Accelerated training of bootstrap aggregation-based deep information extraction systems from cancer pathology reports

Yoon, Hong Jun; Klasky, Hilda B.; Gounley, John; Alawad, Mohammed; Gao, Shang; Durbin, Eric B.; Wu, Xiao Cheng; Stroup, Antoinette M.; Doherty, Jennifer A.; Coyle, Linda; Penberthy, Lynne; Christian, J. Blair; Tourassi, Georgia

doi:10.1016/j.jbi.2020.103564

“…Neural networks and deep learning techniques can be applied for electronic health records (EHR) analysis [6]. The joint application of deep learning models and bootstrap aggregation was able to improve the quality of information extraction systems from cancer pathology reports [7]. In managing diabetes, machine learning (ML) methods can facilitate both diagnosis and prognosis.…”

Section: Introductionmentioning

confidence: 99%

Development and validation of a type 2 diabetes machine learning classification model for clinical decision support framework

Glanz¹,

Dudenkov²,

Velikorodny³

2022

Preprint

0

View full text Add to dashboard Cite

Undiagnosed type 2 diabetes is very common and represents a significant challenge for all national healthcare systems. Although diagnostic criteria and laboratory screening procedures are well-established, clinical tests have limitations, and in many cases diagnosis confirmation and more precise interpretation of the tests results are required. Machine learning methods, when applied to clinical outcomes risk prediction, demonstrate great effectiveness as they recognize specific patterns in data dynamics and thus can be used for identification of at-risk cases where diabetes and complications can be delayed or even prevented. We developed and validated a machine learning classification model for type 2 diabetes that possesses several important advantages over conventional methods (FINDRISC, ADA risk score), including mean values of 0.959, 0.92 and 0.89 for AUC, specificity and sensitivity, respectively. The study results potentially have major clinical implication, as the model is able to exchange data with electronic health records. Thus, it can be employed in clinical decision support framework together with other diabetes, cardiovascular disease models and models for related conditions.

show abstract

“…Neural networks and deep learning techniques can be applied for electronic health records (EHR) analysis (6). The cooperative application of deep learning models and bootstrap aggregation was able to improve the quality of information extraction systems from cancer pathology reports (7). In managing diabetes, machine learning (ML) methods can facilitate both diagnosis and prognosis.…”

Section: Introductionmentioning

confidence: 99%

Development and validation of a type 2 diabetes machine learning classification model for EHR-based diagnostics and clinical decision support

Glanz¹,

Dudenkov²,

Velikorodny³

2022

Preprint

0

View full text Add to dashboard Cite

Undiagnosed type 2 diabetes is very common and represents a significant challenge for all national healthcare systems. Although diagnostic criteria and laboratory screening procedures are well-established, clinical tests have limitations, and in many cases diagnosis confirmation and more precise interpretation of the tests results are required. Machine learning methods, when applied to clinical outcomes risk prediction, demonstrate great effectiveness as they recognize specific patterns in data dynamics and thus can be used for identification of at-risk cases where diabetes and complications can be delayed or even prevented. We developed and validated a machine learning classification model for type 2 diabetes that possesses several important advantages over conventional methods (FINDRISC, ADA risk score), including mean values of 0.959, 0.92 and 0.89 for AUC, specificity and sensitivity, respectively. The study results potentially have major clinical implication, as the model is able to exchange data with electronic health records. Thus, it can be employed in clinical decision support framework together with other diabetes, cardiovascular disease models and models for related conditions

show abstract

“…They also explore grade, locality, and behavior classification, but they are still limited to 29 total classes. Recently, there have been a few large-scale studies of ICD-O-3 classification [3,8,40].…”

Section: Introductionmentioning

confidence: 99%

“…However, their work combines infrequent labels into a single "other" class, thus, not directly training all classes. Furthermore, Yoon et al (2020) [40] introduced a large-scale ICD-O-3 classification method using massive ensembles trained on a high performance computing (HPC) system. Our work focuses on improving the predictive performance of tail codes while training a large-scale ICD-O-3 classification system on standard hardware (i.e., a single GPU).…”

Section: Introductionmentioning

confidence: 99%

“…To the best of our knowledge, this is the first extensive study of machine learning techniques for large-scale ICD-O-3 code detection in pathology reports using a single neural network model on commodity hardware. Prior work has focused on smaller code sets, grouped infrequent classes into an "other" class [1], focused on devleoping techniques more suited for HPC systems [40], or assumed access to large number of pathology reports (e.g., > 900k pathology reports) [40] (which is not accessible to all cancer registires).…”

mentioning

confidence: 99%

See 1 more Smart Citation

Assigning ICD-O-3 codes to pathology reports using neural multi-task training with hierarchical regularization

Rios

¹

,

Durbin

²

,

Hands³

2021

Proceedings of the 12th ACM Conference on Bioinformatics, Computational Biology, and Health Informatics

Self Cite

View full text Add to dashboard Cite

Tracking population-level cancer information is essential for researchers, clinicians, policymakers, and the public. Unfortunately, much of the information is stored as unstructured data in pathology reports. Thus, too process the information, we require either automated extraction techniques or manual curation. Moreover, many of the cancer-related concepts appear infrequently in real world training datasets. Automated extraction is difficult because of the limited data. This study introduces a novel technique that incorporates structured expert knowledge to improve histology and topography code classification models. Using pathology reports collected from the Kentucky Cancer Registry, we introduce a novel multi-task training approach with hierarchical regularization that incorporates structured information about the International Classification of Diseases for Oncology, 3rd Edition classes to improve predictive performance. Overall, we find that our method improves both micro and macro F1. For macro F1, we achieve up to a 6% absolute improvement for topography codes and up to 4% absolute improvement for histology codes.

show abstract

Accelerated training of bootstrap aggregation-based deep information extraction systems from cancer pathology reports

Cited by 15 publications

References 30 publications

Development and validation of a type 2 diabetes machine learning classification model for clinical decision support framework

Development and validation of a type 2 diabetes machine learning classification model for clinical decision support framework

Development and validation of a type 2 diabetes machine learning classification model for EHR-based diagnostics and clinical decision support

Assigning ICD-O-3 codes to pathology reports using neural multi-task training with hierarchical regularization

Contact Info

Product

Resources

About