration-induced organ motion is one of the major uncertainties in lung cancer radiotherapy and is crucial to be able to accurately model the lung motion. Most work so far has focused on the study of the motion of a single point (usually the tumor center of mass), and much less work has been done to model the motion of the entire lung. Inspired by the work of Zhang et al (2007 Med. Phys. 34 4772-81), we believe that the spatiotemporal relationship of the entire lung motion can be accurately modeled based on principle component analysis (PCA) and then a sparse subset of the entire lung, such as an implanted marker, can be used to drive the motion of the entire lung (including the tumor). The goal of this work is twofold. First, we aim to understand the underlying reason why PCA is effective for modeling lung motion and find the optimal number of PCA coefficients for accurate lung motion modeling. We attempt to address the above important problems both in a theoretical framework and in the context of real clinical data. Second, we propose a new method to derive the entire lung motion using a single internal marker based on the PCA model. The main results of this work are as follows. We derived an important property which reveals the implicit regularization imposed by the PCA model. We then studied the model using two mathematical respiratory phantoms and 11 clinical 4DCT scans for eight lung cancer patients. For the mathematical phantoms with cosine and an even power (2n) of cosine motion, we proved that 2 and 2n PCA coefficients and eigenvectors will completely represent the lung motion, respectively. Moreover, for the cosine phantom, we derived the equivalence conditions for the PCA motion model and the physiological 5D lung motion model (Low et al 2005 Int. J. Radiat. Oncol. Biol. Phys. 63 921-9). For the clinical 4DCT data, we demonstrated the modeling power and generalization performance of the PCA model. The average 3D modeling error using PCA was within 1 mm (0.7 ± 0.1 mm). When a single artificial internal marker was used to derive the lung motion, the average 3D error was found to be within 2 mm (1.8 ± 0.3 mm) through comprehensive statistical analysis. The optimal number of PCA coefficients needs to be determined on a patient-by-patient basis and two PCA coefficients seem to be sufficient for accurate modeling of the lung motion for most patients. In conclusion, we have presented thorough theoretical analysis and clinical validation of the PCA lung motion model. The feasibility of deriving the entire lung motion using a single marker has also been demonstrated on clinical data using a simulation approach.     

螺旋CT间接淋巴造影及增强扫描评估卵巢癌前哨淋巴结方法及临床意义

目的分析螺旋CT间接淋巴造影及增强扫描评估卵巢癌前哨淋巴结方法及临床意义。方法回顾性分析我院82例卵巢癌患者临床资料,均行前哨淋巴结影像检查、活检、腹腔镜切除卵巢术等,且均予以螺旋CT间接淋巴造影及多期增强扫描。结果 T4时点多期增强扫描+间接淋巴造影检出率,显著高于单一检出率(P0.05);N1、N于多期增强扫描、间接淋巴造影、活检及多期增强扫描联合间接淋巴造影方法检出率对比均显示高度差异(P0.05)。结论卵巢癌前哨淋巴结行螺旋CT间接淋巴造影技术与多期增强扫描进行术前评估,为卵巢癌SLN清扫提供术前、术中参考依据,同时能准确地对卵巢癌分期诊断,判断淋巴结的累及情况,为临床提供科学参考。