Search

Background and Objective: Determining the most informative features for predicting the overall survival of patients diagnosed with high-grade gastroenteropancreatic neuroendocrine neoplasms is crucial to improve individual treatment plans for patients, as well as the biological understanding of the disease. The main objective of this study is to evaluate the use of modern ensemble feature selection techniques for this purpose with respect to (a) quantitative performance measures such as predictive performance, (b) clinical interpretability, and (c) the effect of integrating prior expert knowledge., Methods: The Repeated Elastic Net Technique for Feature Selection (RENT) and the User-Guided Bayesian Framework for Feature Selection (UBayFS) are recently developed ensemble feature selectors investigated in this work. Both allow the user to identify informative features in datasets with low sample sizes and focus on model interpretability. While RENT is purely data-driven, UBayFS can integrate expert knowledge a priori in the feature selection process. In this work, we compare both feature selectors on a dataset comprising 63 patients and 110 features from multiple sources, including baseline patient characteristics, baseline blood values, tumor histology, imaging, and treatment information., Results: Our experiments involve data-driven and expert-driven setups, as well as combinations of both. In a five-fold cross-validated experiment without expert knowledge, our results demonstrate that both feature selectors allow accurate predictions: A reduction from 110 to approximately 20 features (around 82%) delivers near-optimal predictive performances with minor variations according to the choice of the feature selector, the predictive model, and the fold. Thereafter, we use findings from clinical literature as a source of expert knowledge. In addition, expert knowledge has a stabilizing effect on the feature set (an increase in stability of approximately 40%), while the impact on predictive performance is limited., Conclusions: The features WHO Performance Status, Albumin, Platelets, Ki-67, Tumor Morphology, Total MTV, Total TLG, and SUV max are the most stable and predictive features in our study. Overall, this study demonstrated the practical value of feature selection in medical applications not only to improve quantitative performance but also to deliver potentially new insights to experts., Competing Interests: Declaration of Competing Interest The authors declare no conflicts of interest., (Copyright © 2023 The Author(s). Published by Elsevier B.V. All rights reserved.)

Background: 18 F-flutemetamol positron emission tomography (PET) is used to assess cortical amyloid-β burden in patients with cognitive impairment to support a clinical diagnosis. Visual classification is the most widely used method in clinical practice although semi-quantification is beneficial to obtain an objective and continuous measure of the Aβ burden. The aims were: first to evaluate the correspondence between standardized uptake value ratios (SUVRs) from three different software, Centiloids and visual classification, second to estimate thresholds for supporting visual classification and last to assess differences in semi-quantitative measures between clinical diagnoses., Methods: This observational study included 195 patients with cognitive impairment who underwent 18 F-flutemetamol PET. PET images were semi-quantified with SyngoVia, CortexID suite, and PMOD. Receiver operating characteristics curves were used to compare visual classification with composite SUVR normalized to pons (SUVRpons) and cerebellar cortex (SUVRcer), and Centiloids. We explored correlations and differences between semi-quantitative measures as well as differences in SUVR between two clinical diagnosis groups: Alzheimer's disease-group and non-Alzheimer's disease-group., Results: PET images from 191 patients were semi-quantified with SyngoVia and CortexID and 86 PET-magnetic resonance imaging pairs with PMOD. All receiver operating characteristics curves showed a high area under the curve (>0.98). Thresholds for a visually positive PET was for SUVRcer: 1.87 (SyngoVia) and 1.64 (CortexID) and for SUVRpons: 0.54 (SyngoVia) and 0.55 (CortexID). The threshold on the Centiloid scale was 39.6 Centiloids. All semi-quantitative measures showed a very high correlation between different software and normalization methods. Composite SUVRcer was significantly different between SyngoVia and PMOD, SyngoVia and CortexID but not between PMOD and CortexID. Composite SUVRpons were significantly different between all three software. There were significant differences in the mean rank of SUVRpons, SUVRcer, and Centiloid between Alzheimer's disease-group and non-Alzheimer's disease-group., Conclusions: SUVR from different software performed equally well in discriminating visually positive and negative 18 F-Flutemetamol PET images. Thresholds should be considered software-specific and cautiously be applied across software without preceding validation to categorize scans as positive or negative. SUVR and Centiloid may be used alongside a thorough clinical evaluation to support a clinical diagnosis., Competing Interests: Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://dx.doi.org/10.21037/qims-21-188). HLS has received consultancy fees from Siemens Healthineers. THE has been a rater in two clinical trials (Roche BN29553 and Boehringer-Ingelheim 1346.0023) outside the submitted work. ABK has been a national coordinator and principal investigator in three clinical trials (Roche BN29553, Boehringer-Ingelheim 1346.0023 and Novo Nordisk, EVOKE NN6535-4730) outside the submitted work. The other authors have no conflicts of interest to declare., (2022 Quantitative Imaging in Medicine and Surgery. All rights reserved.)

Background: The aim was to compare resource utilization across the four health trusts within the Western Norway Regional Health Authority since the establishment of positron emission tomography (PET) at Haukeland University Hospital in Bergen in 2009., Methods: Metadata from all PET examinations from 2009 to 2014 were automatically imported from the PET centre's central production database into a custom-developed database system, MDCake. A PET examination was defined as a procedure based on a single injection of radioactive tracer. The patients' place of residence and tentative diagnosis were coded based on the available clinical information., Results: The total number of PET examinations increased from 293 in 2009 to 1616 in 2014. The number of PET examinations per year increased across all diagnostic groups, but plateaued for lung cancer, gastrointestinal cancer and malignant melanoma since 2013. The number of examinations per capita was evenly distributed between the three northern health trusts with an average of 1260 PET studies per million inhabitants in 2014. However, patients residing in the most southerly health trust received between 44% (2010) and 27% (2014; P<0·001, repeated measures ANOVA) fewer examinations per capita per year., Conclusion: Centralized PET in the Western Norwegian health region meets the current clinical demand for patients residing in the three northern health trusts while patients from the most southern health trust receive approximately 30% fewer PET examinations. Access to specialized health care should be monitored routinely in order to identify inequalities in referral patterns and resource utilization., (© 2015 The Authors. Clinical Physiology and Functional Imaging published by John Wiley & Sons Ltd on behalf of Scandinavian Society of Clinical Physiology and Nuclear Medicine.)