Data-Efficient Computational Pathology Platform for Faster and Cheaper Breast Cancer Subtype Identifications: Development of a Deep Learning Model.

BACKGROUND: Breast cancer subtyping is a crucial step in determining therapeutic options, but the molecular examination based on immunohistochemical staining is expensive and time-consuming. Deep learning opens up the possibility to predict the subtypes based on the morphological information from hematoxylin and eosin staining, a much cheaper and faster alternative. However, training the predictive model conventionally requires a large number of histology images, which is challenging to collect by a single institute. OBJECTIVE: We aimed to develop a data-efficient computational pathology platform, 3DHistoNet, which is capable of learning from z-stacked histology images to accurately predict breast cancer subtypes with a small sample size. METHODS: We retrospectively examined 401 cases of patients with primary breast carcinoma diagnosed between 2018 and 2020 at the Department of Pathology, National Cancer Center, South Korea. Pathology slides of the patients with breast carcinoma were prepared according to the standard protocols. Age, gender, histologic grade, hormone receptor (estrogen receptor [ER], progesterone receptor [PR], and androgen receptor [AR]) status, erb-B2 receptor tyrosine kinase 2 (HER2) status, and Ki-67 index were evaluated by reviewing medical charts and pathological records. RESULTS: The area under the receiver operating characteristic curve and decision curve were analyzed to evaluate the performance of our 3DHistoNet platform for predicting the ER, PR, AR, HER2, and Ki67 subtype biomarkers with 5-fold cross-validation. We demonstrated that 3DHistoNet can predict all clinically important biomarkers (ER, PR, AR, HER2, and Ki67) with performance exceeding the conventional multiple instance learning models by a considerable margin (area under the receiver operating characteristic curve: 0.75-0.91 vs 0.67-0.8). We further showed that our z-stack histology scanning method can make up for insufficient training data sets without any additional cost incurred. Finally, 3DHistoNet offered an additional capability to generate attention maps that reveal correlations between Ki67 and histomorphological features, which renders the hematoxylin and eosin image in higher fidelity to the pathologist. CONCLUSIONS: Our stand-alone, data-efficient pathology platform that can both generate z-stacked images and predict key biomarkers is an appealing tool for breast cancer diagnosis. Its development would encourage morphology-based diagnosis, which is faster, cheaper, and less error-prone compared to the protein quantification method based on immunohistochemical staining.

FCSN: Global Context Aware Segmentation by Learning the Fourier Coefficients of Objects in Medical Images.

The encoder-decoder model is a commonly used Deep Neural Network (DNN) model for medical image segmentation. Conventional encoder-decoder models make pixel-wise predictions focusing heavily on local patterns around the pixel. This makes it challenging to give segmentation that preserves the object's shape and topology, which often requires an understanding of the global context. In this work, we propose a Fourier Coefficient Segmentation Network (FCSN)-a novel global context-aware DNN model that segments an object by learning the complex Fourier coefficients of the object's masks. The Fourier coefficients are calculated by integrating over the whole contour. Therefore, for our model to make a precise estimation of the coefficients, the model is motivated to incorporate the global context of the object, leading to a more accurate segmentation of the object's shape. This global context awareness also makes our model robust to unseen local perturbations during inference, such as additive noise or motion blur that are prevalent in medical images. We compare FCSN with other state-of-the-art global context-aware models (UNet++, DeepLabV3+, UNETR) on 5 medical image segmentation tasks, of which 3 are camera imaging datasets (ISIC_2018, RIM_CUP, RIM_DISC) and 2 are medical imaging datasets (PROSTATE, FETAL). When FCSN is compared with UNETR, FCSN attains significantly lower Hausdorff scores with 19.14 (6%), 17.42 (6%), 9.16 (14%), 11.18 (22%), and 5.98 (6%) for ISIC_2018, RIM_CUP, RIM_DISC, PROSTATE, and FETAL tasks respectively. Moreover, FCSN is lightweight by discarding the decoder module, which incurs significant computational overhead. FCSN only requires 29.7 M parameters which are 75.6 M and 9.9 M fewer parameters than UNETR and DeepLabV3+, respectively. FCSN attains inference and training speeds of 1.6 ms/img and 6.3 ms/img, which is 8× and 3× faster than UNet and UNETR. The code for FCSN is made publicly available at https://github.com/nus-mornin-lab/FCSN.

Transformation of Electronic Health Records and Questionnaire Data to OMOP CDM: A Feasibility Study Using SG_T2DM Dataset.

BACKGROUND: Diabetes mellitus (DM) is an important public health concern in Singapore and places a massive burden on health care spending. Tackling chronic diseases such as DM requires innovative strategies to integrate patients' data from diverse sources and use scientific discovery to inform clinical practice that can help better manage the disease. The Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) was chosen as the framework for integrating data with disparate formats. OBJECTIVE: The study aimed to evaluate the feasibility of converting Singapore based data source, comprising of electronic health records (EHR), cognitive and depression assessment questionnaire data to OMOP CDM standard. Additionally, we also validate whether our OMOP CDM instance is fit for the purpose of research by executing a simple treatment pathways study using Atlas, a graphical user interface tool to conduct analysis on OMOP CDM data as a proof of concept. METHODS: We used de-identified EHR, cognitive, and depression assessment questionnaires data from a tertiary care hospital in Singapore to convert it to version 5.3.1 of OMOP CDM standard. We evaluate the OMOP CDM conversion by (1) assessing the mapping coverage (that is the percentage of source terms mapped to OMOP CDM standard); (2) local raw dataset versus CDM dataset analysis; and (3) Implementing Harmonized Intrinsic Data Quality Framework using an open-source R package called Data Quality Dashboard. RESULTS: The content coverage of OMOP CDM vocabularies is more than 90% for clinical data, but only around 11% for questionnaire data. The comparison of characteristics between source and target data returned consistent results and our transformed data did not pass 38 (1.4%) out of 2,622 quality checks. CONCLUSION: Adoption of OMOP CDM at our site demonstrated that EHR data are feasible for standardization with minimal information loss, whereas challenges remain for standardizing cognitive and depression assessment questionnaire data that requires further work.

A New Model for Risk Stratification of Patients With Acute Pulmonary Embolism.

Pulmonary embolism (PE) is associated with mortality. There are many clinical prediction tools to predict early mortality in acute PE but little consensus on which is best. Our study aims to validate existing prediction tools and derive a predictive model that can be applied to all patients with acute PE in both inpatient and outpatient settings. This is a retrospective cohort study of patients with acute PE. For each patient, the Pulmonary Embolism Severity Index (PESI), simplified PESI (sPESI), European Society of Cardiology (ESC), and Angriman scores were calculated. Scores were assessed by the area under the receive-operating curve (AUC) for 30-day, all-cause mortality. To develop a new prognostic model, elastic logistic regression was used on the derivation cohort to estimate ß-coefficients of 8 different variables; these were normalized to weigh them. A total of 321 patients (mean age 60±17 years) were included. Overall 30-day mortality was 10.3%. None of the scores performed well; the AUCs for the PESI, sPESI, ESC, and Angriman scores were 0.67 (95% confidence interval [CI], 0.57-0.77), 0.58 (0.48-0.69), 0.65 (0.55-0.75), and 0.67 (0.57-0.76), respectively. Our new prediction model outperformed PESI, with an AUC of 0.82 (95% CI, 0.76-0.88). At a cutoff score of 100, 195 (60.1%) patients were classified as low risk. Thirty-day mortality was 2.1% (95% CI, 0.8%-5.2%) and 23.0% (16.5%-31.1%) for low- and high-risk groups, respectively (P < .001). In conclusion, we have developed a new model that outperforms existing prediction tools in all comers with PE. However, further validation on external cohorts is required before application.