Irbm最新文献

Objectives

The accurate preoperative segmentation of the uterus and uterine fibroids from magnetic resonance images (MRI) is an essential step for diagnosis and real-time ultrasound guidance during high-intensity focused ultrasound (HIFU) surgery. Conventional supervised methods are effective techniques for image segmentation. Recently, semi-supervised segmentation approaches have been reported in the literature. One popular technique for semi-supervised methods is to use pseudo-labels to artificially annotate unlabeled data. However, many existing pseudo-label generations rely on a fixed threshold used to generate a confidence map, regardless of the proportion of unlabeled and labeled data.

Materials and Methods

To address this issue, we propose a novel semi-supervised framework called Confidence-based Threshold Adaptation Network (CTANet) to improve the quality of pseudo-labels. Specifically, we propose an online pseudo-labels method to automatically adjust the threshold, producing high-confident unlabeled annotations and boosting segmentation accuracy. To further improve the network's generalization to fit the diversity of different patients, we design a novel mixup strategy by regularizing the network on each layer in the decoder part and introducing a consistency regularization loss between the outputs of two sub-networks in CTANet.

Results

We compare our method with several state-of-the-art semi-supervised segmentation methods on the same uterine fibroids dataset containing 297 patients. The performance is evaluated by the Dice similarity coefficient, the precision, and the recall. The results show that our method outperforms other semi-supervised learning methods. Moreover, for the same training set, our method approaches the segmentation performance of a fully supervised U-Net (100% annotated data) but using 4 times less annotated data (25% annotated data, 75% unannotated data).

Conclusion

Experimental results are provided to illustrate the effectiveness of the proposed semi-supervised approach. The proposed method can contribute to multi-class segmentation of uterine regions from MRI for HIFU treatment.

Purpose

Three main non-invasive imaging methods are routinely used for the assessment of liver fibrosis and steatosis in patients with nonalcoholic fatty liver disease (NAFLD): the vibration-controlled transient elastography (VCTE) using the FibroScan device, the magnetic resonance imaging (MRI) based on proton density fat fraction (PDFF), and the magnetic resonance elastography (MRE). The purpose of our study is to evaluate the efficiency of the VCTE findings compared to the two others methods, and to analyze the impact of hepatic iron overload on these comparisons.

Methods

A clinical study was performed on 94 patients with NAFLD in the radiology department of ACRIM-Polyclinic Saint-Côme (France). The study also included 17 patients with hemochromatosis, measured from $T{2}^{⁎}$ MRI. The liver tissues of all the patients were evaluated with 1) VCTE (including the controlled attenuation (CAP) and stiffness parameters), 2) MRI (fat fraction parameter), and 3) MRE (stiffness parameter) techniques. The performance of VCTE was assessed by estimating the area under the ROC curve (AUC) for patients without or with hemochromatosis. Spearman's correlation was used for the comparison of VCTE measurements to MRI and MRE.

Results

VCTE-based stiffness and CAP were significantly correlated with PDFF and MRE measurements ($P<0.01$) for the subgroup without hemochromatosis. The correlations failed for the subgroup with hemochromatosis.

Conclusion

VCTE and CAP measurements were not correlated with those from MR PDFF and MRE for patients with hemochromatosis. VCTE, PDFF and MRE modalities don't give concordant results for patients with hemochromatosis.

Objectives

Esophageal cancer is a high occult malignant tumor. Even with good diagnosis and treatment, the 5-year survival rate of esophageal cancer patients is still less than 30%. Considering the influence of clinical characteristics on postoperative esophageal cancer patients, the construction of a neural network model will help improve the poor prognosis of patients in the five years.

Material and methods

In this study, genetic algorithm optimized deep neural network is exploited to the clinical dataset of esophageal cancer. The independent prognostic factors are screened by Relief algorithm and Cox proportional risk regression. FTD prognostic staging system is established to assess the risk level of esophageal cancer patients.

Results

FTD staging system and independent prognostic factors are integrated into the genetic algorithm optimized deep neural network. The Area Under Curve (AUC) of FTD staging system is 0.802. FTD staging system is verified by the Kaplan-Meier survival curve, and the median survival time is divided for different risk grades. The FTD staging system is superior to the TNM stages in the prognosis effect. The AUC of deep neural network optimized by genetic algorithm is 0.91.

Conclusion

The deep neural network optimized by genetic algorithm has good performance in predicting the 5-year survival status of esophageal cancer patients. The FTD staging system has a significant prognostic effect. The FTD staging system and genetic algorithm optimized deep neural network can be successfully availed in clinical diagnosis and treatment.

Background: An open challenge of P300-based BCI systems focuses on recognizing ERP signals using a reduced number of trials with enough classification rate.

Methods: Three novel methods based on Filter Bank and Canonical Correlation Analysis (CCA) are proposed for the recognition of P300 ERPs using a reduced number of trials. The proposed methods were evaluated with two freely available EEG datasets based on 6x6 speller and were compared with five standard methods: Mean-Amplitude, Step-Wise, Principal Component Analysis, Peak, and CCA.

Results: The proposed methods outperform significantly standard algorithms for P300 identification with a maximum AUC of 0.93 and 0.98, and an average of 0.73 and 0.76, using a single trial.

Conclusion: Proposed methods based on Filter Bank are robust for the identification of P300 using a reduced number of trials, which could be used in real-time BCI spellers for rehabilitation engineering.

Background and Objectives

Tumor Treating Fields (TTFields) is an FDA-approved technique used in the treatment of glioblastoma. It consists in applying an electric field (EF) with a frequency of 200 kHz using two pairs of transducer arrays. During treatment planning, the NovoTAL system is used to strategically place the arrays on the head in regions that maximize the EF at the tumor. Current should be injected at least 18 hours/day and induce a minimum EF of 1 V/cm at the tumor. To avoid any thermal harm to the patient, the temperature of the scalp is kept around 39.5 °C by changing the injected current. The goal of this study was to investigate how accounting for the temperature of the scalp during treatment planning might affect the choice of the best layout suggested by NovoTAL. Furthermore, we also studied the sensitivity of the results to the metric used to evaluate the layouts.

Methods

We used a realistic head model with a virtual glioblastoma in our studies. Through the NovoTAL system we obtained five realistic array layouts and we predicted the best one for our model based on the approach currently implemented in this system. We then repeated the same type of analysis, but also accounting for the temperature during planning. In both cases we ranked the layouts based on three different criteria: the LMiPD and the LAPD (local minimum and local average power densities, respectively) in the tumor and the SAR (specific absorption rate) in the head

Results

Accounting for the temperature does not significantly affect the choice of the best layout provided that the arrays are at least 1 cm apart from each other. Otherwise, a common temperature hotspot occurs in the scalp between the closest transducers of the adjacent arrays, which limits how much current can be injected and consequently treatment effectiveness. Also, the choice of the best layout depends on the criterion used.

Conclusions

Accounting for the temperature might led to significant variations in the current injected. The LMiPD might be used as a first criterion to choose the best treatment layout, followed by the LAPD and then the SAR.

Objectives

Pressure ulcers are a great handicap for those who develop one. Pressure ulcers can take a long time to heal especially if detected late. These afflictions require a lot of time from the medical personnel and thus a great amount of money. We aim here to check the impact of continuous measurement on the performance of early pressure ulcer detection algorithms.

Material and methods

To detect pressure ulcers early on we use a simulation of a human buttocks to simulate the reaction of it to pressure. This simulation considers the most recent findings about pressure ulcers. In particular, the phenomenon of muscle stiffening when pressure is applied for a long period of time, and the reperfusion phenomenon. We can then simulate pressure captors on the outside interface of the buttocks to use these measurements for detection. We then determine the best threshold on the measured pressures to create standard algorithms that we compare to novel algorithms using an optimized threshold on a calculated damage based on the pressure measurement of the last 2 hours.

Results

We compare these different algorithms for the early detection of pressure ulcers and show the need to take the measurement variation in time for a better detection. The detection error is improved by 7.3% for balanced classes and 2.7% for a dataset with a majority of healthy buttocks.

Conclusion

We showed that taking the evolution of pressure instead of only instantaneous measurement can improve the early detection of pressure ulcer.

Objective

this paper presents an innovative graphical user interface to visualize the attitude of a sensing device in a three-dimensional space, serving a wide-range of medical applications.

Material and methods

based on inertial measurement units (IMU) or on magnetic, angular rate and gravity (MARG) sensors, a processing unit provides Euler angles using a sensor fusion technique to display the orientation of the device relative to the Earth frame in real-time. The device is schematized by linking six polygonal regions, and is subject to sequential rotations by updating the graph each 350 ms. We conduct comparative studies between the two sensing devices, i.e. IMUs and MARGs, as well as two orientation filters, namely Madgwick's algorithm and Mahony's algorithm.

Results

the accuracy of the system is reported as a function of (i) the sampling frequency, (ii) the sensing unit, and (iii) the orientation filter, following two elderly care applications, namely fall risk assessment and body posture monitoring. The experiments are conducted using public datasets. The corresponding results show that Madgwick's algorithm is best suited for low sampling rates, whereas MARG sensors are best suited for the detection of postural transitions.

Conclusion

this paper addresses the different aspects and discusses the limitations of attitude estimation systems, which are important tools to help clinicians in their diagnosis.

Objective

The most widespread and intrusive cancer type among women is breast cancer. Globally, this type of cancer causes more mortality among women, next to lung cancer. This made the researchers to focus more on developing effective Computer-Aided Detection (CAD) methodologies for the classification of such deadly cancer types. In order to improve the rate of survival and earlier diagnosis, an optimistic research methodology is required in the classification of breast cancer. Consequently, an improved methodology that integrates the principle of deep learning with metaheuristic and classification algorithms is proposed for the severity classification of breast cancer. Hence to enhance the recent findings, an improved CAD methodology is proposed for redressing the healthcare problem.

Material and Methods

The work intends to cast a light-of-research towards classifying the severities present in digital mammogram images. For evaluating the work, the publicly available MIAS, INbreast, and WDBC databases are utilized. The proposed work employs transfer learning for extricating the features. The novelty of the work lies in improving the classification performance of the weighted k-nearest neighbor (wKNN) algorithm using particle swarm optimization (PSO), dragon-fly optimization algorithm (DFOA), and crow-search optimization algorithm (CSOA) as a transformation technique i.e., transforming non-linear input features into minimal linear separable feature vectors.

Results

The results obtained for the proposed work are compared then with the Gaussian Naïve Bayes and linear Support Vector Machine algorithms, where the highest accuracy for classification is attained for the proposed work (CSOA-wKNN) with 84.35% for MIAS, 83.19% for INbreast, and 97.36% for WDBC datasets respectively.

Conclusion

The obtained results reveal that the proposed Computer-Aided-Diagnosis (CAD) tool is robust for the severity classification of breast cancer.

Objectives

In this paper, we present a plugin for the optimal placement of sensors in a smart home. Our approach includes the Building Information Modeling (BIM) which is a plan that describes the building layout.

Material and methods

This plugin uses the CSTB EveBim viewer for loading IFC file representing the digital building's model. We use then, a mathematical model based on a mixed integer linear program, to determine the optimal sensor placement according to building and sensors characteristics.

Results

The results show the efficiency of the proposed algorithm and the developed plugin. We obtain an optimal solution after few seconds, and we show the sensor placement on the building digital model.

Conclusion

We show the relevance of the proposed plugin to equip room of retirement home or ambient assisted living in order to identify occupant activity for home support application.

Objectives

Because there are many categories, large morphological differences and few labels of urinary sediment components, and uneven data in urine sediment images recognition, the accuracy and recall rate of the existing urine sediment images recognition methods are not ideal. The main purpose of this paper is to improve the accuracy and recall of urine sediment image recognition by proposing a urine sediment image classification method based on semi-supervised learning.

Methods

This paper proposes a method based on semi-supervised learning to classify urine sediment images. This algorithm designs a re-parameterization network (US-RepNet) for low-resolution urine sediment microscopic images to extract complex features of urine sediment images. The dual attention module is introduced on Us-RepNet to increase the extraction of fine-grained features from urine sediment images. And the cross-entropy loss (C.E. loss) function is optimized to train an unbiased classifier to improve long-tailed distribution image classification.

Results

The experimental results show that the accuracy of proposed method can reach 94% with only a small amount of labeled data for 16 types of urine sediment images under long-tail distribution.

Conclusion

The algorithm can recognize most types, and reduces the need for labeled information, while achieving excellent recognition and classification performance. Comprehensive analysis shows that it can be used as a practical reference for urine sediment analysis.