Average Number Of False Positives Research Articles

Low Latency Signature-Based 4-Step Random Access Procedure for Massive Machine-Type Communications

Massive machine-type communication (mMTC) is one of the major services in the 5G new radio (NR) system. In the mMTC application scenarios, many machine-type communication devices (MTCDs) communicate with each other or through an MTC server. However, how to solve the problem of excessive MTCDs accessing the network simultaneously is an important task. To overcome the large number of random access collisions and retransmission problems, this paper proposes a method to solve the mMTC random access problem. The proposed method consists of two parts, modified Contention Based Random Access (CBRA) with 4-step RA type and 4-step Signature RA type. With the modified CBRA with 4-step RA type, each MTCD will obtain the information to generate the signature. After that, the 4-step Signature RA type is used for MTCDs to proceed with the random access with the gNB through the generated signatures. In our proposed method, the average number of false positives and average process time are evaluated under different numbers of active MTCDs and different parameter values for the signature generation. Besides, the average number of preamble attempts and average process time are used as the metrics to compare the proposed scheme with the 5G NR 4-step RA type. Finally, simulation results show that the proposed scheme performs better than the 5G NR 4-step RA type in the mMTC scenarios.

Feasibility of a deep learning artificial intelligence model for the diagnosis of pediatric ileocolic intussusception with grayscale ultrasonography.

This study explored the feasibility of utilizing a deep learning artificial intelligence (AI) model to detect ileocolic intussusception on grayscale ultrasound images. This retrospective observational study incorporated ultrasound images of children who underwent emergency ultrasonography for suspected ileocolic intussusception. After excluding video clips, Doppler images, and annotated images, 40,765 images from two tertiary hospitals were included (positive-to-negative ratio: hospital A, 2,775:35,373; hospital B, 140:2,477). Images from hospital A were split into a training set, a tuning set, and an internal test set (ITS) at a ratio of 7:1.5:1.5. Images from hospital B comprised an external test set (ETS). For each image indicating intussusception, two radiologists provided a bounding box as the ground-truth label. If intussusception was suspected in the input image, the model generated a bounding box with a confidence score (0-1) at the estimated lesion location. Average precision (AP) was used to evaluate overall model performance. The performance of practical thresholds for the modelgenerated confidence score, as determined from the ITS, was verified using the ETS. The AP values for the ITS and ETS were 0.952 and 0.936, respectively. Two confidence thresholds, CTopt and CTprecision, were set at 0.557 and 0.790, respectively. For the ETS, the perimage precision and recall were 95.7% and 80.0% with CTopt, and 98.4% and 44.3% with CTprecision. For per-patient diagnosis, the sensitivity and specificity were 100.0% and 97.1% with CTopt, and 100.0% and 99.0% with CTprecision. The average number of false positives per patient was 0.04 with CTopt and 0.01 for CTprecision. The feasibility of using an AI model to diagnose ileocolic intussusception on ultrasonography was demonstrated. However, further study involving bias-free data is warranted for robust clinical validation.

Automated Rib Fracture Detection on Chest X-Ray Using Contrastive Learning

To develop a deep learning-based model for detecting rib fractures on chest X-Ray and to evaluate its performance based on a multicenter study. Chest digital radiography (DR) images from 18,631 subjects were used for the training, testing, and validation of the deep learning fracture detection model. We first built a pretrained model, a simple framework for contrastive learning of visual representations (simCLR), using contrastive learning with the training set. Then, simCLR was used as the backbone for a fully convolutional one-stage (FCOS) objective detection network to identify rib fractures from chest X-ray images. The detection performance of the network for four different types of rib fractures was evaluated using the testing set. A total of 127 images from Data-CZ and 109 images from Data-CH with the annotations for four types of rib fractures were used for evaluation. The results showed that for Data-CZ, the sensitivities of the detection model with no pretraining, pretrained ImageNet, and pretrained DR were 0.465, 0.735, and 0.822, respectively, and the average number of false positives per scan was five in all cases. For the Data-CH test set, the sensitivities of three different pretraining methods were 0.403, 0.655, and 0.748. In the identification of four fracture types, the detection model achieved the highest performance for displaced fractures, with sensitivities of 0.873 and 0.774 for the Data-CZ and Data-CH test sets, respectively, with 5 false positives per scan, followed by nondisplaced fractures, buckle fractures, and old fractures. A pretrained model can significantly improve the performance of the deep learning-based rib fracture detection based on X-ray images, which can reduce missed diagnoses and improve the diagnostic efficacy.

Knowledge-guided 2.5D CNN for cerebral microbleeds detection

Cerebral microbleed (CMB) based on magnetic resonance imaging has been recently investigated as key biomarker in the diagnosis of cerebral small-vessel diseases and vascular cognitive impairment. Because the CMB lesions are typically small in size, and easily confused with various analogs such as calcified deposits, artifacts, and especially blood vessels when they are observed from a single MRI slice, reducing false positives in CMB detection is quite challenging. In addition, the lack of available medical image data, which inevitably leads to the imbalance between positive and negative samples, is also a challenge to existing deep learning algorithms. To address these problems, this paper proposes a simple but effective CMB detection method based on a novel deep architecture. First, in contrast to the current local patches-based approach, we make full use of the information about the distribution of CMBs in the whole brain based on training data as priori knowledge to guide the model to obtain candidate CMB patches. Second, we propose a 2.5D convolutional neural network based on morphological differences in cerebral blood vessels and cerebral microbleeds. Specifically, we further use information of the candidates in the coronal and sagittal planes and combine the inference based on three planes to determine the CMB probability of each patch. This approach strikes a balance between the high computational cost and the loss of spatial information. The effectiveness of the proposed method is demonstrated through experimental results that show that our KBPNet model has a sensitivity of 98.24%, an accuracy of 94.10% and an average number of false positives per patient of 1.72 on the SWI-CMB dataset.

Cervical spine fracture detection in computed tomography using convolutional neural networks

Objective. In the context of primary in-hospital trauma management timely reading of computed tomography (CT) images is critical. However, assessment of the spine is time consuming, fractures can be very subtle, and the potential for under-diagnosis or delayed diagnosis is relevant. Artificial intelligence is increasingly employed to assist radiologists with the detection of spinal fractures and prioritization of cases. Currently, algorithms focusing on the cervical spine are commercially available. A common approach is the vertebra-wise classification. Instead of a classification task, we formulate fracture detection as a segmentation task aiming to find and display all individual fracture locations presented in the image. Approach. Based on 195 CT examinations, 454 cervical spine fractures were identified and annotated by radiologists at a tertiary trauma center. We trained for the detection a U-Net via four-fold-cross validation to segment spine fractures and the spine via a multi-task loss. We further compared advantages of two image reformation approaches—straightened curved planar reformatted (CPR) around the spine and spinal canal aligned volumes of interest (VOI)—to achieve a unified vertebral alignment in comparison to processing the Cartesian data directly. Main results. Of the three data versions (Cartesian, reformatted, VOI) the VOI approach showed the best detection rate and a reduced computation time. The proposed algorithm was able to detect 87.2% of cervical spine fractures at an average number of false positives of 3.5 per case. Evaluation of the method on a public spine dataset resulted in 0.9 false positive detections per cervical spine case. Significance. The display of individual fracture locations as provided with high sensitivity by the proposed voxel classification based fracture detection has the potential to support the trauma CT reading workflow by reducing missed findings.

Artificial intelligence-powered software detected more than half of the liver metastases overlooked by radiologists on contrast-enhanced CT

PurposeTo evaluate the sensitivity of artificial intelligence (AI)-powered software in detecting liver metastases, especially those overlooked by radiologists. MethodsRecords of 746 patients diagnosed with liver metastases (November 2010–September 2017) were reviewed. Images from when radiologists first diagnosed liver metastases were reviewed, and prior contrast-enhanced CT (CECT) images were checked for availability. Two abdominal radiologists classified the lesions into overlooked lesions (all metastases missed by radiologists on prior CECT) and detected lesions (all metastases if any of them were correctly identified and invisible on prior CECT or those with no prior CECT). Finally, images from 137 patients were identified, 68 of which were classified as “overlooked cases.” The same radiologists created the ground truth for these lesions and compared them with the software’s output at 2-month intervals. The primary endpoint was the sensitivity in detecting all liver lesion types, liver metastases, and liver metastases overlooked by radiologists. ResultsThe software successfully processed images from 135 patients. The per-lesion sensitivity for all liver lesion types, liver metastases, and liver metastases overlooked by radiologists was 70.1%, 70.8%, and 55.0%, respectively. The software detected liver metastases in 92.7% and 53.7% of patients in detected and overlooked cases, respectively. The average number of false positives was 0.48 per patient. ConclusionThe AI-powered software detected more than half of liver metastases overlooked by radiologists while maintaining a relatively low number of false positives. Our results suggest the potential of AI-powered software in reducing the frequency of overlooked liver metastases when used in conjunction with the radiologists’ clinical interpretation.

MSANet: Multiscale Aggregation Network Integrating Spatial and Channel Information for Lung Nodule Detection.

Improving the detection accuracy of pulmonary nodules plays an important role in the diagnosis and early treatment of lung cancer. In this paper, a multiscale aggregation network (MSANet), which integrates spatial and channel information, is proposed for 3D pulmonary nodule detection. MSANet is designed to improve the network's ability to extract information and realize multiscale information fusion. First, multiscale aggregation interaction strategies are used to extract multilevel features and avoid feature fusion interference caused by large resolution differences. These strategies can effectively integrate the contextual information of adjacent resolutions and help to detect different sized nodules. Second, the feature extraction module is designed for efficient channel attention and self-calibrated convolutions (ECA-SC) to enhance the interchannel and local spatial information. ECA-SC also recalibrates the features in the feature extraction process, which can realize adaptive learning of feature weights and enhance the information extraction ability of features. Third, the distribution ranking (DR) loss is introduced as the classification loss function to solve the problem of imbalanced data between positive and negative samples. The proposed MSANet is comprehensively compared with other pulmonary nodule detection networks on the LUNA16 dataset, and a CPM score of 0.920 is obtained. The results show that the sensitivity for detecting pulmonary nodules is improved and that the average number of false-positives is effectively reduced. The proposed method has advantages in pulmonary nodule detection and can effectively assist radiologists in pulmonary nodule detection.

Freely available artificial intelligence for pelvic lymph node metastases in PSMA PET-CT that performs on par with nuclear medicine physicians

PurposeThe aim of this study was to develop and validate an artificial intelligence (AI)-based method using convolutional neural networks (CNNs) for the detection of pelvic lymph node metastases in scans obtained using [18F]PSMA-1007 positron emission tomography-computed tomography (PET-CT) from patients with high-risk prostate cancer. The second goal was to make the AI-based method available to other researchers.Methods[18F]PSMA PET-CT scans were collected from 211 patients. Suspected pelvic lymph node metastases were marked by three independent readers. A CNN was developed and trained on a training and validation group of 161 of the patients. The performance of the AI method and the inter-observer agreement between the three readers were assessed in a separate test group of 50 patients.ResultsThe sensitivity of the AI method for detecting pelvic lymph node metastases was 82%, and the corresponding sensitivity for the human readers was 77% on average. The average number of false positives was 1.8 per patient. A total of 5–17 false negative lesions in the whole cohort were found, depending on which reader was used as a reference. The method is available for researchers at www.recomia.org.ConclusionThis study shows that AI can obtain a sensitivity on par with that of physicians with a reasonable number of false positives. The difficulty in achieving high inter-observer sensitivity emphasizes the need for automated methods. On the road to qualifying AI tools for clinical use, independent validation is critical and allows performance to be assessed in studies from different hospitals. Therefore, we have made our AI tool freely available to other researchers.

LungSeek: 3D Selective Kernel residual network for pulmonary nodule diagnosis.

Early detection and diagnosis of pulmonary nodules is the most promising way to improve the survival chances of lung cancer patients. This paper proposes an automatic pulmonary cancer diagnosis system, LungSeek. LungSeek is mainly divided into two modules: (1) Nodule detection, which detects all suspicious nodules from computed tomography (CT) scan; (2) Nodule Classification, classifies nodules as benign or malignant. Specifically, a 3D Selective Kernel residual network (SK-ResNet) based on the Selective Kernel Network and 3D residual network is located. A deep 3D region proposal network with SK-ResNet is designed for detection of pulmonary nodules while a multi-scale feature fusion network is designed for the nodule classification. Both networks use the SK-Net module to obtain different receptive field information, thereby effectively learning nodule features and improving diagnostic performance. Our method has been verified on the luna16 data set, reaching 89.06, 94.53% and 97.72% when the average number of false positives is 1, 2 and 4, respectively. Meanwhile, its performance is better than the state-of-the-art method and other similar networks and experienced doctors. This method has the ability to adaptively adjust the receptive field according to multiple scales of the input information, so as to better detect nodules of various sizes. The framework of LungSeek based on 3D SK-ResNet is proposed for nodule detection and nodule classification from chest CT. Our experimental results demonstrate the effectiveness of the proposed method in the diagnosis of pulmonary nodules.

Deep Learning-Based Automatic Detection and Segmentation of Gross Tumor for Stereotactic Ablative Radiotherapy in Small-Volume Brain Metastases

<h3>Purpose/Objective(s)</h3> Brain metastases (BM) are the most common intracranial tumors in adults. Benefiting from efficient local control, stereotactic ablative radiotherapy (SABR) is becoming popular in the treatment of BM. Detection and manual delineation of BM are labor-intense processes, and BM are occasionally mistaken for normal brain, especially if volumes are very small. Recently, several efforts have been made to develop automatic detection and segmentation for BM using advanced deep learning (DL) algorithms. In this study, we investigated the efficacy and accuracy of a DL model for the detection and segmentation of BM. <h3>Materials/Methods</h3> To train and validate the DL model, magnetic resonance imaging (MRI) for 65 patients and 603 BM were collected and analyzed. Manual contouring for BM was done by a radiation oncologist. We utilized 2.5-dimensional training using a two-dimensional U-Net as the DL architecture to avoid inefficiencies of the three-dimensional architecture. Because of MRI characteristic, we applied N4 bias field correction and gamma correction as pre-processing techniques. Also, we utilized the overlapping patch technique and under-sampling technique to extract small-volume BM from the normal brain that occupies most of the slice. During the training, data augmentations (horizontal flip, vertical flip, random rotation, random blur) were applied on the fly randomly. The detection performance was measured with sensitivity and average false positive rate and the segmentation performance with the Dice coefficient with dilation (DWD) with dilation factor set to a radius of 4 pixels. BMs with volumes < 0.04 cc were considered small-volume BM. <h3>Results</h3> As a test-set, 12 patients with 58 BM were randomly selected to evaluate the performance of the model. In the test-set, 41.2% BM had a volume of < 0.04 cc, Max volume of 1.219 cc, min volume of 0.021 cc, median volume of 0.068 cc, and mean volume of 0.158 cc. The DL showed 100% detection sensitivity for BM with volumes ≥0.04 cc, and 85.7% sensitivity for BM with volumes < 0.04 cc volume, showing a sensitivity of 97%, and the average number of false positives was 2.08 per patient. Out of the 12 test-set patients, 9 (75%) had less than 10 BM and 3 (25%) had 10 or more BM. In patients with less than 10 BM, the detection sensitivity was 100%. In patients with 10 or more BM, the overall detection sensitivity was 93.5%. The detection sensitivity was 100% and 86% for BM with volumes ≥0.04 cc and < 0.04 cc, respectively. DWD showed a value of over 67% in BM with volumes > 0.04 cc and 49.7% in BM with volumes < 0.04 cc. For all BM, the DWD value was 63%. <h3>Conclusion</h3> Our DL model can detect and segment BM with good performance; it is especially effective even for small-volume BM, suggesting its considerable benefit for gross tumor volume detection and segmentation for stereotactic radiosurgery or SABR.

Drug Target Identification with Machine Learning: How to Choose Negative Examples.

Identification of the protein targets of hit molecules is essential in the drug discovery process. Target prediction with machine learning algorithms can help accelerate this search, limiting the number of required experiments. However, Drug-Target Interactions databases used for training present high statistical bias, leading to a high number of false positives, thus increasing time and cost of experimental validation campaigns. To minimize the number of false positives among predicted targets, we propose a new scheme for choosing negative examples, so that each protein and each drug appears an equal number of times in positive and negative examples. We artificially reproduce the process of target identification for three specific drugs, and more globally for 200 approved drugs. For the detailed three drug examples, and for the larger set of 200 drugs, training with the proposed scheme for the choice of negative examples improved target prediction results: the average number of false positives among the top ranked predicted targets decreased, and overall, the rank of the true targets was improved.Our method corrects databases’ statistical bias and reduces the number of false positive predictions, and therefore the number of useless experiments potentially undertaken.

A Two Cascaded Network Integrating Regional-based YOLO and 3D-CNN for Cerebral Microbleeds Detection.

Cerebral Microbleeds (CMBs) are small chronic brain hemorrhages, which have been considered as diagnostic indicators for different cerebrovascular diseases including stroke, dysfunction, dementia, and cognitive impairment. In this paper, we propose a fully automated two-stage integrated deep learning approach for efficient CMBs detection, which combines a regional-based You Only Look Once (YOLO) stage for potential CMBs candidate detection and three-dimensional convolutional neural networks (3D-CNN) stage for false positives reduction. Both stages are conducted using the 3D contextual information of microbleeds from the MR susceptibility-weighted imaging (SWI) and phase images. However, we average the adjacent slices of SWI and complement the phase images independently and utilize them as a two- channel input for the regional-based YOLO method. The results in the first stage show that the proposed regional-based YOLO efficiently detected the CMBs with an overall sensitivity of 93.62% and an average number of false positives per subject (FPavg) of 52.18 throughout the five-folds cross-validation. The 3D-CNN based second stage further improved the detection performance by reducing the FPavg to 1.42. The outcomes of this work might provide useful guidelines towards applying deep learning algorithms for automatic CMBs detection.

Automatic detection of pulmonary nodules in CT images based on 3D Res-I network

It is difficult for the existing detection methods of the pulmonary nodules to take into account the global and local features simultaneously. It will lead to over-fitting and lower sensitivity since the extracted features of 3D pulmonary nodules is too complex. To solve these problems, a model based an improved 3D residual structure (3D Res-I) was proposed to detect pulmonary nodules. In the model, the basic residual structure is improved by using rectangular convolution kernel, grouping convolution and pre-activation. Rectangular convolution kernel expands the receptive filed of the convolution, which effectively takes into account the global and local features of the pulmonary nodules. Grouping convolution reduces the computational cost of the model. Pre-activation operation alleviates over-fitting phenomenon. 3D Res-I structure is combined with the improved U-Net network as the feature extraction network of Faster R-CNN. The experimental results on LUNA16 dataset show that the proposed model improves the detection accuracy of pulmonary nodules and reduces the average number of false positives and the size of the generated model.

A 2-Phase Merge Filter Approach to Computer-Aided Detection of Breast Tumors on 3-Dimensional Ultrasound Imaging.

The role of image analysis in 3-dimensional (3D) automated breast ultrasound (ABUS) images is increasingly important because of its widespread use as a screening tool in whole-breast examinations. However, reviewing a large number of images acquired from ABUS is time-consuming and sometimes error prone. The aim of this study, therefore, was to develop an efficient computer-aided detection (CADe) algorithm to assist the review process. The proposed CADe algorithm consisted of 4 major steps. First, initial tumor candidates were formed by extracting and merging hypoechoic square cells on 2-dimensional (2D) transverse images. Second, a feature-based classifier was then constructed using 2D features to filter out nontumor candidates. Third, the remaining 2D candidates were merged longitudinally into 3D masses. Finally, a 3D feature-based classifier was used to further filter out nontumor masses to obtain the final detected masses. The proposed method was validated with 176 passes of breast images acquired by an Acuson S2000 automated breast volume scanner (Siemens Medical Solutions USA, Inc., Malvern, PA), including 44 normal passes and 132 abnormal passes containing 162 proven lesions (79 benign and 83 malignant). The proposed CADe system could achieve overall sensitivity of 100% and 90% with 6.71 and 5.14 false-positives (FPs) per pass, respectively. Our results also showed that the average number of FPs per normal pass (7.16) was more than the number of FPs per abnormal pass (6.56) at 100% sensitivity. The proposed CADe system has a great potential for becoming a good companion tool with ABUS imaging by ensuring high sensitivity with a relatively small number of FPs.

Automated detection of cerebral microbleeds in MR images: A two-stage deep learning approach

Cerebral Microbleeds (CMBs) are small chronic brain hemorrhages, which have been considered as diagnostic indicators for different cerebrovascular diseases including stroke, dysfunction, dementia, and cognitive impairment. However, automated detection and identification of CMBs in Magnetic Resonance (MR) images is a very challenging task due to their wide distribution throughout the brain, small sizes, and the high degree of visual similarity between CMBs and CMB mimics such as calcifications, irons, and veins. In this paper, we propose a fully automated two-stage integrated deep learning approach for efficient CMBs detection, which combines a regional-based You Only Look Once (YOLO) stage for potential CMBs candidate detection and three-dimensional convolutional neural networks (3D-CNN) stage for false positives reduction. Both stages are conducted using the 3D contextual information of microbleeds from the MR susceptibility-weighted imaging (SWI) and phase images. However, we average the adjacent slices of SWI and complement the phase images independently and utilize them as a two-channel input for the regional-based YOLO method. This enables YOLO to learn more reliable and representative hierarchal features and hence achieve better detection performance. The proposed work was independently trained and evaluated using high and low in-plane resolution data, which contained 72 subjects with 188 CMBs and 107 subjects with 572 CMBs, respectively. The results in the first stage show that the proposed regional-based YOLO efficiently detected the CMBs with an overall sensitivity of 93.62% and 78.85% and an average number of false positives per subject (FPavg) of 52.18 and 155.50 throughout the five-folds cross-validation for both the high and low in-plane resolution data, respectively. These findings outperformed results by previously utilized techniques such as 3D fast radial symmetry transform, producing fewer FPavg and lower computational cost. The 3D-CNN based second stage further improved the detection performance by reducing the FPavg to 1.42 and 1.89 for the high and low in-plane resolution data, respectively. The outcomes of this work might provide useful guidelines towards applying deep learning algorithms for automatic CMBs detection.

A total variation-undecimated wavelet approach to chest radiograph image enhancement

Most often medical images such as X-Rays have a low dynamic range and many of their targeted features are difficult to identify. Intensity transformations that improve image quality usually rely onwavelet denoising and enhancement typically use the technique of thresholding to obtain better quality medical images. A disadvantage of wavelet thresholding is that even though it adequately removes noise in an image, it introduces unwanted artifacts into the image near discontinuities. We utilize a total variation method and an undecimated wavelet image enhancing algorithm for improving the image quality of chest radiographs. Our approach achieves a high level chest radiograph image deniosing in lung nodules detection while preserving the important features. Moreover, our method results in a high image sensitivity that reduces the average number of false positives on a test set of medical data.

A CAD system for pulmonary nodule prediction based on deep three-dimensional convolutional neural networks and ensemble learning.

BackgroundDetection of pulmonary nodules is an important aspect of an automatic detection system. Incomputer-aided diagnosis (CAD) systems, the ability to detect pulmonary nodules is highly important, which plays an important role in the diagnosis and early treatment of lung cancer. Currently, the detection of pulmonary nodules depends mainly on doctor experience, which varies. This paper aims to address the challenge of pulmonary nodule detection more effectively.MethodsA method for detecting pulmonary nodules based on an improved neural network is presented in this paper. Nodules are clusters of tissue with a diameter of 3 mm to 30 mm in the pulmonary parenchyma. Because pulmonary nodules are similar to other lung structures and have a low density, false positive nodules often occur. Thus, our team proposed an improved convolutional neural network (CNN) framework to detect nodules. First, a nonsharpening mask is used to enhance the nodules in computed tomography (CT) images; then, CT images of 512×512 pixels are segmented into smaller images of 96×96 pixels. Second, in the 96×96 pixel images which contain or exclude pulmonary nodules, the plaques corresponding to positive and negative samples are segmented. Third, CT images segmented into 96×96 pixels are down-sampled to 64×64 and 32×32 size respectively. Fourth, an improved fusion neural network structure is constructed that consists of three three-dimensional convolutional neural networks, designated as CNN-1, CNN-2, and CNN-3, to detect false positive pulmonary nodules. The networks’ input sizes are 32×32×32, 64×64×64, and 96×96×96 and include 5, 7, and 9 layers, respectively. Finally, we use the AdaBoost classifier to fuse the results of CNN-1, CNN-2, and CNN-3. We call this new neural network framework the Amalgamated-Convolutional Neural Network (A-CNN) and use it to detect pulmonary nodules.FindingsOur team trained A-CNN using the LUNA16 and Ali Tianchi datasets and evaluated its performance using the LUNA16 dataset. We discarded nodules less than 5mm in diameter. When the average number of false positives per scan was 0.125 and 0.25, the sensitivity of A-CNN reached as high as 81.7% and 85.1%, respectively.

PNS194 COMPARISON OF SIMULATED CLINICIAN RANKINGS WITH SEMI-AUTOMATED AND AUTOMATED ALGORITHMS FOR VARIABLE SELECTION WHEN CONDUCTING ANCHORED MATCHING-ADJUSTED INDIRECT COMPARISONS

Anchored matching-adjusted indirect comparisons (MAICs) leverage individual patient-level data (IPD) to compare two treatments across separate trials with a common comparator arm. MAICs only require adjustments for effect modifiers, typically identified by expert clinicians a priori. Emerging semi-automated and automated variable selection algorithms may have advantages over clinician rankings in terms of considering effect modifiers on appropriate scale, reproducibility, being data driven, and ability to leverage large datasets. This analysis compared simulated clinician rankings with semi-automated and automated algorithms for variable selection when conducting anchored MAICs. Clinician ranking (60% chance of correct identification, which increased by 10% for every one unit increase in the log effect modifier size), semi-automated high dimensional propensity score [HDPS] algorithm, and four automated algorithms (hierNet, Bayesian projected prediction, Bayesian additive regression trees [BART], and random forest) were evaluated for variable selection in MAIC using 100 simulated trials with two true effect modifiers and eight noise variables. Effect modifiers added were 50% of the size of the treatment effect. Success was measured using an average number of false positives and negatives, and additive absolute bias of missed effect modifiers. Simulated clinician ranking identified one false negative and four false positive variables, with an absolute bias of 0.42. HDPS, hierNet, and Bayesian projected prediction had similar results (1.27, 0.88, and 0.60 false negatives, 1.01, 2.32, and 4.19 false positives, and absolute biases of 0.52, 0.36, and 0.25, respectively). Tree-based algorithms had the lowest absolute bias; BART and random forest had 0.12 and 0.02 false negatives, 2.02 and 1.92 false positives, with absolute biases of 0.018 and 0.003, respectively. Based on the simulations, tree-based algorithms performed best for variable selection in MAIC. These findings suggest that semi-automated and automated variable selection algorithms should also be considered to augment analyses based on clinician rankings when undertaking anchored MAICs.

Toward Automatic Detection of Radiation-Induced Cerebral Microbleeds Using a 3D Deep Residual Network

Cerebral microbleeds, which are small focal hemorrhages in the brain that are prevalent in many diseases, are gaining increasing attention due to their potential as surrogate markers of disease burden, clinical outcomes, and delayed effects of therapy. Manual detection is laborious and automatic detection and labeling of these lesions is challenging using traditional algorithms. Inspired by recent successes of deep convolutional neural networks in computer vision, we developed a 3D deep residual network that can distinguish true microbleeds from false positive mimics of a previously developed technique based on traditional algorithms. A dataset of 73 patients with radiation-induced cerebral microbleeds scanned at 7T with susceptibility-weighted imaging was used to train and evaluate our model. With the resulting network, we maintained 95% of the true microbleeds in 12 test patients and the average number of false positives was reduced by 89%, achieving a detection precision of 71.9%, higher than existing published methods. The likelihood score predicted by the network was also evaluated by comparing to a neuroradiologist's rating, and good correlation was observed.

Quantitative comparison of clustered microcalcifications in for-presentation and for-processing mammograms in full-field digital mammography.

Mammograms acquired with full-field digital mammography (FFDM) systems are provided in both "for-processing'' and "for-presentation'' image formats. For-presentation images are traditionally intended for visual assessment by the radiologists. In this study, we investigate the feasibility of using for-presentation images in computerized analysis and diagnosis of microcalcification (MC) lesions. We make use of a set of 188 matched mammogram image pairs of MC lesions from 95 cases (biopsy proven), in which both for-presentation and for-processing images are provided for each lesion. We then analyze and characterize the MC lesions from for-presentation images and compare them with their counterparts in for-processing images. Specifically, we consider three important aspects in computer-aided diagnosis (CAD) of MC lesions. First, we quantify each MC lesion with a set of 10 image features of clustered MCs and 12 textural features of the lesion area. Second, we assess the detectability of individual MCs in each lesion from the for-presentation images by a commonly used difference-of-Gaussians (DoG) detector. Finally, we study the diagnostic accuracy in discriminating between benign and malignant MC lesions from the for-presentation images by a pretrained support vector machine (SVM) classifier. To accommodate the underlying background suppression and image enhancement in for-presentation images, a normalization procedure is applied. The quantitative image features of MC lesions from for-presentation images are highly consistent with that from for-processing images. The values of Pearson's correlation coefficient between features from the two formats range from 0.824 to 0.961 for the 10 MC image features, and from 0.871 to 0.963 for the 12 textural features. In detection of individual MCs, the FROC curve from for-presentation is similar to that from for-processing. In particular, at sensitivity level of 80%, the average number of false-positives (FPs) per image region is 9.55 for both for-presentation and for-processing images. Finally, for classifying MC lesions as malignant or benign, the area under the ROC curve is 0.769 in for-presentation, compared to 0.761 in for-processing (P = 0.436). The quantitative results demonstrate that MC lesions in for-presentation images are highly consistent with that in for-processing images in terms of image features, detectability of individual MCs, and classification accuracy between malignant and benign lesions. These results indicate that for-presentation images can be compatible with for-processing images for use in CAD algorithms for MC lesions.