False Positive Per Image Research Articles

GFNet: A Deep Learning Framework for Breast Mass Detection

Background: Breast mass is one of the main symptoms of breast cancer. Effective and accurate detection of breast masses at an early stage would be of great value for clinical breast cancer analysis. Methods: We developed a novel mass detection framework named GFNet. The GFNet is comprised of three modules, including patch extraction, feature extraction, and mass detection. The developed breast mass detection framework is of high robustness and generality that can be self-adapted to images collected by different imaging devices. The patch-based detection is deployed to improve performance. A novel feature extraction technique based on gradient field convergence features (GFCF) is proposed to enhance the information of breast mass and, therefore, provide useful information for the following patch extraction module. A novel false positives reduction method is designed by combining the texture and morphological features in breast mass patch. This is the first attempt at fusing morphological and texture features for breast mass false positive reduction. Results: Compared to other state-of-the-art methods, the proposed GFNet showed the best performance on CBIS-DDSM and INbreast with an accuracy of 0.90 at 2.91 false positive per image (FPI) and 0.99 at only 0.97 FPI, respectively. Conclusions: The GFNet is an effective tool for detecting breast mass.

Breast Masses Detection and Segmentation in Full-Field Digital Mammograms using Unified Convolution Neural Network.

Breast Cancer has been the primary reason for mortality in women of age between twenties and sixties worldwide; moreover early detection and treatment provides patients to get absolute treatment and decrease the mortality rate. Furthermore, recent research indicates that most experienced physicians have plenty of limitations, hence the plethora of work has been carried out to develop an automated mechanism of segmentation and classification of affected area and type of cancer; however, it is still considered to be highly challenging due to the variability of tumor in shape, low signal to noise ratio, shape, size and location of tumor. Furthermore, mammographic mass segmentation and detection are performed as a separate task and a convolution neural network is a highly adopted architecture for the same. In this research, we have designed and developed unified CNN architecture to perform the segmentation and detection of a breast mass. The unified-CNN architecture comprises a novel module for convolution which is combined through additional offset. Further RRS aka Random Region Selection mechanism is applied for data augmentation approach and high-level feature map is implied to achieve the high prediction. Furthermore, unified-CNN is evaluated using the metrics like true positive Rate at FPI (False Positive per Image) and Dice Index on INBreast dataset, also comparative analysis is out carried with various existing methodology. Unified-CNN is developed through improvising CNN. It introduces a novel module at the convolution layer to aim for a high-level feature map in order to get a high prediction. RRS (Random Region Selection) algorithm is used as the data augmentation approach to select the boundary region of the affected area; further robust model training is designed and optimized for process to make optimal. Unified-CNN introduces novel module at the convolution layer to aim for high level feature map in order to get high prediction; further ROI pooling is utilized for boundary detection in images.

Oil spill identification in visible sensor imaging using automated cross correlation with crude oil image filters

An algorithm for detection of crude oil spills in visible light images has been developed and tested on 50 documented crude oil spill images from Shell Petroleum Development Company (SPDC) Nigeria. A set of three 25 x 25 pixels crude oil filters, with unique red, green, and blue (RGB) colour values, homogeneity, and power spectrum density (PSD) features were cross-correlated with the documented spill images. The final crude oil spill Region of Interest (ROI) was determined by grouping interconnected pixels based on their proximity, and only selecting ROIs with an area greater than 5,000 pixels. The crude oil filter cross correlation algorithm demonstrated a sensitivity of 84% with a False Positive per Image (FPI) of 0.82. Future work includes volume estimation of detected spills using crude oil filters, and utilizing this information in the recommendation of appropriate spill clean-up and remediation procedures for the detected spills.
 Keywords: Crude Oil Spill Detection, Crude oil image filters, Cross correlation, Visible sensor imaging, Oil Spill Segmentation.

Deep learning for mass detection in Full Field Digital Mammograms

In recent years, the use of Convolutional Neural Networks (CNNs) in medical imaging has shown improved performance in terms of mass detection and classification compared to current state-of-the-art methods. This paper proposes a fully automated framework to detect masses in Full-Field Digital Mammograms (FFDM). This is based on the Faster Region-based Convolutional Neural Network (Faster-RCNN) model and is applied for detecting masses in the large-scale OPTIMAM Mammography Image Database (OMI-DB), which consists of ∼80,000 FFDMs mainly from Hologic and General Electric (GE) scanners. This research is the first to benchmark the performance of deep learning on OMI-DB. The proposed framework obtained a True Positive Rate (TPR) of 0.93 at 0.78 False Positive per Image (FPI) on FFDMs from the Hologic scanner. Transfer learning is then used in the Faster R-CNN model trained on Hologic images to detect masses in smaller databases containing FFDMs from the GE scanner and another public dataset INbreast (Siemens scanner). The detection framework obtained a TPR of 0.91±0.06 at 1.69 FPI for images from the GE scanner and also showed higher performance compared to state-of-the-art methods on the INbreast dataset, obtaining a TPR of 0.99±0.03 at 1.17 FPI for malignant and 0.85±0.08 at 1.0 FPI for benign masses, showing the potential to be used as part of an advanced CAD system for breast cancer screening.

Detection of partially overlapped masses in mammograms

<span>Breast cancer remains one of the major causes of cancer deaths among women. For decades, screening mammography has been one of the most common methods for early cancer detection and diagnosis. Digital mammography images are created by applying a small burst of x-rays that pass through the breast to a solid-state detector, which transmits the electronic signals to a computer to form a digital image. However, due to projection, some mass areas may be partially covered, which makes them difficult to be interprated. This paper addresses the issue of potential mass regions being distorted by other normal breast tissues, which will negatively affect some of the features being extracted from the mass and in turn deteriorate the classification accuracy. The goal was to estimate the overlapped parts of the mass border using Euclidean distance in order to give more accurate results in next stages. The presented method achieved 95.744% region sensitivity at 0.333 False Positive per Image (FPI), outperforming other researches in this branch of mammography analysis.</span>

Reduction of False-Positive Markings on Mammograms: a Retrospective Comparison Study Using an Artificial Intelligence-Based CAD

The aim was to determine whether an artificial intelligence (AI)-based, computer-aided detection (CAD) software can be used to reduce false positive per image (FPPI) on mammograms as compared to an FDA-approved conventional CAD. A retrospective study was performed on a set of 250 full-field digital mammograms between January 1, 2013, and March 31, 2013, and the number of marked regions of interest of two different systems was compared for sensitivity and specificity in cancer detection. The count of false-positive marks per image (FPPI) of the two systems was also evaluated as well as the number of cases that were completely mark-free. All results showed statistically significant reductions in false marks with the use of AI-CAD vs CAD (confidence interval = 95%) with no reduction in sensitivity. There is an overall 69% reduction in FPPI using the AI-based CAD as compared to CAD, consisting of 83% reduction in FPPI for calcifications and 56% reduction for masses. Almost half (48%) of cases showed no AI-CAD markings while only 17% show no conventional CAD marks. There was a significant reduction in FPPI with AI-CAD as compared to CAD for both masses and calcifications at all tissue densities. A 69% decrease in FPPI could result in a 17% decrease in radiologist reading time per case based on prior literature of CAD reading times. Additionally, decreasing false-positive recalls in screening mammography has many direct social and economic benefits.

Performance of Deep Learning Model in Detecting Operable Lung Cancer With Chest Radiographs.

The aim of this study was to evaluate the diagnostic performance of a trained deep convolutional neural network (DCNN) model for detecting operable lung cancer with chest radiographs (CXRs). The institutional review board approved this study. A deep learning model (DLM) based on DCNN was trained with 17,211 CXRs (5700 CT-confirmed lung nodules in 3500 CXRs and 13,711 normal CXRs), finally augmented to 600,000 images. For validation, a trained DLM was tested with 1483 CXRs with surgically resected lung cancer, marked and scored by 2 radiologists. Furthermore, diagnostic performances of DLM and 6 human observers were compared with 500 cases (200 visible T1 lung cancer on CXR and 300 normal CXRs) and analyzed using free-response receiver-operating characteristics curve (FROC) analysis. The overall detection rate of DLM for resected lung cancers (27.2±14.6 mm) was a sensitivity of 76.8% (1139/1483) with a false positive per image (FPPI) of 0.3 and area under the FROC curve (AUC) of 0.732. In the comparison with human readers, DLM demonstrated a sensitivity of 86.5% at 0.1 FPPI and a sensitivity of 92% at 0.3 FPPI with AUC of 0.899 at an FPPI range of 0.03 to 0.44 for detecting visible T1 lung cancers, which were superior to the average of 6 human readers [mean sensitivity; 78% (range, 71.6% to 82.6%) at an FPPI of 0.1% and 85% (range, 80.2% to 89.2%) at an FPPI of 0.3, AUC of 0.819 (range, 0.754 to 0.862) at an FPPI of 0.03 to 0.44). A DLM has high diagnostic performance in detecting operable lung cancer with CXR, demonstrating a potential of playing a pivotal role for lung cancer screening.

Morphological operation and scaled Réyni entropy based approach for masses detection in mammograms

Detection of suspicious masses in mammograms play a vital role in early diagnosis of breast cancer, to reduce the death rate among women. The presence of masses and calcification’s in mammograms is distinguishing signs for breast cancer diagnosis. But in some cases, due to contrast variation, fuzzy boundaries and presence of noise in mammograms, segmentation and detection of masses is challenging assignment. This paper presents a new segmentation approach to detect masses in mammographic images. The proposed approach consist of artifacts elimination and pectoral region extraction, suspicious mass enhancement using dual morphological operation technique and finally, extraction of Regions of Interest (ROIs) from background using scaled Reyni entropy. The proposed system has been tested on two data-sets i.e. mini- Mammographic Image Analysis Society (mini-MIAS) and Digital Database for Screening Mammography (DDSM), over 50 and 90 mammograms respectively. Performance achieved with proposed system in terms of True Positive Fraction (TPF) yields 93.2% and 93.9% respectively, at the rate of 1.48 and 0.74 average False Positive per Image (FP/I), tested on both Cranio-Caudal (CC) and Medio-Lateral Oblique (MLO) views. The obtained experimental results demonstrates that proposed method gives improved results for mass detection and can be useful for radiologists in diagnosis of breast cancer at early stage.

Breast Mass Detection in Digital Mammogram Based on Gestalt Psychology.

Inspired by gestalt psychology, we combine human cognitive characteristics with knowledge of radiologists in medical image analysis. In this paper, a novel framework is proposed to detect breast masses in digitized mammograms. It can be divided into three modules: sensation integration, semantic integration, and verification. After analyzing the progress of radiologist's mammography screening, a series of visual rules based on the morphological characteristics of breast masses are presented and quantified by mathematical methods. The framework can be seen as an effective trade-off between bottom-up sensation and top-down recognition methods. This is a new exploratory method for the automatic detection of lesions. The experiments are performed on Mammographic Image Analysis Society (MIAS) and Digital Database for Screening Mammography (DDSM) data sets. The sensitivity reached to 92% at 1.94 false positive per image (FPI) on MIAS and 93.84% at 2.21 FPI on DDSM. Our framework has achieved a better performance compared with other algorithms.

A cost-sensitive Bayesian combiner for reducing false positives in mammographic mass detection.

Mammography is the most widely used modality for early breast cancer detection. This work proposes a new computer-aided mass detection approach, in which a denoising method called BM3D is first applied to mammograms. Afterwards, using an adaptive segmentation algorithm, images are segmented to suspicious regions of interest (ROIs) and then a classifier is used to understand the features of true positive (TP) and false positive (FP) patterns. In this way, from selected suspicious ROIs, fractal dimension, texture and intensity features are extracted. Subsequently, a discretization approach followed by correlation-based feature selection (CFS) is combined with a genetic algorithm to obtain the most representative features. To neutralize the classifier's bias in favor of the major class in imbalanced datasets, an oversampling algorithm is used. In the next step, a cost-sensitive ensemble classifier based on a trainable combiner is proposed in order to reduce the number of FP samples. Finally, the presented method is validated on miniMIAS and INBreast datasets. The free-response receiver operating characteristic (FROC) analysis results prove the efficiency of the proposed approach. A sensitivity of 88% and false positive per image (FPpI) of 0.78 for miniMIAS and also a sensitivity of 86% and FPpI of 0.75 for INBreast dataset were obtained.

Rapid Pedestrian Detection Based on Motion Information and Static Low-Level Feature

This thesis introduces a pedestrian detection algorithm of medium-low resolution rate in complicated environment. This method uses quick feature pyramid to extract static pedestrian pattern and calculates pedestrian motion feature through Rectangular Block Pixel Mean Value of Difference, and finally fusions both patterns at the stage of classification and identification. With False Positive per Image (FPPI) being 1 and false detection ratio being 14%, this method proves to be effective in reducing the out ratio.

Mass Detection in Mammographic Images Using Wavelet Processing and Adaptive Threshold Technique.

Detection of mass in mammogram for early diagnosis of breast cancer is a significant assignment in the reduction of the mortality rate. However, in some cases, screening of mass is difficult task for radiologist, due to variation in contrast, fuzzy edges and noisy mammograms. Masses and micro-calcifications are the distinctive signs for diagnosis of breast cancer. This paper presents, a method for mass enhancement using piecewise linear operator in combination with wavelet processing from mammographic images. The method includes, artifact suppression and pectoral muscle removal based on morphological operations. Finally, mass segmentation for detection using adaptive threshold technique is carried out to separate the mass from background. The proposed method has been tested on 130 (45 + 85) images with 90.9 and 91 % True Positive Fraction (TPF) at 2.35 and 2.1 average False Positive Per Image(FP/I) from two different databases, namely Mammographic Image Analysis Society (MIAS) and Digital Database for Screening Mammography (DDSM). The obtained results show that, the proposed technique gives improved diagnosis in the early breast cancer detection.

Computer-assisted grading of diabetic macular edema on retinal color fundus images.

Diabetic macular edema (DME) is one of the vision-impairing manifestations of Diabetic Retinopathy (DR). Early detection and treatment of DME can prevent permanent vision loss in people suffering from DR. However, the clinical detection through biomicroscopy is time-consuming. In this paper, a computer-assisted grading method has been proposed to determine the DME severity based on the spatial distribution of exudative lesions around macula. The region around macula is classified into zonal levels and severity of the DME is graded based on the presence of exudative lesions in each zone. The proposed method has been evaluated on diverse public and local databases, and produced the sensitivity of 89.54% for 9.1 false positive per image (FPPI) for exudate detection and 98.8% accuracy for DME grading.

Scene-Specific Pedestrian Detection for Static Video Surveillance

The performance of a generic pedestrian detector may drop significantly when it is applied to a specific scene due to the mismatch between the source training set and samples from the target scene. We propose a new approach of automatically transferring a generic pedestrian detector to a scene-specific detector in static video surveillance without manually labeling samples from the target scene. The proposed transfer learning framework consists of four steps. 1) Through exploring the indegrees from target samples to source samples on a visual affinity graph, the source samples are weighted to match the distribution of target samples. 2) It explores a set of context cues to automatically select samples from the target scene, predicts their labels, and computes confidence scores to guide transfer learning. 3) The confidence scores propagate among target samples according to their underlying visual structures. 4) Target samples with higher confidence scores have larger influence on training scene-specific detectors. All these considerations are formulated under a single objective function called confidence-encoded SVM, which avoids hard thresholding on confidence scores. During test, only the appearance-based detector is used without context cues. The effectiveness is demonstrated through experiments on two video surveillance data sets. Compared with a generic detector, it improves the detection rates by 48 and 36 percent at one false positive per image (FPPI) on the two data sets, respectively. The training process converges after one or two iterations on the data sets in experiments.

Automated Detection of Architectural Detection in Mammograms Using Template Matching

Breast Cancer is one of the leading causes of death, and its early detection increases the survival rate and treatment options available to patients. Computer-Aided-Detection (CAD) systems have been developed to assist radiologists with the task of locating cancer in mammograms. Unfortunately, these CAD systems have demonstrated less than 50% efficiency in detecting Architectural Distortions (AD), which is a sign of breast cancer. This paper presents a method of detecting AD with better sensitivity results. Forty mammograms containing AD were obtained from the Digital Database for Screening Mammography (DDSM). Each mammogram was preprocessed using breast segmentation techniques to extract the breast region from the mammogram. The mammograms were enhanced using contrast-limited adaptive histogram equalization (CLAHE). Next, the enhanced mammograms were filtered with a bank of 180 Gabor filters to extract the texture orientation from the images. Based on the fact that ADs contain spicules radiating in all direction, AD templates were designed in MATLAB. These templates were cross-correlated with the Gabor filtered mammograms to obtain ROIs that were most likely to contain ADs. The developed algorithm is intended to assist the radiologist by flagging regions likely to contain AD, prompting the radiologist to take a closer look at those regions. The current algorithm developed in MATLAB automatically flags seven suspicious blocks of 25 by 25 pixels per image, and demonstrated a sensitivity of 87.5% with a False Positive per Image (FPI) of 5.2. Future work will focus on the reduction of FPI.

多部位集合的人体检测

A part-based mixture model(Poselets)was ultilized to alleviate the problem of pedestrian detection under occlusion and articulation.Firstly,Procrustes distances between similar pedestrian parts were calculated to gather training exmaples with the same configurations.Then,common parts of frontal pedestrian described by Histogram of Oriented Gradient(HOG)were trained to get discriminative models.A test was carried out on a validation set to find out which parts were more accurate,and those pictures which were not detected formed the hard set.Other complementary parts were explored on the hard set,afterwards,all these parts formed the final Poselets detector.While detecting,Poselets were clusterd if the symmetrized KL-divergence between two Poselet activations were small.The bounding box of a pedestrian was inferred by the bounding boxes of the Poselets in a cluster.Based on the test of the INRIA pedestrian dataset,it is showed that the detection rate increases from 77% of Poselets to81%in this paper while False Positive Per Image(FPPI)is 0.5.It is concluded that the part selection mechanism proposed in this paper promotes the detection rate with speedup rate about twice of the traditional method,meanwhile it reduces the number of models.

Breast masses detection using phase portrait analysis and fuzzy inference systems

Breast masses exhibit variability in margins, shapes, and dimensions, so their detection is a difficult task in mammographic computer-aided diagnosis. Mass detection is usually a two-step procedure: mass identification and false-positive reduction. A new method to automatically detect mass lesions in mammographic images with tuning according to the breast tissue density was developed and tested. A modified phase portrait analysis method was introduced, based on the eigenvalue condition number and an eigenvalue intensity map. The method uses an iterative and tissue density-adaptive segmentation procedure with extraction of geometric features. False-positive reduction is accomplished using a fuzzy inference-based classifier. A leave-one-image-out cross-validation procedure was implemented, and stepwise regression analysis was used to automatically extract an optimal set of features. Testing and validation were performed on two different data sets containing at least one malignant mass D1 (388 images) and D2 (674 images), and a third data set N1 (50 images) was used consisting of normal controls. These three data sets were taken from the Digital Database for Screening Mammography. For sensitivities of 0.9, 0.85, 0.80, and 0.75, the best results on cancer images exhibit an False-Positive per Image (FPpI) equal to 0.6, 0.45, 0.35, and 0.3, respectively, using a Bayes Linear Discriminant Analysis (LDA) classifier and an FPpI of 0.85, 0.7, 0.55, and 0.45 using a fuzzy inference system (FIS) for false-positive reduction. When the algorithm is tested on normal images, an FPpI equal to 0.4, 0.3, 0.25, and 0.2 was observed using LDA and 0.3, 0.25, 0.2, and 0.15 using the FIS. A preclinical study of an automatic breast mass detection algorithm provided promising results in terms of sensitivity and low false-positive rate. Further development and clinical testing are justified based on the results.

Detection of Cancerous Masses in Mammograms by Template Matching: Optimization of Template Brightness Distribution by Means of Evolutionary Algorithm

Optimization of brightness distribution in the template used for detection of cancerous masses in mammograms by means of correlation coefficient is presented. This optimization is performed by the evolutionary algorithm using an auxiliary mass classifier. Brightness along the radius of the circularly symmetric template is coded indirectly by its second derivative. The fitness function is defined as the area under curve (AUC) of the receiver operating characteristic (ROC) for the mass classifier. The ROC and AUC are obtained for a teaching set of regions of interest (ROIs), for which it is known whether a ROI is true-positive (TP) or false-positive (F). The teaching set is obtained by running the mass detector using a template with a predetermined brightness. Subsequently, the evolutionary algorithm optimizes the template by classifying masses in the teaching set. The optimal template (OT) can be used for detection of masses in mammograms with unknown ROIs. The approach was tested on the training and testing sets of the Digital Database for Screening Mammography (DDSM). The free-response receiver operating characteristic (FROC) obtained with the new mass detector seems superior to the FROC for the hemispherical template (HT). Exemplary results are the following: in the case of the training set in the DDSM, the true-positive fraction (TPF) = 0.82 for the OT and 0.79 for the HT; in the case of the testing set, TPF = 0.79 for the OT and 0.72 for the HT. These values were obtained for disease cases, and the false-positive per image (FPI) = 2.