Mean False Positive Rate Research Articles

Application of Machine Learning in the Diagnosis of Early Gastric Cancer Using the Kyoto Classification Score and Clinical Features Collected from Medical Consultations.

The early detection accuracy of early gastric cancer (EGC) determines the choice of the optimal treatment strategy and the related medical expenses. We aimed to develop a simple, affordable, and time-saving diagnostic model using six machine learning (ML) algorithms for the diagnosis of EGC. It is based on the endoscopy-based Kyoto classification score obtained after the completion of endoscopy and other clinical features obtained after medical consultation. We retrospectively evaluated 1999 patients who underwent gastrointestinal endoscopy at the China Beijing Hospital. Of these, 203 subjects were diagnosed with EGC. The data were randomly divided into training and test sets (ratio 4:1). We constructed six ML models, and the developed models were evaluated on the testing set. This procedure was repeated five times. The Kolmogorov-Arnold Networks (KANs) model achieved the best performance (mean AUC value: 0.76; mean balanced accuracy: 70.96%; mean precision: 58.91%; mean recall: 70.96%; mean false positive rate: 26.11%; mean false negative rate: 31.96%; and mean F1 score value: 58.46). The endoscopy-based Kyoto classification score was the most important feature with the highest feature importance score. The results suggest that the KAN model, the optimal ML model in this study, has the potential to identify EGC patients, which may result in a reduction in both the time cost and medical expenses in clinical practice.

Limitations of Out-of-Distribution Detection in 3D Medical Image Segmentation.

Deep learning models perform unreliably when the data come from a distribution different from the training one. In critical applications such as medical imaging, out-of-distribution (OOD) detection methods help to identify such data samples, preventing erroneous predictions. In this paper, we further investigate OOD detection effectiveness when applied to 3D medical image segmentation. We designed several OOD challenges representing clinically occurring cases and found that none of the methods achieved acceptable performance. Methods not dedicated to segmentation severely failed to perform in the designed setups; the best mean false-positive rate at a 95% true-positive rate (FPR) was 0.59. Segmentation-dedicated methods still achieved suboptimal performance, with the best mean FPR being 0.31 (lower is better). To indicate this suboptimality, we developed a simple method called Intensity Histogram Features (IHF), which performed comparably or better in the same challenges, with a mean FPR of 0.25. Our findings highlight the limitations of the existing OOD detection methods with 3D medical images and present a promising avenue for improving them. To facilitate research in this area, we release the designed challenges as a publicly available benchmark and formulate practical criteria to test the generalization of OOD detection beyond the suggested benchmark. We also propose IHF as a solid baseline to contest emerging methods.

Fluorescence molecular optomic signatures improve identification of tumors in head and neck specimens.

Fluorescence molecular imaging using ABY-029, an epidermal growth factor receptor (EGFR)-targeted, synthetic Affibody peptide labeled with a near-infrared fluorophore, is under investigation for surgical guidance during head and neck squamous cell carcinoma (HNSCC) resection. However, tumor-to-normal tissue contrast is confounded by intrinsic physiological limitations of heterogeneous EGFR expression and non-specific agent uptake. In this preliminary study, radiomic analysis was applied to optical ABY-029 fluorescence image data for HNSCC tissue classification through an approach termed "optomics." Optomics was employed to improve tumor identification by leveraging textural pattern differences in EGFR expression conveyed by fluorescence. The study objective was to compare the performance of conventional fluorescence intensity thresholding and optomics for binary classification of malignant vs. non-malignant HNSCC tissues. Fluorescence image data collected through a Phase 0 clinical trial of ABY-029 involved a total of 20,073 sub-image patches (size of 1.8 × 1.8 mm2) extracted from 24 bread-loafed slices of HNSCC surgical resections originating from 12 patients who were stratified into three dose groups (30, 90, and 171 nanomoles). Each dose group was randomly partitioned on the specimen-level 75%/25% into training/testing sets, then all training and testing sets were aggregated. A total of 1,472 standardized radiomic features were extracted from each patch and evaluated by minimum redundancy maximum relevance feature selection, and 25 top-ranked features were used to train a support vector machine (SVM) classifier. Predictive performance of the SVM classifier was compared to fluorescence intensity thresholding for classifying testing set image patches with histologically confirmed malignancy status. Optomics provided consistent improvement in prediction accuracy and false positive rate (FPR) and similar false negative rate (FNR) on all testing set slices, irrespective of dose, compared to fluorescence intensity thresholding (mean accuracies of 89% vs. 81%, P = 0.0072; mean FPRs of 12% vs. 21%, P = 0.0035; and mean FNRs of 13% vs. 17%, P = 0.35). Optomics outperformed conventional fluorescence intensity thresholding for tumor identification using sub-image patches as the unit of analysis. Optomics mitigate diagnostic uncertainties introduced through physiological variability, imaging agent dose, and inter-specimen biases of fluorescence molecular imaging by probing textural image information. This preliminary study provides a proof-of-concept that applying radiomics to fluorescence molecular imaging data offers a promising image analysis technique for cancer detection in fluorescence-guided surgery.

Suspicious Actions Detection System Using Enhanced CNN and Surveillance Video

Suspicious pre- and post-activity detection in crowded places is essential as many suspicious activities may be carried out by culprits. Usually, there will be installations of surveillance cameras. These surveillance cameras capture videos or images later investigated by authorities and post-event such suspicious activity would be detected. This leads to high human intervention to detect suspicious activity. However, there are no systems available to protect valuable things from such suspicious incidents. Nowadays machine learning (ML)- and deep learning (DL)-based pre-incident warning alarm systems could be adapted to monitor suspicious activity. Suspicious activity prediction would be based on human gestures and unusual activity detection. Even though some methods based on ML or DL have been proposed, the need for a highly accurate, highly precise, low-false-positive and low-false-negative prediction system can be enhanced by hybrid or enhanced ML- or DL-based systems. This proposed research work has introduced an enhanced convolutional neural network (ECNN)-based suspicious activity detection system. The experiment was carried out and the results were claimed. The results are analyzed with the Statistical Package for the Social Sciences (SPSS) tool. The results showed that the mean accuracy, mean precision, mean false-positive rate, and mean false-negative rate of suspicious activity detections were 97.050%, 96.743%, 2.957%, and 2.927% respectively. This result was also compared with the convolutional neural network (CNN) algorithm. This research work can be applied to enhance the pre-suspicious activity alert security system to avoid risky situations.

Laboratory-Reported Normal Value Ranges Should Not Be Used to Diagnose Periprosthetic Joint Infection.

Introduction: Clinical laboratories offer several multipurpose tests, such as the erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP), which are not intended to diagnose any specific disease but are used by clinicians in multiple fields. The results and laboratory interpretation (normal/abnormal) of these multipurpose tests are based on laboratory-reported normal thresholds, which vary across clinical laboratories. In 2018, the International Consensus Meeting on Musculoskeletal Infection (2018 ICM) provided a gold-standard definition to diagnose periprosthetic joint infection (PJI) which included many multipurpose laboratory tests, along with thresholds optimized to diagnose PJI. The discrepancy between laboratory-reported normal thresholds and 2018 ICM-recommended PJI-optimized test thresholds has never been studied. The purpose of this study was to assess the existing variation in laboratory-reported normal thresholds for tests commonly used to diagnose PJI and evaluate the potential diagnostic impact of using laboratory-reported normal thresholds instead of 2018 ICM-recommended PJI-optimized thresholds.Methods: Clinical laboratories (N=85) were surveyed to determine the laboratory-reported units of measure and normal thresholds for common multipurpose tests to diagnose PJI, including the ESR, CRP, D-dimer, synovial fluid white blood cells (SF-WBC), and polymorphonuclear cell percent (SF-PMN%). The variability of units of measure and normal thresholds for each test was then assessed among the 85 included clinical laboratories. A representative dataset from patients awaiting a revision arthroplasty was used to determine the clinical significance of the existing discrepancy between laboratory-reported normal test interpretations and 2018 ICM-recommended PJI-optimized test interpretations.Results: Two units of measure for the CRP and six units of measure for the D-dimer were observed, with only 59% of laboratories reporting the CRP in terms of mg/L and only 16% reporting the D-dimer in ng/ml, as needed to utilize the 2018 ICM definition of PJI. Across clinical laboratories surveyed, the mean laboratory-reported normal thresholds for the ESR (20 mm/h), CRP (7.69 mg/L), D-dimer (500 ng/mL), SF-WBC (5 cells/uL), and SF-PMN% (25%) were substantially lower than the 2018 ICM-recommended PJI-optimized thresholds of 30 mm/h, 10 mg/L, 860 ng/mL, 3,000 cells/uL, and 70%, respectively. Interpretation of test results from a representative PJI dataset using each laboratory’s normal test thresholds yielded mean false-positive rates of 14% (ESR), 18% (CRP), 42% (D-dimer), 93% (SF-WBC), and 36% (SF-PMN%) versus the ICM-recommended PJI-optimized thresholds.Conclusion: When reporting the results for multipurpose laboratory tests, such as the ESR, CRP, D-dimer, SF-WBC, and SF-PMN%, clinical laboratories utilize laboratory-reported units of measure and normal thresholds that are not intended to diagnose PJI, and therefore may not match the 2018 ICM recommendations. Our findings reveal that laboratory-reported normal thresholds for these multipurpose tests are well below the 2018 ICM recommendations to diagnose PJI. Clinical reliance on laboratory-reported results and interpretations, instead of strict use of the 2018 ICM-recommended units and PJI-optimized thresholds, may lead to false-positive interpretation of multipurpose laboratory tests.

Automatic construction of filter tree by genetic programming for ultrasound guidance image segmentation

Segmentation of ultrasound guidance images (UGIs) is a critical step in ultrasound-guided high intensity focused ultrasound (HIFU) therapy. However, the low signal-to-noise ratio characteristic of UGIs makes it difficult to acquire enough annotations. This paper proposes a novel genetic programming-based approach to achieve automatic construction of an image filter tree (IFT) for UGI segmentation since genetic programming has a natural advantage in training on small datasets. In the new approach, a set of predefined functions are adapted with better anti-noise performance to deal with noise interference. Moreover, a position-determined function is designed for incorporating preoperative information in each IFT to form a closed-loop system thereby facilitating the segmentation process. The optimal IFT evolved by genetic programming, along with a preprocessing step and a postprocessing step, constructs the pipeline for the segmentation of UGIs. The quantitative evaluation of the segmentation results shows the mean true positive rate, the mean false positive rate, the mean intersection over union, the mean norm Hausdorff distance and the mean norm maximum average distance are found to be 94.86%, 6.72%, 89.14%, 3.20% and 0.83%, respectively, outperforming the popular convolutional neural network-based segmentation methods. The segmentation results reveal that the evolved IFT can achieve accurate segmentation of UGIs and indicate that the proposed approach can be a promising option for medical image segmentation when there are only a few training samples available.

Diagnostic validation of a deep learning nodule detection algorithm in low-dose chest CT: determination of optimized dose thresholds in a virtual screening scenario.

This study was conducted to evaluate the effect of dose reduction on the performance of a deep learning (DL)-based computer-aided diagnosis (CAD) system regarding pulmonary nodule detection in a virtual screening scenario. Sixty-eight anthropomorphic chest phantoms were equipped with 329 nodules (150 ground glass, 179 solid) with four sizes (5mm, 8mm, 10mm, 12mm) and scanned with nine tube voltage/current combinations. The examinations were analyzed by a commercially available DL-based CAD system. The results were compared by a comparison of proportions. Logistic regression was performed to evaluate the impact of tube voltage, tube current, nodule size, nodule density, and nodule location. The combination with the lowest effective dose (E) and unimpaired detection rate was 80kV/50 mAs (sensitivity: 97.9%, mean false-positive rate (FPR): 1.9, mean CTDIvol: 1.2 ± 0.4mGy, mean E: 0.66mSv). Logistic regression revealed that tube voltage and current had the greatest impact on the detection rate, while nodule size and density had no significant influence. The optimal tube voltage/current combination proposed in this study (80kV/50 mAs) is comparable to the proposed combinations in similar studies, which mostly dealt with conventional CAD software. Modification of tube voltage and tube current has a significant impact on the performance of DL-based CAD software in pulmonary nodule detection regardless of their size and composition. • Modification of tube voltage and tube current has a significant impact on the performance of deep learning-based CAD software. • Nodule size and composition have no significant impact on the software's performance. • The optimal tube voltage/current combination for the examined software is 80kV/50 mAs.

Clinical evaluation of a deep-learning-based computer-aided detection system for the detection of pulmonary nodules in a large teaching hospital

To evaluate a deep-learning-based computer-aided detection (DL-CAD) software system for pulmonary nodule detection on computed tomography (CT) images and assess its added value in the clinical practice of a large teaching hospital. A retrospective analysis was performed of 145 chest CT examinations by comparing the output of the DL-CAD software with a reference standard based on the consensus reading of three radiologists. For every nodule in each scan, the location, composition, and maximum diameter in the axial plane were recorded. The subgroup of chest CT examinations (n = 97) without any nodules was used to determine the negative predictive value at the given clinical sensitivity threshold setting. The radiologists found 91 nodules and the CAD system 130 nodules of which 80 were true positive. The measured sensitivity was 88% and the mean false-positive rate was 1.04 false positives/scan. The negative predictive value was 95%. For 23 nodules, there was a size discrepancy of which 19 (83%) were measured smaller by the radiologist. The agreement of nodule composition between the CAD results and the reference standard was 95%. The present study found a sensitivity of 88% and a false-positive rate of 1.04 false positives/scan, which match the vendor specification. Together with the measured negative predictive value of 95% the system performs very well; however, these rates are still not good enough to replace the radiologist, even for the specific task of nodule detection. Furthermore, a surprisingly high rate of overestimation of nodule size was observed, which can lead to too many follow-up examinations.

Performance comparison: exome sequencing as a single test replacing Sanger sequencing.

Next generation sequencing tests are used routinely as first-choice tests in the clinic. However, systematic performance comparing the results of exome sequencing as a single test replacing Sanger sequencing of targeted gene(s) is still lacking. Performance comparison data are critically important for clinical case management. In this study, we compared Sanger-sequencing results of 258 genes to those obtained from next generation sequencing (NGS) using two exome-sequencing enrichment kits: Agilent-SureSelectQXT and Illumina-Nextera. Sequencing was performed on leukocytes and buccal-derived DNA from a single individual, and all 258 genes were sequenced a total of 11 times (using different sequencing methods and DNA sources). Sanger sequencing was completed for all exons, including flanking ± 8bp regions. For the 258 genes, NGS mean coverage was > 20 × for > 98 and > 91% of the regions targeted by SureSelect and Nextera, respectively. Overall, 449 variants were identified in at least one experiment, and 407/449 (90.6%) were detected by all. Of the 42 discordant variants, 23 were determined as true calls, summing-up to a truth set of 430 variants. Sensitivity of true-variant detection was 99% for Sanger sequencing and 97-100% for the NGS experiments. Mean false-positive rates were 3.7E-6 for Sanger sequencing, 2.5E-6 for SureSelect-NGS and 5.2E-6 for Nextera-NGS. Our findings suggest a high overall concordance between Sanger sequencing and NGS performances. Both methods demonstrated false-positive and false-negative calls. High clinical suspicion for a specific diagnosis should, therefore, override negative results of either Sanger sequencing or NGS.

A Robust Method of Eye Torsion Measurement for Medical Applications

The detection of eye torsion is an important element for diagnosis of balance disorders, although it is rarely available in existing eye tracking systems. A novel method is proposed in this paper to provide robust measurement of torsional eye movements. A numerical approach is presented to estimate the iris boundary only according to the gaze direction, so the segmentation of the iris is more robust against occlusions and ambiguities. The perspective distortion of the iris pattern at eccentric eye positions is also corrected, benefiting from the transformation relation that is established for the iris estimation. The angle of the eye torsion is next measured on the unrolled iris patterns via a TM (Template Matching) technique. The principle of the proposed method is validated and its robustness in practice is assessed. A very low mean FPR (False Positive Rate) is reported (i.e., 3.3%) in a gaze test when testing on five participants with very different eye morphologies. The present method always gave correct measurement on the iris patterns with simulated eye torsions and rarely provided mistaken detections in the absence of eye torsion in practical conditions. Therefore, it shows a good potential to be further applied in medical applications.

The Milan system for reporting salivary gland cytopathology: A comprehensive review of the literature.

The Milan system for Reporting Salivary Gland Cytopathology (MSRSGC) was published in 2018. Since then, many authors have published their institutional experience by retrospectively assigning salivary gland fine-needle aspiration cases to each of the MSRSGC categories and calculated their risk of malignancy (ROM) accordingly. We reviewed all published articles available online in English that used the MSRSGC since or near its publication. We calculated the risk of neoplasm and ROM for each diagnostic category. In addition, the false-negative and false-positive rates from all studies were examined. Thirty-seven articles were identified in the English literature; 2 were published in 2017, 14 in 2018, 18 in 2019, and 3 in 2020. The total number of cases was 16 394, and 8 468 had surgical follow-up. The mean ROM was 16.9% for category I, 10.5% for category II, 39.3% for category III, 2.9% for category IVa, 39.4% for category IVb, 84.2% for category V, and 97.5% for category VI. The mean false-negative rate for MSRSGC categories II and IVa was 4.5%. Similarly, the mean false-positive rate for MSRSGC categories V and VI was 5.1%. A tiered classification scheme of MSRSGC is helpful in effectively guiding clinical management of patients with salivary gland lesions. The reported mean ROM for each category in most studies is within the recommended range published by the MSRSGC.

Accuracy of radiographer comment following a two-month experiential and blended learning in appendicular skeleton X-ray interpretation: The Singapore experience

IntroductionThe value of combined blended and experiential learning on radiographer diagnostic comment has not been explored. This study aims to examine the accuracy of image interpretation comment of radiographers who received a period of blended and experiential learning in Radiographer Abnormality Detection Systems (RADS). MethodsWe evaluated the diagnostic opinions of 13 radiographers who received a blended training and experiential learning (a process of self-learning and reflection) in RADS. Radiographers' opinions on 16,483 images were examined using the final radiologists’ report as a reference standard. For each radiographer, we recorded the number of true positive, true negative, false positive and false negative opinions and MedCal® was used to calculate diagnostic performance and error rates. A t-test was used to assess whether the number of images read was associated with performance and whether the radiographers retained performance over time. ResultsSensitivity ranged from 87.4 (84.0–90.2) to 98.9 (97.5–99.7) with a mean of 94.3 (93.6–94.8). Specificity varied from 96.4 (94.8–97.5) to 99.9 (99.41–100.0) with a mean of 98.2 (97.9–98.4). Diagnostic accuracy ranged from 93.1 (91.5–94.4) to 99.5 (98.9–99.8) with a mean of 96.9 (96.6–97.1). The mean false positive rate was 0.018 (range: 0.010–0.031) with a false negative rate of 0.057 (range: 0.026–0.11). There were no differences in performance between the first and latter nine months of providing opinions and the number of images reviewed was not associated with performance. ConclusionRadiographers who received blended and experiential learning in RADS provide accurate diagnostic comments on plain emergency appendicular skeleton radiographs. Implication for practiceA combined blended and experiential learning can equip radiographers to provide diagnostic opinion on plain appendicular skeleton radiographs.

Aristotle: A single blood test for pan-cancer screening.

e15037 Background: Blood participates in physiological and pathological changes throughout the body which are reflected as changes in cellular gene expression. The involvement of the immune system in cancer suppression and promotion through immunoediting exemplifies this. These characteristics make blood a promising non-invasive surrogate tissue for cancer diagnosis and monitoring. Methods: A multi-class cancer classification model, Aristotle, was developed with logistic regression using Affymetrix gene expression profiles generated from 2,845 unique human whole peripheral blood samples. The model was evaluated through 100 iterations of 2-fold cross-validation. Results: The 1,013 cancer patient samples (age 61±13, 40.6% ♀), include cancer of the bladder (n = 87, age 65±11, 23.3% ♀), breast (n = 94, age 54±10, 98.4% ♀), colon (n = 205, age 64±13, 39.5% ♀)/colon polyps (n = 226, age 61±11, 36.0% ♀), cervix (n = 36, age 55±13), endometrium (n = 25, age 49±12), liver (n = 16, age 57±11, 12.5% ♀), nasopharynx (n = 78, age 51±13, 26.9% ♀), ovaries (n = 64, age 55±12), prostate (n = 160, age 68±9) or stomach (n = 22, age 59±11, 28.6% ♀); and 1,832 control samples (age 53±16, 38.9% ♀), including healthy subjects (n = 1,042, age 55±16, 37.0 ♀) and subjects with ankylosing spondylitis (n = 59, age 39±12, 18.6% ♀), heart failure (n = 193, age 67±12, 40.0% ♀), inflammatory bowel disease (n = 51, age 37±9, 40.0% ♀), osteoarthritis (n = 248, age 53±15, 45.6% ♀), rheumatoid arthritis (n = 40, age 57±17, 84.6% ♀), schizophrenia (n = 110, age 36±13, 43.6% ♀) or other (n = 89, age 50±14, 41.1% ♀). Aristotle classified cancer profiles with sensitivities from 55.6-100% and positive predictive values from 5.6-77.7% at a specificity of 99.0% (Table). The mean false positive rate for the 11 cancer classes ranged from 0.3- 6.8%. Conclusions: Our study demonstrates the clinical viability of using whole peripheral blood for multi-class cancer classification. Aristotle was capable of classifying patients with different types of cancer with clinically relevant accuracy. We are confident that Aristotle holds great promise to help address the need for practical and non-invasive early-stage cancer diagnostics. [Table: see text]

Computer-assisted EEG diagnostic review for idiopathic generalized epilepsy

Epilepsy diagnosis can be costly, time-consuming, and not uncommonly inaccurate. The reference standard diagnostic monitoring is continuous video-electroencephalography (EEG) monitoring, ideally capturing all events or concordant interictal discharges. Automating EEG data review would save time and resources, thus enabling more people to receive reference standard monitoring and also potentially heralding a more quantitative approach to therapeutic outcomes. There is substantial research into the automated detection of seizures and epileptic activity from EEG. However, automated detection software is not widely used in the clinic, and despite numerous published algorithms, few methods have regulatory approval for detecting epileptic activity from EEG.This study reports on a deep learning algorithm for computer-assisted EEG review. Deep convolutional neural networks were trained to detect epileptic discharges using a preexisting dataset of over 6000 labelled events in a cohort of 103 patients with idiopathic generalized epilepsy (IGE). Patients underwent 24-hour ambulatory outpatient EEG, and all data were curated and confirmed independently by two epilepsy specialists (Seneviratne et al., 2016). The resulting automated detection algorithm was then used to review diagnostic scalp EEG for seven patients (four with IGE and three with events mimicking seizures) to validate performance in a clinical setting.The automated detection algorithm showed state-of-the-art performance for detecting epileptic activity from clinical EEG, with mean sensitivity of >95% and corresponding mean false positive rate of 1 detection per minute. Importantly, diagnostic case studies showed that the automated detection algorithm reduced human review time by 80%–99%, without compromising event detection or diagnostic accuracy. The presented results demonstrate that computer-assisted review can increase the speed and accuracy of EEG assessment and has the potential to greatly improve therapeutic outcomes.This article is part of the Special Issue "NEWroscience 2018".

P576The first handheld device for autonomic self-detection of symptomatic acute coronary artery occlusion: feasibility, performance and implications for time-efficient self-triage of outpatients with CAD

Abstract Background Time delay between onset of symptoms and seeking medical attention is a major determinant of mortality and morbidity in patients with acute coronary artery occlusion (ACAO). Response time might be reduced by reliable self-detection of ACAO. Currently no self-applicable device can accurately detect ACAO. We have recently shown that an automatic algorithm based upon a three lead system (RELF method) accurately detects ACAO. Purpose In this multicenter observational study we tested the operational feasibility, sensitivity and specificity of our RELF method, built into a mobile handheld device, for detection of ACAO. Methods Patients with planned elective percutaneous coronary intervention (PCI), who were able to use a smartphone, were asked to perform random ambulatory self-recordings for at least one week. A similar self-recording was made before angioplasty and at 60 s of balloon occlusion. Results The operational feasibility of ambulatory self-recordings in enrolled patients with CAD was 59/64 (92.1%, 95% CI: 83.0–96.6). Of all self-recordings 91.1% (n=4567) were automatically classified as successful within one minute. The mean false positive rate during daily life conditions was 4.19% (95% CI: 3.29–5.10). Amongst 65 balloon occlusions, 63 index-tests at 60 s of occlusion were available. The sensitivity for the target conditions “ACAO”, “ACAO with ECG changes” and “ACAO with ECG changes and ST segment elevation myocardial infarction (STEMI) criteria” was respectively 55/63 (0.87; 95% CI: 0.77–0.93), 54/57 (0.95; 95% CI: 0.86–0.98) and 35/35 (1.00). The figure depicts all (n=3936) ST difference vector (STDVn) measurements obtained during ambulatory postural changes, exercise and coronary artery occlusion with and without ECG changes and/or STEMI criteria. Receiver Operator Curve (ROC) for ACAO at different cut-off values of the magnitude of STDVn was 0.973 (95% CI: 0.956–0.990). Boxplots of all STDVn test recordings Conclusions Self-recording with our RELF device is feasible for the majority of patients with CAD. The sensitivity and specificity for automatic detection of the earliest phase of acute coronary artery occlusion support the concept of our RELF device for patient empowerment to reduce delay and increase survival without overloading emergency services. This is the first clinical study that confirms the proof-of-concept of self-detection of acute coronary artery occlusion in outpatients with CAD. Acknowledgement/Funding Ghent University, Industrial Research Fund (IOF reference: F2015/IOF-advanced/084).

Feasibility and performance of a device for automatic self-detection of symptomatic acute coronary artery occlusion in outpatients with coronary artery disease: a multicentre observational study

Time delay between onset of symptoms and seeking medical attention is a major determinant of mortality and morbidity in patients with acute coronary artery occlusion. Response time might be reduced by reliable self-detection. We aimed to formally assess the proof-of-concept and accuracy of self-detection of acute coronary artery occlusion by patients during daily life situations and during the very early stages of acute coronary artery occlusion. In this multicentre, observational study, we tested the operational feasibility, specificity, and sensitivity of our RELF method, a three-lead detection system with an automatic algorithm built into a mobile handheld device, for detection of acute coronary artery occlusion. Patients were recruited continuously by physician referrals from three Belgian hospitals until the desired sample size was achieved, had been discharged with planned elective percutaneous coronary intervention, and were able to use a smartphone; they were asked to perform random ambulatory self-recordings for at least 1 week. A similar self-recording was made before percutaneous coronary intervention and at 60 s of balloon occlusion. Patients were clinically followed up until 1 month after discharge. We quantitatively assessed the operational feasibility with an automated dichotomous quality check of self-recordings. Performance was assessed by analysing the receiver operator characteristics of the ST difference vector magnitude. This trial is registered with ClinicalTrials.gov, number NCT02983396. From Nov 18, 2016, to April 25, 2018, we enrolled 64 patients into the study, of whom 59 (92%) were eligible for self-applications. 58 (91%) of 64 (95% CI 81·0-95·6) patients were able to perform ambulatory self-recordings. Of all 5011 self-recordings, 4567 (91%) were automatically classified as successful within 1 min. In 65 balloon occlusions, 63 index tests at 60 s of occlusion in 55 patients were available. The mean specificity of daily life recordings was 0·96 (0·95-0·97). The mean false positive rate during daily life conditions was 4·19% (95% CI 3·29-5·10). The sensitivity for the target conditions was 0·87 (55 of 63; 95% CI 0·77-0·93) for acute coronary artery occlusion, 0·95 (54 of 57; 0·86-0·98) for acute coronary artery occlusion with electrocardiogram (ECG) changes, and 1·00 (35 of 35) for acute coronary artery occlusion with ECG changes and ST-segment elevation myocardial infarction criteria (STEMI). The index test was more sensitive to detect a 60 s balloon occlusion than the STEMI criteria on 12-lead ECG (87% vs 56%; p<0·0001). The proportion of total variation in study estimates due to heterogeneity between patients (I2) was low (12·6%). The area under the receiver operator characteristics curve was 0·973 (95% CI 0·956-0·990) for acute coronary artery occlusion at different cutoff values of the magnitude of the ST difference vector. No patients died during the study. Self-recording with our RELF device is feasible for most patients with coronary artery disease. The sensitivity and specificity for automatic detection of the earliest phase of acute coronary artery occlusion support the concept of our RELF device for patient empowerment to reduce delay and increase Survival without overloading emergency services. Ghent University, Industrial Research Fund.

A cloud-based computer-aided detection system improves identification of lung nodules on computed tomography scans of patients with extra-thoracic malignancies.

To compare unassisted and CAD-assisted detection and time efficiency of radiologists in reporting lung nodules on CT scans taken from patients with extra-thoracic malignancies using a Cloud-based system. Three radiologists searched for pulmonary nodules in patients with extra-thoracic malignancy who underwent CT (slice thickness/spacing 2 mm/1.7 mm) between September 2015 and March 2016. All nodules detected by unassisted reading were measured and coordinates were uploaded on a cloud-based system. CAD marks were then reviewed by the same readers using the cloud-based interface. To establish the reference standard all nodules ≥ 3 mm detected by at least one radiologist were validated by two additional experienced radiologists in consensus. Reader detection rate and reporting time with and without CAD were compared. The study was approved by the local ethics committee. All patients signed written informed consent. The series included 225 patients (age range 21-90 years, mean 62 years), including 75 patients having at least one nodule, for a total of 215 nodules. Stand-alone CAD sensitivity for lesions ≥ 3 mm was 85% (183/215, 95% CI: 82-91); mean false-positive rate per scan was 3.8. Sensitivity across readers in detecting lesions ≥ 3 mm was statistically higher using CAD: 65% (95% CI: 61-69) versus 88% (95% CI: 86-91, p<0.01). Reading time increased by 11% using CAD (296 s vs. 329 s; p<0.05). In patients with extra-thoracic malignancies, CAD-assisted reading improves detection of ≥ 3-mm lung nodules on CT, slightly increasing reading time. • CAD-assisted reading improves the detection of lung nodules compared with unassisted reading on CT scans of patients with primary extra-thoracic tumour, slightly increasing reading time. • Cloud-based CAD systems may represent a cost-effective solution since CAD results can be reviewed while a separated cloud back-end is taking care of computations. • Early identification of lung nodules by CAD-assisted interpretation of CT scans in patients with extra-thoracic primary tumours is of paramount importance as it could anticipate surgery and extend patient life expectancy.

A novel multistage system for the detection and removal of pectoral muscles in mammograms

In this paper, a novel multistage scheme for pectoral muscle removal from mammography images is proposed, and the performance of this system is verified using the publicly available Mammographic Image Analysis Society digital mammogram database. This database is composed of mediolateral oblique mammography images including three different tissue types (fatty, fatty-glandular, and dense-glandular) with three health status types (normal, benign cancer, and malignant cancer). In the implementation of the proposed system, a mammography image is first preprocessed by performing noise reduction background removal followed by artifact suppression processes. Then a presegmentation procedure is applied using region growing and line fitting is executed. Finally, pixels including pectoral muscle regions are removed from mammography images with an accuracy of 94.40%, sensitivity of 89.62%, and specificity of 99.99% after some postprocessing operations. Although the mean false positive rate obtained by the proposed approach is higher than that of other studies in the literature, not only the lower mean false negative rate but also the enhancement in the quality of pectoral muscle segmentation for all 322 images in the database evidently show the success of the proposed approach.

Cluster signal-to-noise analysis for evaluation of the information content in an image.

(1) To develop an observer-free method of analysing image quality related to the observer performance in the detection task and (2) to analyse observer behaviour patterns in the detection of small mass changes in cone-beam CT images. 13 observers detected holes in a Teflon phantom in cone-beam CT images. Using the same images, we developed a new method, cluster signal-to-noise analysis, to detect the holes by applying various cut-off values using ImageJ and reconstructing cluster signal-to-noise curves. We then evaluated the correlation between cluster signal-to-noise analysis and the observer performance test. We measured the background noise in each image to evaluate the relationship with false positive rates (FPRs) of the observers. Correlations between mean FPRs and intra- and interobserver variations were also evaluated. Moreover, we calculated true positive rates (TPRs) and accuracies from background noise and evaluated their correlations with TPRs from observers. Cluster signal-to-noise curves were derived in cluster signal-to-noise analysis. They yield the detection of signals (true holes) related to noise (false holes). This method correlated highly with the observer performance test (R2 = 0.9296). In noisy images, increasing background noise resulted in higher FPRs and larger intra- and interobserver variations. TPRs and accuracies calculated from background noise had high correlation with actual TPRs from observers; R2 was 0.9244 and 0.9338, respectively. Cluster signal-to-noise analysis can simulate the detection performance of observers and thus replace the observer performance test in the evaluation of image quality. Erroneous decision-making increased with increasing background noise.

Validation studies in forensic odontology – Part 1: Accuracy of radiographic matching

As part of a series of studies aimed at validating techniques in forensic odontology, this study aimed to validate the accuracy of ante-mortem (AM)/postmortem (PM) radiographic matching by dentists and forensic odontologists. This study used a web-based interface with 50 pairs of AM and PM radiographs from real casework, at varying degrees of difficulty. Participants were shown both radiographs as a pair and initially asked to decide if they represented the same individual using a yes/no binary choice forced-decision. Participants were asked to assess their level of confidence in their decision, and to make a conclusion using one of the ABFO (American Board of Forensic Odontology), INTERPOL (International Criminal Police Organisation) and DVISys™ (DVI System International, Plass Data Software) identification scale degrees. The mean false-positive rate using the binary choice scale was 12%. Overall accuracy was 89% using this model, however, 13% of participants scored below 80%. Only 25% of participants accurately answered yes or no >90% of the time, with no individual making the correct yes/no decision for all 50 pairs of radiographs. Non-odontologists (lay participants) scored poorly, with a mean accuracy of only 60%. Use of the graded ABFO, DVISYS and INTERPOL scales resulted in general improvements in performance, with the false-positive and false-negative rates falling to approximately 2% overall. Inter-examiner agreement in assigning scale degrees was good (ICC=0.64), however there was little correlation between confidence and both accuracy or agreement among practitioners. These results suggest that use of a non-binary scale is supported over a match/non-match call as it reduces the frequency of false positives and negatives. The use of the terms “possible” and “insufficient information” in the same scale appears to create confusion, reducing inter-examiner agreement. The lack of agreement between higher-performing and lower-performing groups suggests that there is an inconsistency in the cognitive processes used to determine similarity between radiographs.