Artificial Intelligence-Assisted Cancer Status Detection in Radiology Reports.

Cancer research is dependent on accurate and relevant information of patient's medical journey. Data in radiology reports are of extreme value but lack consistent structure for direct use in analytics. At Memorial Sloan Kettering Cancer Center (MSKCC), the radiology reports are curated using gold-standard approach of using human annotators. However, the manual process of curating large volume of retrospective data slows the pace of cancer research. Manual curation process is sensitive to volume of reports, number of data elements and nature of reports and demand appropriate skillset. In this work, we explore state of the art methods in artificial intelligence (AI) and implement end-to-end pipeline for fast and accurate annotation of radiology reports. Language models (LM) are trained using curated data by approaching curation as multiclass or multilabel classification problem. The classification tasks are to predict multiple imaging scan sites, presence of cancer and cancer status from the reports. The trained natural language processing (NLP) model classifiers achieve high weighted F1 score and accuracy. We propose and demonstrate the use of these models to assist in the manual curation process which results in higher accuracy and F1 score with lesser time and cost, thus improving efforts of cancer research. SIGNIFICANCE: Extraction of structured data in radiology for cancer research with manual process is laborious. Using AI for extraction of data elements is achieved using NLP models' assistance is faster and more accurate.

How to use replicate weights in health survey analysis using the National Nutrition and Physical Activity Survey as an example.

OBJECTIVE: To conduct nutrition-related analyses on large-scale health surveys, two aspects of the survey must be incorporated into the analysis: the sampling weights and the sample design; a practice which is not always observed. The present paper compares three analyses: (1) unweighted; (2) weighted but not accounting for the complex sample design; and (3) weighted and accounting for the complex design using replicate weights. DESIGN: Descriptive statistics are computed and a logistic regression investigation of being overweight/obese is conducted using Stata. SETTING: Cross-sectional health survey with complex sample design where replicate weights are supplied rather than the variables containing sample design information. PARTICIPANTS: Responding adults from the National Nutrition and Physical Activity Survey (NNPAS) part of the Australian Health Survey (2011-2013). RESULTS: Unweighted analysis produces biased estimates and incorrect estimates of se. Adjusting for the sampling weights gives unbiased estimates but incorrect se estimates. Incorporating both the sampling weights and the sample design results in unbiased estimates and the correct se estimates. This can affect interpretation; for example, the incorrect estimate of the OR for being a current smoker in the unweighted analysis was 1·20 (95 % CI 1·06, 1·37), t= 2·89, P = 0·004, suggesting a statistically significant relationship with being overweight/obese. When the sampling weights and complex sample design are correctly incorporated, the results are no longer statistically significant: OR = 1·06 (95 % CI 0·89, 1·27), t = 0·71, P = 0·480. CONCLUSIONS: Correct incorporation of the sampling weights and sample design is crucial for valid inference from survey data.