Natural language processing in urology: Automated extraction of clinical information from histopathology reports of uro-oncology procedures

Objectives

We aimed to automate routine extraction of clinically relevant unstructured information from uro-oncological histopathology reports by applying rule-based and machine learning (ML)/deep learning (DL) methods to develop an oncology focused natural language processing (NLP) algorithm.

Methods

Our algorithm employs a combination of a rule-based approach and support vector machines/neural networks (BioBert/Clinical BERT), and is optimised for accuracy. We randomly extracted 5772 uro-oncological histology reports from 2008 to 2018 from electronic health records (EHRs) and split the data into training and validation datasets in an 80:20 ratio. The training dataset was annotated by medical professionals and reviewed by cancer registrars. The validation dataset was annotated by cancer registrars and defined as the gold standard with which the algorithm outcomes were compared. The accuracy of NLP-parsed data was matched against these human annotation results. We defined an accuracy rate of >95% as "acceptable" by professional human extraction, as per our cancer registry definition.

Results

There were 11 extraction variables in 268 free-text reports. We achieved an accuracy rate of between 61.2% and 99.0% using our algorithm. Of the 11 data fields, a total of 8 data fields met the acceptable accuracy standard, while another 3 data fields had an accuracy rate between 61.2% and 89.7%. Noticeably, the rule-based approach was shown to be more effective and robust in extracting variables of interest. On the other hand, ML/DL models had poorer predictive performances due to highly imbalanced data distribution and variable writing styles between different reports and data used for domain-specific pre-trained models.

Conclusion

We designed an NLP algorithm that can automate clinical information extraction accurately from histopathology reports with an overall average micro accuracy of 93.3%.

A comparative analysis between low-dose-rate brachytherapy and external beam radiation therapy for low- and intermediate-risk prostate cancer in Asian men

OBJECTIVE

To report the long-term clinical outcomes of low-risk (LR) and intermediate-risk (IR) prostate cancer patients treated with low-dose-rate brachytherapy (LDR-BT) and external beam radiation therapy (EBRT).

PATIENTS AND METHODS

Men with biopsy-proven low- and intermediate-risk prostate cancer received EBRT and LDR-BT in an Asian academic center from 2000 to 2019 were reviewed. Kaplan-Meier survival analysis was performed to compare biochemical failure-free survival (bFFS) and overall survival (OS) between LDR and EBRT in the low- and intermediate-risk cohorts.

RESULTS

642 patients (521 EBRT and 121 LDR-BT) with low- and intermediate-risk prostate cancer were included for analysis. In the intermediate-risk group, 5- and 10-year bFFS was 96%, 89% and 86%, 61% for LDR-BT and EBRT, respectively. LDR-BT was associated with a statistically significant improvement of bFFS in the intermediate-risk cohort (HR 2.7, p = 0.02). In the low-risk cohort, no difference of bFFS was found between LDR-BT and EBRT (HR 1.9, p = 0.08). Hormone therapy was more common in EBRT than LDR-BT for intermediate-risk group (71% versus 44%, p < 0.05). Prostate cancer-specific mortality was low in both EBRT (1%) and LDR-BT (2%) cohorts. No significant difference in OS was found between LDR-BT and EBRT in low- and intermediate-risk group (HR 2.1, p = 0.2 and HR = 1.7, p = 0.3).

CONCLUSION

In our retrospective study, LDR-BT is associated with superior bFFS compared with EBRT in Asian men with intermediate-risk prostate cancer.

Regional registration of whole slide image stacks containing major histological artifacts

Background

High resolution 2D whole slide imaging provides rich information about the tissue structure. This information can be a lot richer if these 2D images can be stacked into a 3D tissue volume. A 3D analysis, however, requires accurate reconstruction of the tissue volume from the 2D image stack. This task is not trivial due to the distortions such as tissue tearing, folding and missing at each slide. Performing registration for the whole tissue slices may be adversely affected by distorted tissue regions. Consequently, regional registration is found to be more effective. In this paper, we propose a new approach to an accurate and robust registration of regions of interest for whole slide images. We introduce the idea of multi-scale attention for registration.

Results

Using mean similarity index as the metric, the proposed algorithm (mean ± SD \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.84 \pm 0.11$$\end{document}0.84±0.11) followed by a fine registration algorithm (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.86 \pm 0.08$$\end{document}0.86±0.08) outperformed the state-of-the-art linear whole tissue registration algorithm (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.74 \pm 0.19$$\end{document}0.74±0.19) and the regional version of this algorithm (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.81 \pm 0.15$$\end{document}0.81±0.15). The proposed algorithm also outperforms the state-of-the-art nonlinear registration algorithm (original: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.82 \pm 0.12$$\end{document}0.82±0.12, regional: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.77 \pm 0.22$$\end{document}0.77±0.22) for whole slide images and a recently proposed patch-based registration algorithm (patch size 256: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.79 \pm 0.16$$\end{document}0.79±0.16 , patch size 512: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.77 \pm 0.16$$\end{document}0.77±0.16) for medical images.

Conclusion

Using multi-scale attention mechanism leads to a more robust and accurate solution to the problem of regional registration of whole slide images corrupted in some parts by major histological artifacts in the imaged tissue.

Automated Renal Cancer Grading Using Nuclear Pleomorphic Patterns

PURPOSE

Nuclear pleomorphic patterns are essential for Fuhrman grading of clear cell renal cell carcinoma (ccRCC). Manual observation of renal histopathologic slides may lead to subjective and inconsistent assessment between pathologists. An automated, image-based system that classifies ccRCC slides by quantifying nuclear pleomorphic patterns in an objective and consistent interpretable fashion can aid pathologists in histopathologic assessment.

METHODS

In the current study, histopathologic tissue slides of 59 patients with ccRCC who underwent surgery at Singapore General Hospital were assembled retrospectively. An automated image classification pipeline detects and analyzes prominent nucleoli in ccRCC images to classify them as either low (Fuhrman grade 1 and 2) or high (Fuhrman grade 3 and 4). The pipeline uses machine learning and image pixel intensity-based feature extraction techniques for nuclear analysis. We trained classification systems that concurrently analyze different permutations of multiple prominent nucleoli image patches.

RESULTS

Given the parameters for feature combination and extraction, we present experimental results across various configurations for the classification of a given ccRCC histopathologic image. We also demonstrate that the image score used by the pipeline, termed fraction value, is correlated ( R = 0.59) with an existing multigene assay-based scoring system that has previously been demonstrated to be a strong indicator of prognosis in patients with ccRCC.

CONCLUSION

The current method provides an objective and fully automated way by which to process pathologic slides. The correlation study with a multigene assay-based scoring system also allows us to provide quantitative interpretation for already established nuclear pleomorphic patterns in ccRCC. This method can be extended to other cancers whose corresponding grading systems use nuclear pattern information.