True Positive Fraction Research Articles

Evaluation of cumulative rainfall and rainfall event–duration threshold based on triggering and non-triggering rainfalls: Northern Thailand case

Abstract A recently introduced rainfall threshold for landslide early warning systems combined a cumulative rainfall threshold with a rainfall event–duration (ED) threshold. Cumulative rainfall with rainfall event–duration, known as the CED threshold, was reported to perform better than the conventional ED threshold. However, the establishment of the CED threshold was based on a frequentist approach which required an adequate number of landslide-triggering rainfall data. An alternative to the use of landslide-triggering rainfall data is the use of non-triggering rainfall data. These rainfall events supply much bigger amount of data to the susceptibility model. Although the establishment of a rainfall threshold based on non-triggering rainfall events is seldom considered, previous scholars reported that this approach has produced better results than the conventional approach based on landslide-triggering events. This study investigates the reliability and prediction performance of the CED threshold based on non-triggering rainfall data. The performance of this threshold, designated as the negative-CED (CEDN) threshold, was compared with the positive-CED (CEDP) threshold based on landslide-triggering rainfall data. North Thailand, a landslide hot spot, was chosen as the study area. The proposed threshold was assessed from three skill scores, including (1) the true positive fraction (TPF), (2) the false positive fraction (FPF), and (3) the positive predictive value (PPV), and their variations over the range of threshold uncertainties. Rather than possessing lower uncertainties of the threshold parameters, the negative threshold provided better compromise predictions of TPF and FPF scores than the positive thresholds. Integrating the cumulative rainfall threshold and an event-based rainfall threshold resulted in a significant improvement in FPF scores, and hence enhanced the compromise predictions of TPF and FPF scores. Keeping in mind that the negative thresholds were not established from landslide data, care must be taken when using these thresholds and it is recommended that they should be applied only to areas where landslide data are limited.

Atypical architectural distortion detection in digital breast tomosynthesis: a multi-view computer-aided detection model with ipsilateral learning

Objective. Breast architectural distortion (AD), a common imaging symptom of breast cancer, is associated with a particularly high rate of missed clinical detection. In clinical practice, atypical ADs that lack an obvious radiating appearance constitute most cases, and detection models based on single-view images often exhibit poor performance in detecting such ADs. Existing multi-view deep learning methods have overlooked the correspondence between anatomical structures across different views. Approach. To develop a computer-aided detection (CADe) model for AD detection that effectively utilizes the craniocaudal (CC) and mediolateral oblique (MLO) views of digital breast tomosynthesis (DBT) images, we proposed an anatomic-structure-based multi-view information fusion approach by leveraging the related anatomical structure information between these ipsilateral views. To obtain a representation that can effectively capture the similarity between ADs in images from ipsilateral views, our approach utilizes a Siamese network architecture to extract and compare information from both views. Additionally, we employed a triplet module that utilizes the anatomical structural relationship between the ipsilateral views as supervision information. Main results. Our method achieved a mean true positive fraction (MTPF) of 0.05–2.0, false positives (FPs) per volume of 64.40%, and a number of FPs at 80% sensitivity (FPs@0.8) of 3.5754; this indicates a 6% improvement in MPTF and 16% reduction in FPs@0.8 compared to the state-of-the-art baseline model. Significance. From our experimental results, it can be observed that the anatomic-structure-based fusion of ipsilateral view information contributes significantly to the improvement of CADe model performance for atypical AD detection based on DBT. The proposed approach has the potential to lead to earlier diagnosis and better patient outcomes.

Offline Verification of Daily Adaptive Therapy for Locally Advanced Lung Cancer

Daily adaptive therapy (AdaptT) for locally advanced lung cancer has been an intriguing field with the adoption of cone beam CT (CBCT). We have utilized AdaptT for lung cancer. In this workflow, contours for targets and organs at risk from the original CT simulation dataset are deformed and registered to the CBCT for each fraction. A physicist reviews and edits the deformed contours online without MD presence and a new plan is treated daily. Treating physicians are required to attend 1 fraction per week. We sought to evaluate the accuracy of these contours for targets and OARs in comparison to a thoracic physician contouring offline with the same CBCT dataset. The impetus of this study was to evaluate the model of physics directed contouring and ensure that gross tumor volume (GTV) misses are not occurring during treatment, defined as Planning Target Volumes (PTV's) derived from AdaptT failing to encompass the GTV redrawn offline by the physician. Six consecutive patients were treated from 2021-2022 with AdaptT utilizing processes as described above. One CBCT per week was re-contoured by a physician for 5-6 CBCT's per case. The GTVs, heart, and lungs were redrawn without registration. The physician was blinded to the structures approved for that fraction and from clinical information outside of the CT simulation data. The clinical target volume expansions were 5 mm shaved from normal tissue barriers and a uniform 5 mm expansion to form the PTV. The Dice Similarity Coefficient (DSE), mean surface distance (MSD), 95th percentile Hausdorff distance (95HD) for contours were calculated and the True Positive Fraction (TPF) of the redrawn GTVs within the treated PTVs were calculated. Thirty CBCTs from 6 patients were contoured. The TPF was 1 for each redrawn GTV, indicating each GTV drawn by the physician offline would have been encompassed within the PTV for treatment. The median 95HF for the targets was below 5 mm (Table 1) and the MSD for targets was close to 1.07 mm. The DSC for Heart and Lungs approached 1.0, indicating close similarity. The average time the physician spent on contouring was 9.8 minutes compared to 9.6 minutes with AdaptT. All redrawn GTVs were within the treated PTV and there was close agreement of normal structures. We evaluated the current model employing daily physics overview with weekly physician input and it appears safe and feasible. Further prospective study is needed to standardize comparison of contour quality and validate clinical outcomes of this approach.

A virtual-pinhole PET device for improving contrast recovery and enhancing lesion detectability of a one-meter-long PET scanner: a simulation study

This paper presents a simulation study to demonstrate that the contrast recovery coefficients (CRC) and detectability of small lesions of a one-meter-long positron emission tomography (PET) scanner can be further enhanced by the integration of high resolution virtual-pinhole (VP) PET devices. The scanner under investigation is a Siemens Biograph Vision Quadra which has an axial field-of-view (FOV) of 106 cm. The VP-PET devices contain two high-resolution flat panel detectors, each composed of 2 × 8 detector modules each of which consists of 32 × 64 lutetium-oxyorthosilicate crystals (1.0 × 1.0 × 10.0 mm3 each). Two configurations for the VP-PET device placement were evaluated: (1) place the two flat-panel detectors at the center of the scanner’s axial FOV below the patient bed; (2) place one flat-panel detector at the center of the first and the last quarter of the scanner’s axial FOV below the patient bed. Sensitivity profiles were measured by moving a point 22Na source stepwise across the scanner’s FOV axially at different locations. To assess the improvement in CRC and lesion detectability by the VP-PET devices, an elliptical torso phantom (31.6 × 22.8 × 106 cm3) was first imaged by the native scanner then subsequently by the two VP-PET geometry configurations. Spherical lesions (4 mm in diameter) having 5:1 lesion-to-background radioactivity concentration ratio were grouped and placed at nine regions in the phantom to analyze the dependence of the improvement in plane. Average CRCs and their standard deviations of the 7 tumors in each group were computed and the receiver operating characteristic (ROC) curves were drawn to evaluate the improvement in lesion detectability by the VP-PET device over the native long axial PET scanner. The fraction of coincidence events between the inserts and the scanner detectors was 13%–16% (out of the total number of coincidences) for VP-PET configuration 1 and 2, respectively. The VP-PET systems provide higher CRCs for lesions in all regions in the torso, with more significant enhancement at regions closer to the inserts, than the native scanner does. For any given false positive fraction, the VP-PET systems offer higher true positive fraction compared to the native scanner. This work provides a potential solution to further enhance the image resolution of a long axial FOV PET scanner to maximize its lesion detectability afforded by its super high effective sensitivity.

Priors, population sizes, and power in genome-wide hypothesis tests

BackgroundGenome-wide tests, including genome-wide association studies (GWAS) of germ-line genetic variants, driver tests of cancer somatic mutations, and transcriptome-wide association tests of RNAseq data, carry a high multiple testing burden. This burden can be overcome by enrolling larger cohorts or alleviated by using prior biological knowledge to favor some hypotheses over others. Here we compare these two methods in terms of their abilities to boost the power of hypothesis testing.ResultsWe provide a quantitative estimate for progress in cohort sizes and present a theoretical analysis of the power of oracular hard priors: priors that select a subset of hypotheses for testing, with an oracular guarantee that all true positives are within the tested subset. This theory demonstrates that for GWAS, strong priors that limit testing to 100–1000 genes provide less power than typical annual 20–40% increases in cohort sizes. Furthermore, non-oracular priors that exclude even a small fraction of true positives from the tested set can perform worse than not using a prior at all.ConclusionOur results provide a theoretical explanation for the continued dominance of simple, unbiased univariate hypothesis tests for GWAS: if a statistical question can be answered by larger cohort sizes, it should be answered by larger cohort sizes rather than by more complicated biased methods involving priors. We suggest that priors are better suited for non-statistical aspects of biology, such as pathway structure and causality, that are not yet easily captured by standard hypothesis tests.

Atypical architectural distortion detection in digital breast tomosynthesis: a computer-aided detection model with adaptive receptive field

Objective. In digital breast tomosynthesis (DBT), architectural distortion (AD) is a breast lesion that is difficult to detect. Compared with typical ADs, which have radial patterns, identifying a typical ADs is more difficult. Most existing computer-aided detection (CADe) models focus on the detection of typical ADs. This study focuses on atypical ADs and develops a deep learning-based CADe model with an adaptive receptive field in DBT. Approach. Our proposed model uses a Gabor filter and convergence measure to depict the distribution of fibroglandular tissues in DBT slices. Subsequently, two-dimensional (2D) detection is implemented using a deformable-convolution-based deep learning framework, in which an adaptive receptive field is introduced to extract global features in slices. Finally, 2D candidates are aggregated to form the three-dimensional AD detection results. The model is trained on 99 positive cases with ADs and evaluated on 120 AD-positive cases and 100 AD-negative cases. Main results. A convergence-measure-based model and deep-learning model without an adaptive receptive field are reproduced as controls. Their mean true positive fractions (MTPF) ranging from 0.05 to 4 false positives per volume are 0.3846 ± 0.0352 and 0.6501 ± 0.0380, respectively. Our proposed model achieves an MTPF of 0.7148 ± 0.0322, which is a significant improvement (p < 0.05) compared with the other two methods. In particular, our model detects more atypical ADs, primarily contributing to the performance improvement. Significance. The adaptive receptive field helps the model improve the atypical AD detection performance. It can help radiologists identify more ADs in breast cancer screening.

Constructing a pollen proxy from low-cost Optical Particle Counter (OPC) data processed with Neural Networks and Random Forests

Pollen allergies affect a significant proportion of the global population, and this is expected to worsen in years to come. There is demand for the development of automated pollen monitoring systems. Low-cost Optical Particle Counters (OPCs) measure particulate matter and have attractive advantages of real-time high time resolution data and affordable costs. This study asks whether low-cost OPC sensors can be used for meaningful monitoring of airborne pollen.We employ a variety of methods, including supervised machine learning techniques, to construct pollen proxies from hourly-average OPC data and evaluate their performance, holding out 40 % of observations to test the proxies. The most successful methods are supervised machine learning Neural Network (NN) and Random Forest (RF) methods, trained from pollen concentrations collected from a Hirst-type sampler. These perform significantly better than using a simple particle size-filtered proxy or a Positive Matrix Factorisation (PMF) source apportionment pollen proxy. Twelve NN and RF models were developed to construct a pollen proxy, each varying by model type, input features and target variable. The results show that such models can construct useful information on pollen from OPC data. The best metrics achieved (Spearman correlation coefficient = 0.85, coefficient of determination = 0.67) were for the NN model constructing a Poaceae (grass) pollen proxy, based on particle size information, temperature, and relative humidity. Ability to distinguish high pollen events was evaluated using F1 Scores, a score reflecting the fraction of true positives with respect to false positives and false negatives, with promising results (F1 ≤ 0.83). Model-constructed proxies demonstrated the ability to follow monthly and diurnal trends in pollen. We discuss the suitability of OPCs for monitoring pollen and offer advice for future progress. We demonstrate an attractive alternative for automated pollen monitoring that could provide valuable and timely information to the benefit of pollen allergy sufferers.

Assessing a model of Pacific Northwest harmful algal bloom transport as a decision-support tool

In the Pacific Northwest, blooms of the diatom Pseudo-nitzschia (PN) sometimes produce domoic acid, a neurotoxin that causes amnesic shellfish poisoning, leading to a Harmful Algal Bloom (HAB) event. The Pacific Northwest (PNW) HAB Bulletin project, a partnership between academic, government, and tribal stakeholders, uses a combination of beach and offshore monitoring data and ocean forecast modeling to better understand the formation, evolution, and transport of HABs in this region. This project produces periodic Bulletins to inform local stakeholders of current and forecasted conditions. The goal of this study was to help improve how the forecast model is used in the Bulletin's preparation through a retrospective particle-tracking experiment. Using past observations of beach PN cell counts, events were identified that likely originated in the Juan de Fuca eddy, a known PN hotspot, and then particle tracks were used in the model to simulate these events. A variety of “beaching definitions” were tested, based on both water depth and distance offshore, to define when a particle in the model was close enough to the coast that it was likely to correspond to cells appearing in the intertidal zone and in shellfish diets, as well as a variety of observed PN cell thresholds to determine what cell count should be used to describe an event that would warrant further action. The skill of these criteria was assessed by determining the fraction of true positives, true negatives, false positives, and false negatives within the model in comparison with observations, as well as a variety of derived model performance metrics. This analysis suggested that for our stakeholders’ purposes, the most useful beaching definition is the 30 m isobath and the most useful PN cell threshold for coincident field-based sample PN density estimates is 10,000 PN cells/L. Lastly, the performance of a medium-resolution (1.5 km horizontal resolution) version of the model was compared with that of a high-resolution (0.5 km horizontal resolution) version, the latter currently used in forecasting for the PNW HAB Bulletin project. This analysis includes a direct comparison of the two model resolutions for one overlapping year (2017). These results suggested that a narrower, more realistic beaching definition is most useful in a high-resolution model, while a wider beaching definition is more appropriate in a lower resolution model like the medium-resolution version used in this analysis. Overall, this analysis demonstrated the importance of incorporating stakeholder needs into the statistical approach in order to generate the most effective decision-support information from oceanographic modeling.

Chemical Space Exploration with Active Learning and Alchemical Free Energies.

Drug discovery can be thought of as a search for a needle in a haystack: searching through a large chemical space for the most active compounds. Computational techniques can narrow the search space for experimental follow up, but even they become unaffordable when evaluating large numbers of molecules. Therefore, machine learning (ML) strategies are being developed as computationally cheaper complementary techniques for navigating and triaging large chemical libraries. Here, we explore how an active learning protocol can be combined with first-principles based alchemical free energy calculations to identify high affinity phosphodiesterase 2 (PDE2) inhibitors. We first calibrate the procedure using a set of experimentally characterized PDE2 binders. The optimized protocol is then used prospectively on a large chemical library to navigate toward potent inhibitors. In the active learning cycle, at every iteration a small fraction of compounds is probed by alchemical calculations and the obtained affinities are used to train ML models. With successive rounds, high affinity binders are identified by explicitly evaluating only a small subset of compounds in a large chemical library, thus providing an efficient protocol that robustly identifies a large fraction of true positives.

Assessing model accuracy using random data split: a simulation study

ABSTRACT Randomization is considered a safeguard against bias and a gold standard in clinical studies. To assess the generalizability of the accuracy of a model, a common approach is to randomly split a master data set into two parts: one for training and the other for testing. In this paper, we demonstrated the limitations of random split in assessing the generalizability of the accuracy of models through simulation studies. We generated three simulation data for binary or continuous endpoints, each with large sample size (n = 10,000). In each simulation scenario, we randomly split the data into two, one for training and one for testing, and then compare the performance of the model between training and testing data. All simulations were repeated 1,000 times. When random split was used, the model performance based on training and testing data behaved similarly in terms of the true positive fraction and false positive fraction for binary data and mean-squared errors for continuous data. However, when there is a time drift effect in the data, random split will result in large differences between training and testing data. As the training and testing data are similar through a random split, assessing the generalizability of the model on similar data will generate similar results. Generalizability of the accuracy of models is thus best achieved if testing is done in a distinct and independent study.

Observed versus predicted mortality after isolated tricuspid valve surgery.

BackgroundAim of this study is to analyse the performances of Clinical Risk Score (CRS) and European System for Cardiac Operative Risk Evaluation (EuroSCORE)‐II in isolated tricuspid surgery.MethodsThree hundred and eighty‐three patients (54 ± 16 year; 54% female) were enrolled. Receiver operating characteristic analysis was performed to evaluate the relationship between the true positive fraction of test results and the false‐positive fraction for a procedure.ResultsConsidering the 30‐day mortality the area under the curve was 0.6 (95% confidence interval [CI] 0.50–0.72) for EuroSCORE II and 0.7 (95% CI 0.56–0.84) for CRS‐score. The ratio of expected/observed mortality showed underestimation when considering EuroSCORE‐II (min. 0.46–max. 0.6). At multivariate analysis, the CRS score (p = .005) was predictor of late cardiac death.ConclusionWe suggest using both scores to obtain a range of expected mortality. CRS to speculate on late survival.

Architectural distortion detection based on superior-inferior directional context and anatomic prior knowledge in digital breast tomosynthesis.

In 2020, breast cancer becomes the most leading diagnosed cancer all over the world. The burden is increasing in the prevention and treatment of breast cancer. Accurately detecting breast lesions in screening images is important for early detection of cancer. Architectural distortion (AD) is one of the breast lesions that need to be detected. To develop a deep-learning-based computer-aided detection (CADe) model for AD in digital breast tomosynthesis (DBT). This model uses the superior-inferior directional context of DBT and anatomic prior knowledge to reduce false positive (FP). It can identify some negative samples that cannot be distinguished by deep learning features. The proposed CADe model consists of three steps. In the first step, a deep learning detection network detects two-dimensional (2D) candidates of ADs in DBT slices with the inputs preprocessed by Gabor filters and convergence measure. In the second step, three-dimensional (3D) candidates are obtained by stacking 2D candidates along superior-inferior direction. In the last step, FP reduction for 3D candidates is implemented based on superior-inferior directional context and anatomic prior knowledge of breast. DBT data from 99 cases with AD were used as the training set to train the CADe model, and data from 208 cases were used as an independent test set (including 108 cases with AD and 100 cases without AD as the control group). The free-response receiver operating characteristic and mean true positive fraction (MTPF) in the range of 0.05-2.0 FPs per volume are used to evaluate the model. Compared with the baseline model based on convergence measure, our proposed method demonstrates significant improvement (MTPF: 0.2826±0.0321vs. 0.6640±0.0399). Results of an ablation study show that our proposed context- and anatomy-based FP reduction methods improve the detection performance. The number of FPs per DBT volume reduces from 2.47 to 1.66 at 80% sensitivity after employing these two schemes. The deep learning model demonstrates practical value for AD detection. The results indicate that introducing superior-inferior directional context and anatomic prior knowledge into model can indeed reduce FPs and improve the performance of CADe model.

Role of Magnetic Resonance Cholangiopancreatography and Ultrasound as a Diagnostic Tool in Suspected Cases of Biliary Obstruction: A Prospective Cohort Study

Introduction: Obstructive jaundice is a common and grave form of hepatobiliary disease. It can pose problems in diagnosis and management, particularly intrahepatic cholestasis. Magnetic Resonance Cholangiopancreatography (MRCP) and Ultrasound (USG) are the frontline modality for biliary and pancreatic duct imaging. Aim: To evaluate the accuracy of USG and MRCP in assessing the level and cause of biliary obstruction which were compared with pathological or surgical findings. Materials and Methods: This prospective cohort study was conducted in the Department of Radiodiagnosis in a teriary care hospital from August 2013 to August 2015. Total 60 patients who were referred from the Department of Surgery and Medicine with strong clinical suspicion of biliary obstruction and altered Liver Function Test (LFT) were included in the study. All these patients underwent USG examination within a day of MRCP. Patients with previous biliary surgery and in whom Magnetic Resonance Imaging (MRI) was contraindicated were excluded. As a gold standard, the authors used ERCP, histological test and pathological findings after surgical intervention. The data collected was statistically analysed; sensitivity and specificity were calculated by using true and false positive and/or negative fractions. Results: The majority of patients were females and in sixth decade of life. The diagnostic accuracy of USG in detecting level of obstruction was comparable with MRCP. USG was 100% accurate in detecting hilar and suprapancreatic obstructions. In infrapancreatic obstructions diagnostic accuracy of USG was 95% as pancreatic head and peripancreatic region are poorly visualised on ultrasonography due to bowel gas shadows. Diagnostic accuracy of USG in the present study was 98.3%, one false negative case was of a small, soft, 4 mm calculus in distal Common Bile Duct (CBD). Sensitivity and specificity of USG in differentiating cystic from solid masses was 100%. Conclusion: The USG with improved technology and operator experience should be considered as the first choice in the diagnostic imaging of obstructive biliary disease. If the suspicion posed by clinical and laboratory findings are not confirmed at USG, the diagnosis must be achieved with the aid of MRCP.

Detection of acute rib fractures on CT images with convolutional neural networks: effect of location and type of fracture and reader’s experience

The evaluation of all ribs on thin-slice CT images is time consuming and it can be difficult to accurately assess the location and type of rib fracture in an emergency. The aim of our study was to develop and validate a convolutional neural network (CNN) algorithm for the detection of acute rib fractures on thoracic CT images and to investigate the effect of the CNN algorithm on radiologists' performance. The dataset for development of a CNN consisted of 539 thoracic CT scans with 4906 acute rib fractures. A three-dimensional faster region-based CNN was trained and evaluated by using tenfold cross-validation. For an observer performance study to investigate the effect of CNN outputs on radiologists' performance, 30 thoracic CT scans (28 scans with 90 acute rib fractures and 2 without rib fractures) which were not included in the development dataset were used. Observer performance study involved eight radiologists who evaluated CT images first without and second with CNN outputs. The diagnostic performance was assessed by using figure of merit (FOM) values obtained from the jackknife free-response receiver operating characteristic (JAFROC) analysis. When radiologists used the CNN output for detection of rib fractures, the mean FOM value significantly increased for all readers (0.759 to 0.819, P = 0.0004) and for displaced (0.925 to 0.995, P = 0.0028) and non-displaced fractures (0.678 to 0.732, P = 0.0116). At all rib levels except for the 1st and 12th ribs, the radiologists' true-positive fraction of the detection became significantly increased by using the CNN outputs. The CNN specialized for the detection of acute rib fractures on CT images can improve the radiologists' diagnostic performance regardless of the type of fractures and reader's experience. Further studies are needed to clarify the usefulness of the CNN for the detection of acute rib fractures on CT images in actual clinical practice.

Local Gaussian Distribution Fitting Boundary Image Segmentation Algorithm for Ultrasound Images in Avoiding Recurrent Laryngeal Nerve Injury during Thyroid Nodules Treatment

This work aimed to analyze the effect of the boundary segmentation algorithm in thyroid nodules image segmentation and the influence of its adoption in avoiding recurrent laryngeal nerve (RLN) injury during treatment of thyroid nodules. The nodule boundary was extracted aided by the local Gaussian distribution fitting energy (LGDF) segmentation algorithm, which was compared with the normalized cut (Ncut) algorithm and the Canny algorithm. Then, 51 patients treated with microwave ablation for thyroid nodules were taken as a test group, and 51 patients treated with surgical resection were taken as a control group. The incidence of RLN injury and the levels of free triiodothyronine (FT3), free thyroid hormone (FT4), and thyroid-stimulating hormone (TSH) were compared between the two groups before and after treatment. The results showed that the true positive fraction (TPF) of the LGDF segmentation algorithm was 69.45%, the TPF of the Ncut algorithm and the Canny algorithm were 58.65% and 52.37%, respectively. The TPF of LGDF algorithm was higher than that of the Ncut algorithm and Canny algorithm, with notable differences ( P &lt; 0.05 ). In the control group, there were 10 cases of temporary and permanent damage to the RLN after operation, and the total incidence was 19.61%. In the test group, there were 3 cases of temporary and permanent damage to the RLN after operation, and the total incidence was 5.88%, which was lower than that of the control group ( P &lt; 0.05 ). No evident differences were shown in the levels of FT3, FT4, and TSH between the two groups before treatment ( P &gt; 0.05 ). However, after treatment, the TSH level of the test group (4.58 ± 0.79) was higher than that of the control group (3.19 ± 0.17), and the levels of FT3 and FT4 in the test group were lower than those in the control group, and the differences were remarkable ( P &lt; 0.05 ). In short, the LGDF algorithm had more ideal segmentation effect. In addition, ultrasound-guided microwave ablation was effective in treating benign thyroid nodules, which could reduce damage to the RLN and maintain normal thyroid function.

Breast Tumor Detection in MR Images Based on Density

Breast cancer is dangerous in women. It is generally found after the symptoms appear. Detecting the breast cancer at an early stage and understanding the treatment are the most important strategies to prevent death from cancer. Generally, for detection of breast cancer, breast Magnetic Resonance Image (MRI) takes place. It is one of the best approaches to detect tumor in women. In this research paper, a combination of selection methods for seed region growing image segmentation is suggested to detect breast tumor. The suggested method has been divided into following parts: First, the pre-processing of breast image is performed. Second, the automatic threshold for binarization process is calculated. Third, the number of seed points and its position in the breast image are determined automatically using density of pixels value. Fourth, a method for calculation of threshold value is proposed for the purpose of region creation in seed region growing. For the evaluation purpose, the proposed method was applied and tested on the RIDER MRI breast dataset from National Biomedical Imaging Archive (NBIA). After the test was performed, it was observed that proposed algorithm gives 90% accuracy, 88% True Negative Fraction, 91% True Positive Fraction, 10% Misclassification Rate, 94% Precision and 86% Relative Overlap which is better than other existing methods. It not only gives better evaluation measure but also provides segmentation method for multiple tumor detection.

Clinical target volume segmentation for stomach cancer by stochastic width deep neural network

Precise segmentation of clinical target volume (CTV) is the key to stomach cancer radiotherapy. We proposed a novel stochastic width-deep neural network (SW-DNN) for better automatically contouring stomach CTV. Stochastic width-deep neural network was an end-to-end approach, of which the core component was a novel SW mechanism that employed shortcut connections between the encoder and decoder in a random manner, and thus the width of the SW-DNN was stochastically adjustable to obtain improved segmentation results. In total, 150 stomach cancer patient computed tomography (CT) cases with the corresponding CTV labels were collected and used to train and evaluate the SW-DNN. Three common quantitative measures: true positive volume fraction (TPVF), positive predictive value (PPV), and Dice similarity coefficient (DSC) were used to evaluate the segmentation accuracy. Clinical target volumes calculated by SW-DNN had significant quantitative advantages over three state-of-the-art methods. The average DSC value of SW-DNN was 2.1%, 2.8%, and 3.6% higher than that of three state-of-the-art methods. The average DSC, TPVF, and PPV values of SW-DNN were 2.1%, 4.0%, and 0.3% higher than that of the corresponding constant width DNN. Stochastic width-deep neural network provided better performance for contouring stomach cancer CTV accurately and efficiently. It is a promising solution in clinical radiotherapy planning for stomach cancer.

Deep learning of mammary gland distribution for architectural distortion detection in digital breast tomosynthesis

Computer aided detection (CADe) for breast lesions can provide an important reference for radiologists in breast cancer screening. Architectural distortion (AD) is a type of breast lesion that is difficult to detect. A majority of CADe methods focus on detecting the radial pattern, which is a main characteristic of typical ADs. However, a few atypical ADs do not exhibit such a pattern. To improve the performance of CADe for typical and atypical ADs, we propose a deep-learning-based model that used mammary gland distribution as prior information to detect ADs in digital breast tomosynthesis (DBT). First, information about gland distribution, including the Gabor magnitude, the Gabor orientation field, and a convergence map, were produced using a bank of Gabor filters and convergence measures. Then, this prior information and an original slice were input into a Faster R-CNN detection network to obtain the 2-D candidates for each slice. Finally, a 3-D aggregation scheme was employed to fuse these 2-D candidates as 3-D candidates for each DBT volume. Retrospectively, 64 typical AD volumes, 74 atypical AD volumes, and 127 normal volumes were collected. Six-fold cross-validation and mean true positive fraction (MTPF) were used to evaluate the model. Compared to an existing convergence-based model, our proposed model achieved an MTPF of 0.53 ± 0.04, 0.61 ± 0.05, and 0.45 ± 0.04 for all DBT volumes, typical + normal volumes, and atypical + normal volumes, respectively. These results were significantly better than those of 0.36 ± 0.03, 0.46 ± 0.04, and 0.28 ± 0.04 for a convergence-based model (p ≪ 0.01). These results indicate that employing the prior information of gland distribution and a deep learning method can improve the performance of CADe for AD.

Detection and stage classification of Plasmodium falciparum from images of Giemsa stained thin blood films using random forest classifiers

BackgroundThe conventional method for the diagnosis of malaria parasites is the microscopic examination of stained blood films, which is time consuming and requires expertise. We introduce computer-based image segmentation and life stage classification with a random forest classifier. Segmentation and stage classification are performed on a large dataset of malaria parasites with ground truth labels provided by experts.MethodsWe made use of Giemsa stained images obtained from the blood of 16 patients infected with Plasmodium falciparum. Experts labeled the parasite types from each of the images. We applied a two-step approach: image segmentation followed by life stage classification. In segmentation, we classified each pixel as a parasite or non-parasite pixel using a random forest classifier. Performance was evaluated with classification accuracy, Dice coefficient and free-response receiver operating characteristic (FROC) analysis. In life stage classification, we classified each of the segmented objects into one of 8 classes: 6 parasite life stages, early ring, late ring or early trophozoite, mid trophozoite, early schizont, late schizont or segmented, and two other classes, white blood cell or debris.ResultsOur segmentation method gives an average cross-validated Dice coefficient of 0.82 which is a 13% improvement compared to the Otsu method. The Otsu method achieved a True Positive Fraction (TPF) of 0.925 at the expense of a False Positive Rate (FPR) of 2.45. At the same TPF of 0.925, our method achieved an FPR of 0.92, an improvement by more than a factor two. We find that inclusion of average intensity of the whole image as feature for the random forest considerably improves segmentation performance. We obtain an overall accuracy of 58.8% when classifying all life stages. Stages are mostly confused with their neighboring stages. When we reduce the life stages to ring, trophozoite and schizont only, we obtain an accuracy of 82.7%.ConclusionPixel classification gives better segmentation performance than the conventional Otsu method. Effects of staining and background variations can be reduced with the inclusion of average intensity features. The proposed method and data set can be used in the development of automatic tools for the detection and stage classification of malaria parasites. The data set is publicly available as a benchmark for future studies.

RMPPNet: residual multiple pyramid pooling network for subretinal fluid segmentation in SD-OCT images

Automatic assessment of neurosensory retinal detachment (NRD) plays an important role in the diagnosis and treatment for central serous chorioretinopathy (CSC). In this paper, we propose a novel residual multiple pyramid pooling network (RMPPNet) to segment NRD in the spectral-domain optical coherence tomography (SD-OCT) images. Based on the encoder-decoder architecture, RMPPNet can better deal with receptive field and multi-scale features. In the encoder stage, based on the residual architectures, six striding convolutions are utilized to replace the conventional pooling layers to obtain wider receptive fields. To further explore the multi-scale features, three pyramid pooling modules (PPM) are supplemented in the encoder stage. In the decoder stage, we use multiple transpose convolutions to recover the resolution of feature maps and concatenate the feature maps from the encoder for each transpose convolution layer. Finally, for better and faster training, we propose a novel loss function to constrain the different sets between the true label and the prediction label. Three different datasets are utilized to evaluate the proposed model. The first dataset contains 35 cubes from 23 patients, and all the cubes are diagnosed as CSC with only NRD lesions. Based on the first dataset, the second dataset supplements ten normal cubes without NRD lesions. The proposed model obtains a mean dice similarity coefficient 92.6 ± 5.6 and 90.2 ± 20.5, respectively. The last dataset includes 23 cubes from 12 eyes of 12 patients with NRD lesions. The average quantitative results, i.e., mean true positive volume fraction, positive predictive value and dice similarity coefficient, obtained by the proposed model are 96%, 96.45% and 96.2%, respectively. The proposed model can provide a wider receptive field and more abundant multi-scale features to overcome the defects involved in NRD segmentations, such as various size, low contrast, and weak boundaries. Comparing with state-of-the-art methods, the proposed RMPPNet can produce more reliable results for NRD segmentation with higher mean values and lower standard deviations of quantitative criterion, which indicates the practical application for the clinical diagnosis of CSC.