This study investigates the effect of class imbalance in training data when developing neural network classifiers for computer aided medical diagnosis. The investigation is performed in the presence of other characteristics that are typical among medical data, namely small training sample size, large number of features, and correlations between features. Two methods of neural network training are explored: classical backpropagation (BP) and particle swarm optimization (PSO) with clinically relevant training criteria. An experimental study is performed using simulated data and the conclusions are further validated on real clinical data for breast cancer diagnosis. The results show that classifier performance deteriorates with even modest class imbalance in the training data. Further, it is shown that BP is generally preferable over PSO for imbalanced training data especially with small data sample and large number of features. Finally, it is shown that there is no clear preference between oversampling and no compensation approach and some guidance is provided regarding a proper selection.
IMPORTANCE Mammography screening currently relies on subjective human interpretation. Artificial intelligence (AI) advances could be used to increase mammography screening accuracy by reducing missed cancers and false positives. OBJECTIVE To evaluate whether AI can overcome human mammography interpretation limitations with a rigorous, unbiased evaluation of machine learning algorithms. DESIGN, SETTING, AND PARTICIPANTS In this diagnostic accuracy study conducted between September 2016 and November 2017, an international, crowdsourced challenge was hosted to foster AI algorithm development focused on interpreting screening mammography. More than 1100 participants comprising 126 teams from 44 countries participated. Analysis began November 18, 2016. MAIN OUTCOMES AND MEASUREMENTS Algorithms used images alone (challenge 1) or combined images, previous examinations (if available), and clinical and demographic risk factor data (challenge 2) and output a score that translated to cancer yes/no within 12 months. Algorithm accuracy for breast cancer detection was evaluated using area under the curve and algorithm specificity compared with radiologists' specificity with radiologists' sensitivity set at 85.9% (United States) and 83.9% (Sweden). An ensemble method aggregating top-performing AI algorithms and radiologists' recall assessment was developed and evaluated. RESULTS Overall, 144 231 screening mammograms from 85 580 US women (952 cancer positive Յ12 months from screening) were used for algorithm training and validation. A second independent validation cohort included 166 578 examinations from 68 008 Swedish women (780 cancer positive). The top-performing algorithm achieved an area under the curve of 0.858 (United States) and 0.903 (Sweden) and 66.2% (United States) and 81.2% (Sweden) specificity at the radiologists' sensitivity, lower than community-practice radiologists' specificity of 90.5% (United States) and 98.5% (Sweden). Combining top-performing algorithms and US radiologist assessments resulted in a higher area under the curve of 0.942 and achieved a significantly improved specificity (92.0%) at the same sensitivity. CONCLUSIONS AND RELEVANCE While no single AI algorithm outperformed radiologists, an ensemble of AI algorithms combined with radiologist assessment in a single-reader screening environment improved overall accuracy. This study underscores the potential of using machine (continued)
The goal of intensity-modulated radiation therapy (IMRT) treatment plan optimization is to produce a cumulative dose distribution that satisfies both the dose prescription and the normal tissue dose constraints. The typical manual treatment planning process is iterative, time consuming, and highly dependent on the skill and experience of the planner. We have addressed this problem by developing a knowledgebased approach that utilizes a database of prior plans to leverage the planning expertise of physicians and physicists at our institution. We developed a case-similarity algorithm that uses mutual information to identify a similar matched case for a given query case, and various treatment parameters from the matched case are then adapted to derive new treatment plans that are patient specific.We used 10 randomly selected cases matched against a knowledge base of 100 cases to demonstrate that new, clinically acceptable IMRT treatment plans can be developed. This approach substantially reduced planning time by skipping all but the last few iterations of the optimization process. Additionally, we established a simple metric based on the areas under the curve (AUC) of the dose volume histogram (DVH), specifically for the planning target volume (PTV), rectum, and bladder. This plan quality metric was used to successfully rank order the plan quality of a collection of knowledgebased plans. Further, we used 100 pre-optimized plans (20 query x 5 matches) to show that the average normalized MI score can be used as a surrogate of overall plan quality.Plans of lower pre-optimized plan quality tended to improve substantially after optimization, though its final plan quality did not improve to the same level as a plan that has a higher pre-optimized plan quality to begin with. Optimization usually improved PTV coverage slightly while providing substantial dose sparing for both v bladder and rectum of 12.4% and 9.1% respectively. Lastly, we developed new treatment plans for cases selected from an outside institution matched against our sitespecific database. The knowledge-based plans are very comparable to the original manual plan, providing adequate PTV coverage as well as substantial improvement in dose sparing to the rectum and bladder.In conclusion, we found that a site-specific database of prior plans can be effectively used to design new treatment plans for our own institution as well as outside cases. Specifically, knowledge-based plans can provide clinically acceptable planning target volume coverage and clinically acceptable dose sparing to the rectum and bladder. This approach has been demonstrated to improve the efficiency of the treatment planning process, and may potentially improve the quality of patient care by enabling more consistent treatment planning across institutions.vi
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.