Automated prostate gland segmentation in challenging clinical cases: comparison of three artificial intelligence methods

OBJECTIVE

Automated methods for prostate segmentation on MRI are typically developed under ideal scanning and anatomical conditions. This study evaluates three different prostate segmentation AI algorithms in a challenging population of patients with prior treatments, variable anatomic characteristics, complex clinical history, or atypical MRI acquisition parameters.

MATERIALS AND METHODS

A single institution retrospective database was queried for the following conditions at prostate MRI: prior prostate-specific oncologic treatment, transurethral resection of the prostate (TURP), abdominal perineal resection (APR), hip prosthesis (HP), diversity of prostate volumes (large ≥ 150 cc, small ≤ 25 cc), whole gland tumor burden, magnet strength, noted poor quality, and various scanners (outside/vendors). Final inclusion criteria required availability of axial T2-weighted (T2W) sequence and corresponding prostate organ segmentation from an expert radiologist. Three previously developed algorithms were evaluated: (1) deep learning (DL)-based model, (2) commercially available shape-based model, and (3) federated DL-based model. Dice Similarity Coefficient (DSC) was calculated compared to expert. DSC by model and scan factors were evaluated with Wilcox signed-rank test and linear mixed effects (LMER) model.

RESULTS

683 scans (651 patients) met inclusion criteria (mean prostate volume 60.1 cc [9.05-329 cc]). Overall DSC scores for models 1, 2, and 3 were 0.916 (0.707-0.971), 0.873 (0-0.997), and 0.894 (0.025-0.961), respectively, with DL-based models demonstrating significantly higher performance (p < 0.01). In sub-group analysis by factors, Model 1 outperformed Model 2 (all p < 0.05) and Model 3 (all p < 0.001). Performance of all models was negatively impacted by prostate volume and poor signal quality (p < 0.01). Shape-based factors influenced DL models (p < 0.001) while signal factors influenced all (p < 0.001).

CONCLUSION

Factors affecting anatomical and signal conditions of the prostate gland can adversely impact both DL and non-deep learning-based segmentation models.

A multitask approach for automated detection and segmentation of thyroid nodules in ultrasound images

An increase in the incidence and diagnosis of thyroid nodules and thyroid cancer underscores the need for a better approach to nodule detection and risk stratification in ultrasound (US) images that can reduce healthcare costs, patient discomfort, and unnecessary invasive procedures. However, variability in ultrasound technique and interpretation makes the diagnostic process partially subjective. Therefore, an automated approach that detects and segments nodules could improve performance on downstream tasks, such as risk stratification. Ultrasound studies were acquired from 280 patients at UCLA Health, totaling 9888 images, and annotated by collaborating radiologists. Current deep learning architectures for segmentation are typically semi-automated because they are evaluated solely on images known to have nodules and do not assess ability to identify suspicious images. However, the proposed multitask approach both detects suspicious images and segments potential nodules; this allows for a clinically translatable model that aptly parallels the workflow for thyroid nodule assessment. The multitask approach is centered on an anomaly detection (AD) module that can be integrated with any UNet architecture variant to improve image-level nodule detection. Of the evaluated multitask models, a UNet with a ImageNet pretrained encoder and AD achieved the highest F1 score of 0.839 and image-wide Dice similarity coefficient of 0.808 on the hold-out test set. Furthermore, models were evaluated on two external validations datasets to demonstrate generalizability and robustness to data variability. Ultimately, the proposed architecture is an automated multitask method that expands on previous methods by successfully both detecting and segmenting nodules in ultrasound.

Automated Assessment of DWI-FLAIR Mismatch in Patients with Acute Ischemic Stroke: Added Value to Routine Clinical Practice

BACKGROUND AND PURPOSE

DWI-FLAIR mismatch is used to determine thrombolytic eligibility in patients with acute ischemic stroke (AIS) when time since stroke onset (TSS) is unknown. Commercial software packages have been developed for automated DWI-FLAIR classification. We aimed to use e-Stroke software (Brainomix, Oxford, UK) for automated classification of DWI-FLAIR mismatch in a cohort of patients with AIS and in a comparative analysis with two expert neuroradiologists.

MATERIALS AND METHODS

In this retrospective study, patients with AIS who had MRI and known TSS were included. DWI-FLAIR mismatch was evaluated by two neuroradiologists blinded to TSS and automatically by e-Stroke software. After 4 weeks, the neuroradiologists reevaluated the MRIs, this time equipped with automated predicted e-Stroke results as a computer assisted tool (CAT). Diagnostic performances of e-Stroke software and neuroradiologists were evaluated for prediction of DWI-FLAIR mismatch status.

RESULTS

A total of 157 patients met inclusion criteria. A total of 82 patients (52%) had TSS ≤ 4.5 hours. Using consensus reads, 81 patients (51.5%) had DWI-FLAIR mismatch. The diagnostic accuracy (AUC/sensitivity/specificity) of e-Stroke software for determination of DWI-FLAIR mismatch was 0.72/90.0/53.9. The diagnostic accuracy (AUC/sensitivity/specificity) for neuroradiologist 1 and 2 was 0.76/69.1/84.2 and was 0.82/91.4/73.7 respectively, both significantly (p<0.05) improved to 0.83/79.0/86.8 and 0.89/92.6/85.5 respectively following the use of e-Stroke predictions as CAT. The interrater agreement (K) for determination of DWI-FLAIR status was improved from 0.49 to 0.57 following the use of CAT.

CONCLUSIONS

Automated quantitative approach for DWI-FLAIR mismatch provides comparable results to human experts and can improve diagnostic accuracies of expert neuroradiologists in determination of DWI-FLAIR status.ABBREVIATIONS: AIS: Acute ischemic stroke; CAT: Computer assisted tool; TSS: Time since stroke onset.

SLIViT: a general AI framework for clinical-feature diagnosis from limited 3D biomedical-imaging data

We present SLIViT, a deep-learning framework that accurately measures disease-related risk factors in volumetric biomedical imaging, such as magnetic resonance imaging (MRI) scans, optical coherence tomography (OCT) scans, and ultrasound videos. To evaluate SLIViT, we applied it to five different datasets of these three different data modalities tackling seven learning tasks (including both classification and regression) and found that it consistently and significantly outperforms domain-specific state-of-the-art models, typically improving performance (ROC AUC or correlation) by 0.1-0.4. Notably, compared to existing approaches, SLIViT can be applied even when only a small number of annotated training samples is available, which is often a constraint in medical applications. When trained on less than 700 annotated volumes, SLIViT obtained accuracy comparable to trained clinical specialists while reducing annotation time by a factor of 5,000 demonstrating its utility to automate and expedite ongoing research and other practical clinical scenarios. *Oren Avram and Berkin Durmus equally contributed to this work. **Srinivas R. Sadda and Eran Halperin jointly supervised this study.

Federated Learning with Research Prototypes: Application to Multi-Center MRI-based Detection of Prostate Cancer with Diverse Histopathology

RATIONALE AND OBJECTIVES

Early prostate cancer detection and staging from MRI is extremely challenging for both radiologists and deep learning algorithms, but the potential to learn from large and diverse datasets remains a promising avenue to increase their performance within and across institutions. To enable this for prototype-stage algorithms, where the majority of existing research remains, we introduce a flexible federated learning framework for cross-site training, validation, and evaluation of custom deep learning prostate cancer detection algorithms.

MATERIALS AND METHODS

We introduce an abstraction of prostate cancer groundtruth that represents diverse annotation and histopathology data. We maximize use of this groundtruth if and when they are available using UCNet, a custom 3D UNet that enables simultaneous supervision of pixel-wise, region-wise, and gland-wise classification. We leverage these modules to perform cross-site federated training using 1400+ heterogeneous multi-parameteric prostate MRI exams from two University hospitals.

RESULTS

We observe a positive result, with significant improvements in cross-site generalization performance with negligible intra-site performance degradation for both lesion segmentation and per-lesion binary classification of clinically-significant prostate cancer. Cross-site lesion segmentation performance intersection-over-union (IoU) improved by 100%, while cross-site lesion classification performance overall accuracy improved by 9.5-14.8%, depending on the optimal checkpoint selected by each site.

CONCLUSION

Federated learning can improve the generalization performance of prostate cancer detection models across institutions while protecting patient health information and institution-specific code and data. However, even more data and participating institutions are likely required to improve the absolute performance of prostate cancer classification models. To enable adoption of federated learning with limited re-engineering of federated components, we open-source our FLtools system at https://federated.ucsf.edu, including examples that can be easily adapted to other medical imaging deep learning projects.

A Multitask Approach for Automated Detection and Segmentation of Thyroid Nodules in Ultrasound Images

An increase in the incidence and diagnosis of thyroid nodules and thyroid cancer underscores the need for a better approach to nodule detection and risk stratification in ultrasound (US) images that can reduce healthcare costs, patient discomfort, and unnecessary invasive procedures. However, variability in ultrasound technique and interpretation makes the diagnostic process partially subjective. Therefore, an automated approach that detects and segments nodules could improve performance on downstream tasks, such as risk stratification.Current deep learning architectures for segmentation are typically semi-automated because they are evaluated solely on images known to have nodules and do not assess ability to identify suspicious images. However, the proposed multitask approach both detects suspicious images and segments potential nodules; this allows for a clinically translatable model that aptly parallels the workflow for thyroid nodule assessment. The multitask approach is centered on an anomaly detection (AD) module that can be integrated with any U-Net architecture variant to improve image-level nodule detection. Ultrasound studies were acquired from 280 patients at UCLA Health, totaling 9,888 images, and annotated by collaborating radiologists. Of the evaluated models, a multi-scale UNet (MSUNet) with AD achieved the highest F1 score of 0.829 and image-wide Dice similarity coefficient of 0.782 on our hold-out test set. Furthermore, models were evaluated on two external validations datasets to demonstrate generalizability and robustness to data variability. Ultimately, the proposed architecture is an automated multitask method that expands on previous methods by successfully both detecting and segmenting nodules in ultrasound.

Synthesizing MR Image Contrast Enhancement Using 3D High-Resolution ConvNets

OBJECTIVE

Gadolinium-based contrast agents (GBCAs) have been widely used to better visualize disease in brain magnetic resonance imaging (MRI). However, gadolinium deposition within the brain and body has raised safety concerns about the use of GBCAs. Therefore, the development of novel approaches that can decrease or even eliminate GBCA exposure while providing similar contrast information would be of significant use clinically.

METHODS

In this work, we present a deep learning based approach for contrast-enhanced T1 synthesis on brain tumor patients. A 3D high-resolution fully convolutional network (FCN), which maintains high resolution information through processing and aggregates multi-scale information in parallel, is designed to map pre-contrast MRI sequences to contrast-enhanced MRI sequences. Specifically, three pre-contrast MRI sequences, T1, T2 and apparent diffusion coefficient map (ADC), are utilized as inputs and the post-contrast T1 sequences are utilized as target output. To alleviate the data imbalance problem between normal tissues and the tumor regions, we introduce a local loss to improve the contribution of the tumor regions, which leads to better enhancement results on tumors.

RESULTS

Extensive quantitative and visual assessments are performed, with our proposed model achieving a PSNR of 28.24 dB in the brain and 21.2 dB in tumor regions.

CONCLUSION AND SIGNIFICANCE

Our results suggest the potential of substituting GBCAs with synthetic contrast images generated via deep learning.

Institutional Strategies to Maintain and Grow Imaging Research During the COVID-19 Pandemic

Understanding imaging research experiences, challenges, and strategies for academic radiology departments during and after COVID-19 is critical to prepare for future disruptive events. We summarize key insights and programmatic initiatives at major academic hospitals across the world, based on literature review and meetings of the Radiological Society of North America Vice Chairs of Research (RSNA VCR) group. Through expert discussion and case studies, we provide suggested guidelines to maintain and grow radiology research in the postpandemic era.

Identifying acute ischemic stroke patients within the thrombolytic treatment window using deep learning

BACKGROUND AND PURPOSE

Treatment of acute ischemic stroke is heavily contingent upon time, as there is a strong relationship between time clock and tissue progression. Work has established imaging biomarker assessments as surrogates for time since stroke (TSS), namely, by comparing signal mismatch between diffusion-weighted imaging (DWI) and fluid-attenuated inversion recovery (FLAIR) imaging. Our goal was to develop an automatic technique for determining TSS from imaging that does not require subspecialist radiology expertise.

METHODS

Using 772 patients (66 ± 9 years, 319 women), we developed and externally evaluated a deep learning network for classifying TSS from MR images and compared algorithm predictions to neuroradiologist assessments of DWI-FLAIR mismatch. Models were trained to classify TSS within 4.5 hours and performance metrics with confidence intervals were reported on both internal and external evaluation sets.

RESULTS

Three board-certified neuroradiologists' DWI-FLAIR mismatch assessments, based on majority vote, yielded a sensitivity of .62, a specificity of .86, and a Fleiss' kappa of .46 when used to classify TSS. The deep learning method performed similarly to radiologists and outperformed previously reported methods, with the best model achieving an average evaluation accuracy, sensitivity, and specificity of .726, .712, and .741, respectively, on an internal cohort and .724, .757, and .679, respectively, on an external cohort.

CONCLUSION

Our model achieved higher generalization performance on external evaluation datasets than the current state-of-the-art for TSS classification. These results demonstrate the potential of automatic assessment of onset time from imaging without the need for expertly trained radiologists.

PathAL: An Active Learning Framework for Histopathology Image Analysis

Deep neural networks, in particular convolutional networks, have rapidly become a popular choice for analyzing histopathology images. However, training these models relies heavily on a large number of samples manually annotated by experts, which is cumbersome and expensive. In addition, it is difficult to obtain a perfect set of labels due to the variability between expert annotations. This paper presents a novel active learning (AL) framework for histopathology image analysis, named PathAL. To reduce the required number of expert annotations, PathAL selects two groups of unlabeled data in each training iteration: one "informative" sample that requires additional expert annotation, and one "confident predictive" sample that is automatically added to the training set using the model's pseudo-labels. To reduce the impact of the noisy-labeled samples in the training set, PathAL systematically identifies noisy samples and excludes them to improve the generalization of the model. Our model advances the existing AL method for medical image analysis in two ways. First, we present a selection strategy to improve classification performance with fewer manual annotations. Unlike traditional methods focusing only on finding the most uncertain samples with low prediction confidence, we discover a large number of high confidence samples from the unlabeled set and automatically add them for training with assigned pseudo-labels. Second, we design a method to distinguish between noisy samples and hard samples using a heuristic approach. We exclude the noisy samples while preserving the hard samples to improve model performance. Extensive experiments demonstrate that our proposed PathAL framework achieves promising results on a prostate cancer Gleason grading task, obtaining similar performance with 40% fewer annotations compared to the fully supervised learning scenario. An ablation study is provided to analyze the effectiveness of each component in PathAL, and a pathologist reader study is conducted to validate our proposed algorithm.

Electronic Health Record-Integrated Tumor Board Application to Save Preparation Time and Reduce Errors

PURPOSE

Multidisciplinary oncology meetings, or tumor boards (TBs), ensure and facilitate communication between specialties regarding the management of cancer cases to improve patient care. The organization of TB and the preparation and presentation of patient cases are typically inefficient processes that require the exchange of patient information via e-mail, the hunting for data and images in the electronic health record, and the copying and pasting of patient data into desktop presentation software.

METHODS

We implemented a standards-based electronic health record-integrated application that automated several aspects of TB organization and preparation. We hypothesized that this application would increase the efficiency of TB preparation, reduce errors in patient entry, and enhance communication with the clinical team. Our experimental design used a prospective evaluation by pathologists who were timed in preparing for weekly TBs using both the new application and the conventional method. In addition, patient data entry errors associated with each method were tracked, and TB attendees completed a survey evaluating satisfaction with the new application.

RESULTS

The total time savings for TB preparation using the digital TB application over the conventional method was 5 hours and 19 minutes, representing a 45% reduction in preparation time (P < .01). Survey results showed that 91% of respondents preferred the digital method and believed that it improved the flow of the TB meeting. In addition, most believed that the digital method had an impact on subsequent patient care.

CONCLUSION

This study provides further evidence that new electronic systems have the potential to significantly improve the overall TB paradigm by optimizing and enhancing case organization, preparation, and presentation.

A Semi-Supervised Learning Framework to Leverage Proxy Information for Stroke MRI Analysis

Treating acute ischemic stroke (AIS) patients is a time-sensitive endeavor, as therapies target areas experiencing ischemia to prevent irreversible damage to brain tissue. Depending on how an AIS is progressing, thrombolytics such as tissue-plasminogen activator (tPA) may be administered within a short therapeutic window. The underlying conditions for optimal treatment are varied. While previous clinical guidelines only permitted tPA to be administered to patients with a known onset within 4.5 hours, clinical trials demonstrated that patients with signal intensity differences between diffusion-weighted imaging (DWI) and fluid-attenuated inversion recovery (FLAIR) sequences in an MRI study can benefit from thrombolytic therapy. This intensity difference, known as DWI-FLAIR mismatch, is prone to high inter-reader variability. Thus, a paradigm exists where onset time serves as a weak proxy for DWI-FLAIR mismatch. In this study, we sought to detect DWI-FLAIR mismatch in an automated fashion, and we compared this to assessments done by three expert neuroradiologists. Our approach involved training a deep learning model on MRI to classify tissue clock and leveraging time clock as a weak proxy label to supplement training in a semi-supervised learning (SSL) framework. We evaluate our deep learning model by testing it on an unseen dataset from an external institution. In total, our proposed framework was able to improve detection of DWI-FLAIR mismatch, achieving a top ROC-AUC of 74.30%. Our study illustrated that incorporating clinical proxy information into SSL can improve model optimization by increasing the fidelity of unlabeled samples included in the training process.

Bidirectional Representation Learning From Transformers Using Multimodal Electronic Health Record Data to Predict Depression

Advancements in machine learning algorithms have had a beneficial impact on representation learning, classification, and prediction models built using electronic health record (EHR) data. Effort has been put both on increasing models' overall performance as well as improving their interpretability, particularly regarding the decision-making process. In this study, we present a temporal deep learning model to perform bidirectional representation learning on EHR sequences with a transformer architecture to predict future diagnosis of depression. This model is able to aggregate five heterogenous and high-dimensional data sources from the EHR and process them in a temporal manner for chronic disease prediction at various prediction windows. We applied the current trend of pretraining and fine-tuning on EHR data to outperform the current state-of-the-art in chronic disease prediction, and to demonstrate the underlying relation between EHR codes in the sequence. The model generated the highest increases of precision-recall area under the curve (PRAUC) from 0.70 to 0.76 in depression prediction compared to the best baseline model. Furthermore, the self-attention weights in each sequence quantitatively demonstrated the inner relationship between various codes, which improved the model's interpretability. These results demonstrate the model's ability to utilize heterogeneous EHR data to predict depression while achieving high accuracy and interpretability, which may facilitate constructing clinical decision support systems in the future for chronic disease screening and early detection.

Attention-Guided Discriminative Region Localization and Label Distribution Learning for Bone Age Assessment

Bone age assessment (BAA) is clinically important as it can be used to diagnose endocrine and metabolic disorders during child development. Existing deep learning based methods for classifying bone age use the global image as input, or exploit local information by annotating extra bounding boxes or key points. However, training with the global image underutilizes discriminative local information, while providing extra annotations is expensive and subjective. In this paper, we propose an attention-guided approach to automatically localize the discriminative regions for BAA without any extra annotations. Specifically, we first train a classification model to learn the attention maps of the discriminative regions, finding the hand region, the most discriminative region (the carpal bones), and the next most discriminative region (the metacarpal bones). Guided by those attention maps, we then crop the informative local regions from the original image and aggregate different regions for BAA. Instead of taking BAA as a general regression task, which is suboptimal due to the label ambiguity problem in the age label space, we propose using joint age distribution learning and expectation regression, which makes use of the ordinal relationship among hand images with different individual ages and leads to more robust age estimation. Extensive experiments are conducted on the RSNA pediatric bone age data set. {\color{red} Without using extra manual} annotations, our method achieves competitive results compared with existing state-of-the-art deep learning-based methods that require manual annotation. Code is available at \url{https://github.com/chenchao666/Bone-Age-Assessment}.

Harnessing clinical annotations to improve deep learning performance in prostate segmentation

Purpose

Developing large-scale datasets with research-quality annotations is challenging due to the high cost of refining clinically generated markup into high precision annotations. We evaluated the direct use of a large dataset with only clinically generated annotations in development of high-performance segmentation models for small research-quality challenge datasets.

Materials and methods

We used a large retrospective dataset from our institution comprised of 1,620 clinically generated segmentations, and two challenge datasets (PROMISE12: 50 patients, ProstateX-2: 99 patients). We trained a 3D U-Net convolutional neural network (CNN) segmentation model using our entire dataset, and used that model as a template to train models on the challenge datasets. We also trained versions of the template model using ablated proportions of our dataset, and evaluated the relative benefit of those templates for the final models. Finally, we trained a version of the template model using an out-of-domain brain cancer dataset, and evaluated the relevant benefit of that template for the final models. We used five-fold cross-validation (CV) for all training and evaluation across our entire dataset.

Results

Our model achieves state-of-the-art performance on our large dataset (mean overall Dice 0.916, average Hausdorff distance 0.135 across CV folds). Using this model as a pre-trained template for refining on two external datasets significantly enhanced performance (30% and 49% enhancement in Dice scores respectively). Mean overall Dice and mean average Hausdorff distance were 0.912 and 0.15 for the ProstateX-2 dataset, and 0.852 and 0.581 for the PROMISE12 dataset. Using even small quantities of data to train the template enhanced performance, with significant improvements using 5% or more of the data.

Conclusion

We trained a state-of-the-art model using unrefined clinical prostate annotations and found that its use as a template model significantly improved performance in other prostate segmentation tasks, even when trained with only 5% of the original dataset.

Intra-domain task-adaptive transfer learning to determine acute ischemic stroke onset time

Treatment of acute ischemic strokes (AIS) is largely contingent upon the time since stroke onset (TSS). However, TSS may not be readily available in up to 25% of patients with unwitnessed AIS. Current clinical guidelines for patients with unknown TSS recommend the use of MRI to determine eligibility for thrombolysis, but radiology assessments have high inter-reader variability. In this work, we present deep learning models that leverage MRI diffusion series to classify TSS based on clinically validated thresholds. We propose an intra-domain task-adaptive transfer learning method, which involves training a model on an easier clinical task (stroke detection) and then refining the model with different binary thresholds of TSS. We apply this approach to both 2D and 3D CNN architectures with our top model achieving an ROC-AUC value of 0.74, with a sensitivity of 0.70 and a specificity of 0.81 for classifying TSS < 4.5 h. Our pretrained models achieve better classification metrics than the models trained from scratch, and these metrics exceed those of previously published models applied to our dataset. Furthermore, our pipeline accommodates a more inclusive patient cohort than previous work, as we did not exclude imaging studies based on clinical, demographic, or image processing criteria. When applied to this broad spectrum of patients, our deep learning model achieves an overall accuracy of 75.78% when classifying TSS < 4.5 h, carrying potential therapeutic implications for patients with unknown TSS.

HCET: Hierarchical Clinical Embedding With Topic Modeling on Electronic Health Records for Predicting Future Depression

Recent developments in machine learning algorithms have enabled models to exhibit impressive performance in healthcare tasks using electronic health record (EHR) data. However, the heterogeneous nature and sparsity of EHR data remains challenging. In this work, we present a model that utilizes heterogeneous data and addresses sparsity by representing diagnoses, procedures, and medication codes with temporal Hierarchical Clinical Embeddings combined with Topic modeling (HCET) on clinical notes. HCET aggregates various categories of EHR data and learns inherent structure based on hospital visits for an individual patient. We demonstrate the potential of the approach in the task of predicting depression at various time points prior to a clinical diagnosis. We found that HCET outperformed all baseline methods with a highest improvement of 0.07 in precision-recall area under the curve (PRAUC). Furthermore, applying attention weights across EHR data modalities significantly improved the performance as well as the model's interpretability by revealing the relative weight for each data modality. Our results demonstrate the model's ability to utilize heterogeneous EHR information to predict depression, which may have future implications for screening and early detection.

Social media analytics of overactive bladder posts: what do patients know and want to know?

OBJECTIVE

To assess women's knowledge, patient experience, and treatment decision making regarding overactive bladder (OAB) using digital ethnography.

METHODS

Online posts were identified using a data mining service. Two hundred randomized posts were reviewed and coded using grounded theory. We then applied a latent Dirichlet allocation (LDA) probabilistic topic modeling process to review the entire collection of identified posts.

RESULTS

A total of 2618 posts by 1867 unique users from 203 different websites were identified. Our analysis yielded six themes: the impact of OAB on quality of life, patient-physician interactions, online engagement, symptom management, patient knowledge acquisition, and alternative therapies.

CONCLUSION

Overall, online communities are a source of support for women to self-manage the OAB symptom complex and help overcome treatment pathway challenges. Digital ethnography provides insight into patient knowledge and barriers to patient-centered care, which are important to improve patient outreach. Additionally, we identify similar findings to prior work, indicating the reliability of studying social media.

A multi-resolution model for histopathology image classification and localization with multiple instance learning

Large numbers of histopathological images have been digitized into high resolution whole slide images, opening opportunities in developing computational image analysis tools to reduce pathologists' workload and potentially improve inter- and intra-observer agreement. Most previous work on whole slide image analysis has focused on classification or segmentation of small pre-selected regions-of-interest, which requires fine-grained annotation and is non-trivial to extend for large-scale whole slide analysis. In this paper, we proposed a multi-resolution multiple instance learning model that leverages saliency maps to detect suspicious regions for fine-grained grade prediction. Instead of relying on expensive region- or pixel-level annotations, our model can be trained end-to-end with only slide-level labels. The model is developed on a large-scale prostate biopsy dataset containing 20,229 slides from 830 patients. The model achieved 92.7% accuracy, 81.8% Cohen's Kappa for benign, low grade (i.e. Grade group 1) and high grade (i.e. Grade group ≥ 2) prediction, an area under the receiver operating characteristic curve (AUROC) of 98.2% and an average precision (AP) of 97.4% for differentiating malignant and benign slides. The model obtained an AUROC of 99.4% and an AP of 99.8% for cancer detection on an external dataset.

Federated learning improves site performance in multicenter deep learning without data sharing

Objective

To demonstrate enabling multi-institutional training without centralizing or sharing the underlying physical data via federated learning (FL).

Materials and Methods

Deep learning models were trained at each participating institution using local clinical data, and an additional model was trained using FL across all of the institutions.

Results

We found that the FL model exhibited superior performance and generalizability to the models trained at single institutions, with an overall performance level that was significantly better than that of any of the institutional models alone when evaluated on held-out test sets from each institution and an outside challenge dataset.

Discussion

The power of FL was successfully demonstrated across 3 academic institutions while avoiding the privacy risk associated with the transfer and pooling of patient data.

Conclusion

Federated learning is an effective methodology that merits further study to enable accelerated development of models across institutions, enabling greater generalizability in clinical use.

Classifying Breast Histopathology Images with a Ductal Instance-Oriented Pipeline

In this study, we propose the Ductal Instance-Oriented Pipeline (DIOP) that contains a duct-level instance segmentation model, a tissue-level semantic segmentation model, and three-levels of features for diagnostic classification. Based on recent advancements in instance segmentation and the Mask RCNN model, our duct-level segmenter tries to identify each ductal individual inside a microscopic image; then, it extracts tissue-level information from the identified ductal instances. Leveraging three levels of information obtained from these ductal instances and also the histopathology image, the proposed DIOP outperforms previous approaches (both feature-based and CNN-based) in all diagnostic tasks; for the four-way classification task, the DIOP achieves comparable performance to general pathologists in this unique dataset. The proposed DIOP only takes a few seconds to run in the inference time, which could be used interactively on most modern computers. More clinical explorations are needed to study the robustness and generalizability of this system in the future.

Semi-automated PIRADS scoring via mpMRI analysis

Purpose: Prostate cancer (PCa) is the most common solid organ cancer and second leading cause of death in men. Multiparametric magnetic resonance imaging (mpMRI) enables detection of the most aggressive, clinically significant PCa (csPCa) tumors that require further treatment. A suspicious region of interest (ROI) detected on mpMRI is now assigned a Prostate Imaging-Reporting and Data System (PIRADS) score to standardize interpretation of mpMRI for PCa detection. However, there is significant inter-reader variability among radiologists in PIRADS score assignment and a minimal input semi-automated artificial intelligence (AI) system is proposed to harmonize PIRADS scores with mpMRI data. Approach: The proposed deep learning model (the seed point model) uses a simulated single-click seed point as input to annotate the lesion on mpMRI. This approach is in contrast to typical medical AI-based approaches that require annotation of the complete lesion. The mpMRI data from 617 patients used in this study were prospectively collected at a major tertiary U.S. medical center. The model was trained and validated to classify whether an mpMRI image had a lesion with a PIRADS score greater than or equal to PIRADS 4. Results: The model yielded an average receiver-operator characteristic (ROC) area under the curve (ROC-AUC) of 0.704 over a 10-fold cross-validation, which is significantly higher than the previously published benchmark. Conclusions: The proposed model could aid in PIRADS scoring of mpMRI, providing second reads to promote quality as well as offering expertise in environments that lack a radiologist with training in prostate mpMRI interpretation. The model could help identify tumors with a higher PIRADS for better clinical management and treatment of PCa patients at an early stage.

Integrating remote monitoring into heart failure patients' care regimen: A pilot study

Background

Around 50% of hospital readmissions due to heart failure are preventable, with lack of adherence to prescribed self-care as a driving factor. Remote tracking and reminders issued by mobile health devices could help to promote self-care, which could potentially reduce these readmissions.

Objective

We sought to investigate two factors: (1) feasibility of enrolling heart failure patients in a remote monitoring regimen that uses wireless sensors and patient-reported outcome measures; and (2) their adherence to using the study devices and completing patient-reported outcome measures.

Methods

Twenty heart failure patients participated in piloting a remote monitoring regimen. Data collection included: (1) physical activity using wrist-worn activity trackers; (2) body weight using bathroom scales; (3) medication adherence using smart pill bottles; and (4) patient -reported outcomes using patient-reported outcome measures.

Results

We evaluated 150 hospitalized heart failure patients and enrolled 20 individuals. Two factors contributed to 50% (65/130) being excluded from the study: smartphone ownership and patient discharge. Over the course of the study, 60.0% of the subjects wore the activity tracker for at least 70% of the hours, and 45.0% used the scale for more than 70% of the days. The pill bottle was used less than 10% of the days by 55.0% of the subjects.

Conclusions

Our method of recruiting heart failure patients prior to hospital discharge may not be feasible as the enrollment rate was low. Once enrolled, the majority of subjects maintained a high adherence to wearing the activity tracker but low adherence to using the pill bottle and completing the follow-up surveys. Scale usage was fair, but it received positive reviews from most subjects. Given the observed usage and feedback, we suggest mobile health-driven interventions consider including an activity tracker and bathroom scale. We also recommend administering a shorter survey more regularly and through an easier interface.

A Machine Learning Approach to Classifying Self-Reported Health Status in a Cohort of Patients With Heart Disease Using Activity Tracker Data

Constructing statistical models using personal sensor data could allow for tracking health status over time, thereby enabling the possibility of early intervention. The goal of this study was to use machine learning algorithms to classify patient-reported outcomes (PROs) using activity tracker data in a cohort of patients with stable ischemic heart disease (SIHD). A population of 182 patients with SIHD were monitored over a period of 12 weeks. Each subject received a Fitbit Charge 2 device to record daily activity data, and each subject completed eight Patient-Reported Outcomes Measurement Information Systems short form at the end of each week as a self-assessment of their health status. Two models were built to classify PRO scores using activity tracker data. The first model treated each week independently, whereas the second used a hidden Markov model (HMM) to take advantage of correlations between successive weeks. Retrospective analysis compared the classification accuracy of the two models and the importance of each feature. In the independent model, a random forest classifier achieved a mean area under curve (AUC) of 0.76 for classifying the physical function PRO. The HMM model achieved significantly better AUCs for all PROs (p < 0.05) other than Fatigue and Sleep Disturbance, with a highest mean AUC of 0.79 for the physical function-short form 10a. Our study demonstrates the ability of activity tracker data to classify health status over time. These results suggest that patient outcomes can be monitored in real time using activity trackers.

An Electronic Form for Reporting Results of Targeted Prostate Biopsy: Urology Integrated Diagnostic Report (Uro-IDR)

OBJECTIVE

To detail the development of an electronic report that graphically conveys all relevant information from targeted prostate biopsy.

METHODS

The Urology Integrated Diagnostic Report (Uro-IDR) is based on a published framework (RadPath) which enables the compilation of diagnostic data from urology, radiology, and pathology. Each component of the Uro-IDR is generated by the contributing clinician, is assembled in one document, and provides correlation of the 3 inputs at a glance. Upon completion, the Uro-IDR is automatically linked to the electronic medical record as an interactive file and can also be downloaded for offline sharing as a PDF.

RESULTS

At our institution, 1638 individual Uro-IDRs were generated between June 2016 and April 2019. There were 5715 views of these documents via the EMR. The average turnaround time for the creation of an individual report decreased from nearly 8 days at the time of its launch to 2 days after 6 months of use. The average time for report generation was 22 seconds for the pathologist and 69 seconds for the radiologist. An instructive video is linked to this article.

CONCLUSION

The Uro-IDR has proven to be a feasible, efficient, clinically useful form to concisely transmit key information about targeted prostate biopsy to both clinicians and patients.

Estimating the Health-Related Quality of Life of Twitter Users Using Semantic Processing

Social media presents a rich opportunity to gather health information with limited intervention through the analysis of completely unstructured and unlabeled microposts. We sought to estimate the health-related quality of life (HRQOL) of Twitter users using automated semantic processing methods. We collected tweets from 878 Twitter users recruited through online solicitation and in-person contact with patients. All participants completed the four-item Centers for Disease Control Healthy Days Questionnaire at the time of enrollment and 30 days later to measure “ground truth” HRQOL. We used a combination of document frequency analysis, sentiment analysis, topic analysis, and concept mapping to extract features from tweets, which we then used to estimate dichotomized HRQOL (“high” vs. “low”) using logistic regression. Binary HRQOL status was estimated with moderate performance (AUC=0.64). This result indicates that free-range social media data only offers a window into HRQOL, but does not afford direct access to current health status.

A Machine Learning Approach for Classifying Ischemic Stroke Onset Time From Imaging

Current clinical practice relies on clinical history to determine the time since stroke (TSS) onset. Imaging-based determination of acute stroke onset time could provide critical information to clinicians in deciding stroke treatment options, such as thrombolysis. The patients with unknown or unwitnessed TSS are usually excluded from thrombolysis, even if their symptoms began within the therapeutic window. In this paper, we demonstrate a machine learning approach for TSS classification using routinely acquired imaging sequences. We develop imaging features from the magnetic resonance (MR) images and train machine learning models to classify the TSS. We also propose a deep-learning model to extract hidden representations for the MR perfusion-weighted images and demonstrate classification improvement by incorporating these additional deep features. The cross-validation results show that our best classifier achieved an area under the curve of 0.765, with a sensitivity of 0.788 and a negative predictive value of 0.609, outperforming existing methods. We show that the features generated by our deep-learning algorithm correlate with the MR imaging features, and validate the robustness of the model on imaging parameter variations (e.g., year of imaging). This paper advances magnetic resonance imaging analysis one-step-closer to an operational decision support tool for stroke treatment guidance.

Path R-CNN for Prostate Cancer Diagnosis and Gleason Grading of Histological Images

Prostate cancer is the most common and second most deadly form of cancer in men in the United States. The classification of prostate cancers based on Gleason grading using histological images is important in risk assessment and treatment planning for patients. Here, we demonstrate a new region-based convolutional neural network framework for multi-task prediction using an epithelial network head and a grading network head. Compared with a single-task model, our multi-task model can provide complementary contextual information, which contributes to better performance. Our model is achieved a state-of-the-art performance in epithelial cells detection and Gleason grading tasks simultaneously. Using fivefold cross-validation, our model is achieved an epithelial cells detection accuracy of 99.07% with an average area under the curve of 0.998. As for Gleason grading, our model is obtained a mean intersection over union of 79.56% and an overall pixel accuracy of 89.40%.

An EM-based semi-supervised deep learning approach for semantic segmentation of histopathological images from radical prostatectomies

Automated Gleason grading is an important preliminary step for quantitative histopathological feature extraction. Different from the traditional task of classifying small pre-selected homogeneous regions, semantic segmentation provides pixel-wise Gleason predictions across an entire slide. Deep learning-based segmentation models can automatically learn visual semantics from data, which alleviates the need for feature engineering. However, performance of deep learning models is limited by the scarcity of large-scale fully annotated datasets, which can be both expensive and time-consuming to create. One way to address this problem is to leverage external weakly labeled datasets to augment models trained on the limited data. In this paper, we developed an expectation maximization-based approach constrained by an approximated prior distribution in order to extract useful representations from a large number of weakly labeled images generated from low-magnification annotations. This method was utilized to improve the performance of a model trained on a limited fully annotated dataset. Our semi-supervised approach trained with 135 fully annotated and 1800 weakly annotated tiles achieved a mean Jaccard Index of 49.5% on an independent test set, which was 14% higher than the initial model trained only on the fully annotated dataset.

A Multi-scale U-Net for Semantic Segmentation of Histological Images from Radical Prostatectomies

Gleason grading of histological images is important in risk assessment and treatment planning for prostate cancer patients. Much research has been done in classifying small homogeneous cancer regions within histological images. However, semi-supervised methods published to date depend on pre-selected regions and cannot be easily extended to an image of heterogeneous tissue composition. In this paper, we propose a multi-scale U-Net model to classify images at the pixel-level using 224 histological image tiles from radical prostatectomies of 20 patients. Our model was evaluated by a patient-based 10-fold cross validation, and achieved a mean Jaccard index of 65.8% across 4 classes (stroma, Gleason 3, Gleason 4 and benign glands), and 75.5% for 3 classes (stroma, benign glands, prostate cancer), outperforming other methods.

Classifying Acute Ischemic Stroke Onset Time using Deep Imaging Features

Models have been developed to predict stroke outcomes (e.g., mortality) in attempt to provide better guidance for stroke treatment. However, there is little work in developing classification models for the problem of unknown time-since-stroke (TSS), which determines a patient's treatment eligibility based on a clinical defined cutoff time point (i.e., <4.5hrs). In this paper, we construct and compare machine learning methods to classify TSS<4.5hrs using magnetic resonance (MR) imaging features. We also propose a deep learning model to extract hidden representations from the MR perfusion-weighted images and demonstrate classification improvement by incorporating these additional imaging features. Finally, we discuss a strategy to visualize the learned features from the proposed deep learning model. The cross-validation results show that our best classifier achieved an area under the curve of 0.68, which improves significantly over current clinical methods (0.58), demonstrating the potential benefit of using advanced machine learning methods in TSS classification.

Consumers' Patient Portal Preferences and Health Literacy: A Survey Using Crowdsourcing

Background

eHealth apps have the potential to meet the information needs of patient populations and improve health literacy rates. However, little work has been done to document perceived usability of portals and health literacy of specific topics.

Objective

Our aim was to establish a baseline of lung cancer health literacy and perceived portal usability.

Methods

A survey based on previously validated instruments was used to assess a baseline of patient portal usability and health literacy within the domain of lung cancer. The survey was distributed via Amazon’s Mechanical Turk to 500 participants.

Results

Our results show differences in preferences and literacy by demographic cohorts, with a trend of chronically ill patients having a more positive reception of patient portals and a higher health literacy rate of lung cancer knowledge (P<.05).

Conclusions

This article provides a baseline of usability needs and health literacy that suggests that chronically ill patients have a greater preference for patient portals and higher level of health literacy within the domain of lung cancer.

Evaluating topic model interpretability from a primary care physician perspective

BACKGROUND AND OBJECTIVE

Probabilistic topic models provide an unsupervised method for analyzing unstructured text. These models discover semantically coherent combinations of words (topics) that could be integrated in a clinical automatic summarization system for primary care physicians performing chart review. However, the human interpretability of topics discovered from clinical reports is unknown. Our objective is to assess the coherence of topics and their ability to represent the contents of clinical reports from a primary care physician's point of view.

METHODS

Three latent Dirichlet allocation models (50 topics, 100 topics, and 150 topics) were fit to a large collection of clinical reports. Topics were manually evaluated by primary care physicians and graduate students. Wilcoxon Signed-Rank Tests for Paired Samples were used to evaluate differences between different topic models, while differences in performance between students and primary care physicians (PCPs) were tested using Mann-Whitney U tests for each of the tasks.

RESULTS

While the 150-topic model produced the best log likelihood, participants were most accurate at identifying words that did not belong in topics learned by the 100-topic model, suggesting that 100 topics provides better relative granularity of discovered semantic themes for the data set used in this study. Models were comparable in their ability to represent the contents of documents. Primary care physicians significantly outperformed students in both tasks.

CONCLUSION

This work establishes a baseline of interpretability for topic models trained with clinical reports, and provides insights on the appropriateness of using topic models for informatics applications. Our results indicate that PCPs find discovered topics more coherent and representative of clinical reports relative to students, warranting further research into their use for automatic summarization.

Abstract WP41: Predicting Acute Ischemic Stroke Tissue Fate Using Deep Learning on Source Perfusion MRI

Introduction: Perfusion MR and CT parameters are widely used during the assessment of acute ischemic strokes to evaluate the likely lesion growth. However, such perfusion parameters and derived ratios are sensitive to the choice of arterial input function or deconvolution method. Studies have shown that various deconvolution algorithms can lead to inconsistent values and also introduce distortions that influence measurement. In this work, we propose a deep convolutional neural network (CNN) to learn imaging features from source magnetic resonance perfusion images that are best predictive of tissue outcome.

Methods: We developed a deep CNN for voxel-by-voxel prediction. Perfusion-weighted images (PWIs), diffusion-weighted images (DWIs), and apparent diffusion coefficient (ADC) maps were obtained in five ischemic stroke patients (mean age: 69; mean NIHSS: 12) within 6.5 hours of stroke onset from UCLA stroke center database. The final infarct volumes were evaluated in the 5-7 days post-treatment FLAIR images. 53,000 training data were generated; each training data consisted of the raw voxel values obtained across the MR images. The CNN has three separate convolutional and pooling layers for each imaging channel; 128 combined features were learned in the final fully-connected layer, which were fitted to a logistic classifier for single voxel infarct prediction.

Results: The model achieved an area under the curve (AUC) of 0.941 +/- 0.002 (95% CI) in ten-fold cross validation, outperforming existing models with perfusion parameters. With a cut-off threshold of 0.5, the model achieved an accuracy of 0.865, a precision of 0.836, and a recall of 0.910. The model automatically learned spatio-temporal filters that detected temporal and spatial changes in the PWIs.

Conclusion: This pilot study showed that the deep CNN is capable of generating automatically learned features from source perfusion images that are predictive of tissue outcome, providing a potential alternative for stroke image analysis. Our future work includes experimenting on a larger dataset, and optimizing model parameters with pre-training and fine-tuning.

RadPath: A Web-based System for Integrating and Correlating Radiology and Pathology Findings During Cancer Diagnosis

RATIONALE AND OBJECTIVES

The current paradigm of cancer diagnosis involves uncoordinated communication of findings from radiology and pathology to downstream physicians. Discordance between these findings can require additional time from downstream users to resolve, or given incorrect resolution, may adversely impact treatment decisions. To mitigate this problem, we developed a web-based system, called RadPath, for correlating and integrating radiology and pathology reporting.

MATERIALS AND METHODS

RadPath includes interfaces to our institution's clinical information systems, which are used to retrieve reports, images, and test results that are structured into an interactive compendium for a diagnostic patient case. The system includes an editing interface for physicians, allowing for the inclusion of additional clinical data, as well as the ability to retrospectively correlate and contextualize imaging findings following pathology diagnosis.

RESULTS

During pilot deployment and testing over the course of 1 year, physicians at our institution have completed 60 RadPath cases, requiring an average of 128 seconds from a radiologist and an average of 93 seconds from a pathologist per case. Several technical and workflow challenges were encountered during development, including interfacing with diverse clinical information systems, automatically structuring report contents, and determining the appropriate physicians to create RadPath summaries. Reaction to RadPath has been positive, with users valuing the system's ability to consolidate diagnostic information.

CONCLUSIONS

With the increasing complexity of medicine and the movement toward team-based disease management, there is a need for improved clinical communication and information exchange. RadPath provides a platform for generating coherent and correlated diagnostic summaries in cancer diagnosis with minimal additional effort from physicians.

A data-driven approach for quality assessment of radiologic interpretations

Given the increasing emphasis on delivering high-quality, cost-efficient healthcare, improved methodologies are needed to measure the accuracy and utility of ordered diagnostic examinations in achieving the appropriate diagnosis. Here, we present a data-driven approach for performing automated quality assessment of radiologic interpretations using other clinical information (e.g., pathology) as a reference standard for individual radiologists, subspecialty sections, imaging modalities, and entire departments. Downstream diagnostic conclusions from the electronic medical record are utilized as "truth" to which upstream diagnoses generated by radiology are compared. The described system automatically extracts and compares patient medical data to characterize concordance between clinical sources. Initial results are presented in the context of breast imaging, matching 18 101 radiologic interpretations with 301 pathology diagnoses and achieving a precision and recall of 84% and 92%, respectively. The presented data-driven method highlights the challenges of integrating multiple data sources and the application of information extraction tools to facilitate healthcare quality improvement.

Incorporating advanced language models into the P300 speller using particle filtering

OBJECTIVE

The P300 speller is a common brain-computer interface (BCI) application designed to communicate language by detecting event related potentials in a subject's electroencephalogram signal. Information about the structure of natural language can be valuable for BCI communication, but attempts to use this information have thus far been limited to rudimentary n-gram models. While more sophisticated language models are prevalent in natural language processing literature, current BCI analysis methods based on dynamic programming cannot handle their complexity.

APPROACH

Sampling methods can overcome this complexity by estimating the posterior distribution without searching the entire state space of the model. In this study, we implement sequential importance resampling, a commonly used particle filtering (PF) algorithm, to integrate a probabilistic automaton language model.

MAIN RESULT

This method was first evaluated offline on a dataset of 15 healthy subjects, which showed significant increases in speed and accuracy when compared to standard classification methods as well as a recently published approach using a hidden Markov model (HMM). An online pilot study verified these results as the average speed and accuracy achieved using the PF method was significantly higher than that using the HMM method.

SIGNIFICANCE

These findings strongly support the integration of domain-specific knowledge into BCI classification to improve system performance.

Predicting discharge mortality after acute ischemic stroke using balanced data

Several models have been developed to predict stroke outcomes (e.g., stroke mortality, patient dependence, etc.) in recent decades. However, there is little discussion regarding the problem of between-class imbalance in stroke datasets, which leads to prediction bias and decreased performance. In this paper, we demonstrate the use of the Synthetic Minority Over-sampling Technique to overcome such problems. We also compare state of the art machine learning methods and construct a six-variable support vector machine (SVM) model to predict stroke mortality at discharge. Finally, we discuss how the identification of a reduced feature set allowed us to identify additional cases in our research database for validation testing. Our classifier achieved a c-statistic of 0.865 on the cross-validated dataset, demonstrating good classification performance using a reduced set of variables.

Imaging informatics for consumer health: towards a radiology patient portal

OBJECTIVE

With the increased routine use of advanced imaging in clinical diagnosis and treatment, it has become imperative to provide patients with a means to view and understand their imaging studies. We illustrate the feasibility of a patient portal that automatically structures and integrates radiology reports with corresponding imaging studies according to several information orientations tailored for the layperson.

METHODS

The imaging patient portal is composed of an image processing module for the creation of a timeline that illustrates the progression of disease, a natural language processing module to extract salient concepts from radiology reports (73% accuracy, F1 score of 0.67), and an interactive user interface navigable by an imaging findings list. The portal was developed as a Java-based web application and is demonstrated for patients with brain cancer.

RESULTS AND DISCUSSION

The system was exhibited at an international radiology conference to solicit feedback from a diverse group of healthcare professionals. There was wide support for educating patients about their imaging studies, and an appreciation for the informatics tools used to simplify images and reports for consumer interpretation. Primary concerns included the possibility of patients misunderstanding their results, as well as worries regarding accidental improper disclosure of medical information.

CONCLUSIONS

Radiologic imaging composes a significant amount of the evidence used to make diagnostic and treatment decisions, yet there are few tools for explaining this information to patients. The proposed radiology patient portal provides a framework for organizing radiologic results into several information orientations to support patient education.

Abstract TP382: A Bayesian Network for Reasoning on Acute Ischemic Stroke Intervention

INTRODUCTION: As endovascular treatment options for acute ischemic stroke evolve, it is important to study the conditions and decisions that lead to successful outcomes so that future interventions may be better tailored for each individual. Bayesian networks provide a method for studying the conditional dependencies between variables and have been applied in other areas of medicine. In this work, we present a Bayesian network of acute ischemic stroke for analyzing the dependencies between a patient’s clinical and imaging presentation, therapeutic interventions, and outcomes.

METHODS: 790 unique episodes of acute ischemic stroke from the last five years (2006-2012) were retrieved from our institution’s quality improvement repository. A subset of variables from each case was extracted and modeled in a Bayesian network. Variable selection and connectivity was guided by a review of current practice guidelines and the domain knowledge of clinical investigators. Conditional probabilities between variables were then calculated using expectation maximization.

RESULTS: The Bayesian network may be manipulated through a graphical user interface to investigate the likelihood of different clinical scenarios. For example, evidence regarding patient presentation may be set and then combinations of interventions may be applied to observe possible outcomes. Additionally, the model accommodates the integration of utility nodes to support exploration of a patient’s projected quality-adjusted life expectancy (QALE) as a function of their clinical state and variable treatment decisions.

CONCLUSION: The presented model supports a pertinent set of variables and provides an initial tool for clinicians and researchers to study how combinations of patient evidence and interventions affect outcomes in acute stroke. This information may be useful for suggesting subsequent investigations on methods for improving existing treatment protocols.

BabySTEPS: a sugar tracking electronic portal system for gestational diabetes

Gestational diabetes is a condition occurring in up to 18% [1] of pregnant women that results in an increase in blood glucose levels due to the body's inability to produce sufficient insulin given the additional needs of the baby, and/or hormonal changes that lower the body's sensitivity to insulin. If left untreated, the growing baby may become too large, increasing the risk of injury to the mother and baby during delivery. Controlling blood glucose can be a challenging task, especially for women with no previous experience and who may have unhealthy diets. An opportunity exists to further encourage compliance by providing patients electronic access to data generated during their pregnancy. Previous studies have shown the potential of portals for managing general diabetes [2], but no work has targeted glucose control in pregnant women. We present BabySTEPS (Sugar Tracking Electronic Portal System), a patient portal focused on engaging women with gestational diabetes that provides personalized feedback with the goal of reducing complications at birth and subsequent medical problems resulting from poor glucose control.

Unifying acute stroke treatment guidelines for a Bayesian belief network

With the large number of clinical practice guidelines available, there is an increasing need for a comprehensive unified model for acute ischemic stroke treatment to assist in clinical decision making. We present a unified treatment model derived through review of existing clinical practice guidelines, meta-analyses, and clinical trials. Using logic from the treatment model, a Bayesian belief network was defined and fitted to data from our institution's observational quality improvement database for acute stroke patients. The resulting network validates known relationships between variables, treatment decisions and outcomes, and enables the exploration of new correlative relationships not defined in current guidelines.

Hierarchical semantic structures for medical NLP

We present a framework for building a medical natural language processing (NLP) system capable of deep understanding of clinical text reports. The framework helps developers understand how various NLP-related efforts and knowledge sources can be integrated. The aspects considered include: 1) computational issues dealing with defining layers of intermediate semantic structures to reduce the dimensionality of the NLP problem; 2) algorithmic issues in which we survey the NLP literature and discuss state-of-the-art procedures used to map between various levels of the hierarchy; and 3) implementation issues to software developers with available resources. The objective of this poster is to educate readers to the various levels of semantic representation (e.g., word level concepts, ontological concepts, logical relations, logical frames, discourse structures, etc.). The poster presents an architecture for which diverse efforts and resources in medical NLP can be integrated in a principled way.

Integrating UIMA annotators in a web-based text processing framework

The Unstructured Information Management Architecture (UIMA) [1] framework is a growing platform for natural language processing (NLP) applications. However, such applications may be difficult for non-technical users deploy. This project presents a web-based framework that wraps UIMA-based annotator systems into a graphical user interface for researchers and clinicians, and a web service for developers. An annotator that extracts data elements from lung cancer radiology reports is presented to illustrate the use of the system. Annotation results from the web system can be exported to multiple formats for users to utilize in other aspects of their research and workflow. This project demonstrates the benefits of a lay-user interface for complex NLP applications. Efforts such as this can lead to increased interest and support for NLP work in the clinical domain.

Clinical Case-based Retrieval Using Latent Topic Analysis

Clinical reporting is often performed with minimal consideration for secondary computational analysis of concepts. This fact makes the comparison of patients challenging as records lack a representation in a space where their similarity may be judged quantitatively. We present a method by which the entirety of a patient's clinical records may be compared using latent topics. To capture topics at a clinically relevant level, patient reports are partitioned based on their type, allowing for a more granular characterization of topics. The resulting probabilistic patient topic representations are directly comparable to one another using distance measures. To navigate a collection of patient records we have developed a workstation that allows users to weight different report types and displays succinct summarizations of why two patients are deemed similar, tailoring and expediting searches. Results show the system is able to capture clinically significant topics that can be used for case-based retrieval.

A Neuro-Oncology Workstation for Structuring, Modeling, and Visualizing Patient Records

The patient medical record contains a wealth of information consisting of prior observations, interpretations, and interventions that need to be interpreted and applied towards decisions regarding current patient care. Given the time constraints and the large-often extraneous-amount of data available, clinicians are tasked with the challenge of performing a comprehensive review of how a disease progresses in individual patients. To facilitate this process, we demonstrate a neuro-oncology workstation that assists in structuring and visualizing medical data to promote an evidence-based approach for understanding a patient's record. The workstation consists of three components: 1) a structuring tool that incorporates natural language processing to assist with the extraction of problems, findings, and attributes for structuring observations, events, and inferences stated within medical reports; 2) a data modeling tool that provides a comprehensive and consistent representation of concepts for the disease-specific domain; and 3) a visual workbench for visualizing, navigating, and querying the structured data to enable retrieval of relevant portions of the patient record. We discuss this workstation in the context of reviewing cases of glioblastoma multiforme patients.

Informatics in radiology: A prototype Web-based reporting system for onsite-offsite clinician communication

The communication of imaging findings to a referring physician is an important role of the radiologist. However, communication between onsite and offsite physicians is a time-consuming process that can obstruct work flow and frequently involves no exchange of visual information, which is especially problematic given the importance of radiologic images for diagnosis and treatment. A prototype World Wide Web-based image documentation and reporting system was developed for use in supporting a "communication loop" that is based on the concept of a classic "wet-read" system. The proposed system represents an attempt to address many of the problems seen in current communication work flows by implementing a well-documented and easily accessible communication loop that is adaptable to different types of imaging study evaluation. Images are displayed in a native (DICOM) Digital Imaging and Communications in Medicine format with a Java applet, which allows accurate presentation along with use of various image manipulation tools. The Web-based infrastructure consists of a server that stores imaging studies and reports, with Web browsers that download and install necessary client software on demand. Application logic consists of a set of PHP (hypertext preprocessor) modules that are accessible with an application programming interface. The system may be adapted to any clinician-specialist communication loop, and, because it integrates radiologic standards with Web-based technologies, can more effectively communicate and document imaging data.

Field note phytovolatilization of oxygenated gasoline-impacted groundwater at an underground storage tank site via conifers

A stand of five conifers (Pinus sp.) bordering a gasoline service station was studied to estimate the methyl tert-butyl ether (MTBE) emission rate from gasoline-impacted groundwater. Groundwater was impacted with gasoline oxygenates MTBE and tert-butyl alcohol (TBA) at combined concentrations exceeding 200,000 microg/L. Condensate from trees was collected in sealed environmental chambers and analyzed. Concentrations of MTBE in condensate ranged from 0.51 to 460 microg/L; TBA ranged from 12 to 4100 microg/L (n=19). Transpirate concentrations were derived from MTBE air-liquid partitioning data exhibited in controls spiked with known concentrations of analyte. Tree emissions were estimated by multiplying average transpirate concentrations by transpiration rates derived from evapotranspiration data. Stand evapotranspiration was calculated using meteorological data from the California Irrigation Management Information System (CIMIS) applied in the Standardized Reference Evapotranspiration Equation.

The occupational health status of African-American women health care workers

Race, ethnicity, and gender are significant indicators of occupational status, general health status, and thus, occupational health status. Although African-American women constitute only 6.8% of the total U.S. labor force, they hold 20% of the jobs in the health care industry and are disproportionately represented in those jobs that have the highest levels of workplace exposure to hazards. As a result, they are therefore more likely to be at greater exposure and risk to the spectrum of occupational health problems. In order to gain insight into the effects of race and gender on the occupational health status of African-American women health care workers, this article uses three data sources that provide different but complementary sources of information on the demographic characteristics of workers, location of categories of occupations, working conditions of jobs, and other job and worker characteristics. Given the concentration of African-American women in health care positions where there exists a greater likelihood of being exposed to occupational hazards, it is therefore both logical and appropriate for primary care physicians, especially those engaged in office-based practices, to identify this target population for special services and to be more aware of the type of health issues with which these patients are more likely to present and to experience during their working lives. Health care providers have a responsibility to assess occupational factors related to a patient's health problems and to incorporate this information into their treatment protocols and into the design and explanation of each patient's care plan.