利用机器学习和神经网络实现癌症登记的转型.pdf

1、Transforming cancer registries with machine learning and neural networksAdhari Abdullah AlZaabiCollege of Medicine and Health Sciences Sultan Qaboos Universityadharisqu.edu.omOutlineIntroductionMethodsResultsConclusionPlease,consider this is work in progress!Cancer Burden(leading cause of death)Canc

2、er Burden(leading cause of death)Oman and Gulf Cooperation Council cancer incidencehttps:/pmc.ncbi.nlm.nih.gov/articles/PMC11403302/Cancer care continuum&surveillance Cancer Registry Track trends over time(Incidence,mortality and survival)Allocate resources,prevention,screening and treatment Evaluat

3、e effectiveness of cancer programs and policiesCancer RegistrationManual Cancer RegistrationChallenges-Manual abstractionExpensiveProne to errorsAffect quality,completeness,Accuracy and timeliness dataUn-sustainableManual abstraction-delayed reportingCancer incidence reports are often not available

4、until 24 months or greater after a diagnosisSolution using AIApplies natural language processing(NLP)and deep learning algorithms to population-based cancer data To develop scalable NLP tools for deep text comprehension of unstructured clinical text To enable automated and accurate capture of report

5、able cancer surveillance data elementsSolutions-Automate Data collection using ML&NLPUnstructured data Clinical text Clinical text contextcontext is important is important Clinical text is temporalClinical text is temporalNLP ETL LayerMethods Methods-NLP PipelineNLP PipelineMethods-Annotation or lab

6、eling(Gold standard)-Breast-CRC-Thyroid-Prostate-LungObjective 1Consoloditate TNM staging from Clinical text Clinical text to TNM staging“TUMOR INVADES INTO BUT NOT THROUGH VISCERAL PLEURA”=stage T2“8 LYMPH NODES NEGATIVE FOR TUMOR”=stage N0DatasetResults(TNM document-level)Resul