Radiology reports are a rich resource for advancing deep learning applications in medicine by leveraging the large volume of data continuously being updated, integrated, and shared. However, there are significant challenges as well, largely due to the ambiguity and subtlety of natural language. We propose a hybrid strategy that combines semantic-dictionary mapping and word2vec modeling for creating dense vector embeddings of free-text radiology reports. Our method leverages the benefits of both semantic-dictionary mapping as well as unsupervised learning. Using the vector representation, we automatically classify the radiology reports into three classes denoting confidence in the diagnosis of intracranial hemorrhage by the interpreting radiologist. We performed experiments with varying hyperparameter settings of the word embeddings and a range of different classifiers. Best performance achieved was a weighted precision of 88% and weighted recall of 90%. Our work offers the potential to leverage unstructured electronic health record data by allowing direct analysis of narrative clinical notes.

Quantitative imaging approaches compute features within images' regions of interest. Segmentation is rarely completely automatic, requiring time-consuming editing by experts. We propose a new paradigm, called "digital biopsy," that allows for the collection of intensity- and texture-based features from these regions at least 1 order of magnitude faster than the current manual or semiautomated methods. A radiologist reviewed automated segmentations of lung nodules from 100 preoperative volume computed tomography scans of patients with non-small cell lung cancer, and manually adjusted the nodule boundaries in each section, to be used as a reference standard, requiring up to 45 minutes per nodule. We also asked a different expert to generate a digital biopsy for each patient using a paintbrush tool to paint a contiguous region of each tumor over multiple cross-sections, a procedure that required an average of <3 minutes per nodule. We simulated additional digital biopsies using morphological procedures. Finally, we compared the features extracted from these digital biopsies with our reference standard using intraclass correlation coefficient (ICC) to characterize robustness. Comparing the reference standard segmentations to our digital biopsies, we found that 84/94 features had an ICC >0.7; comparing erosions and dilations, using a sphere of 1.5-mm radius, of our digital biopsies to the reference standard segmentations resulted in 41/94 and 53/94 features, respectively, with ICCs >0.7. We conclude that many intensity- and texture-based features remain consistent between the reference standard and our method while substantially reducing the amount of operator time required.

We propose a novel method, the adaptive local window, for improving level set segmentation technique. The window is estimated separately for each contour point, over iterations of the segmentation process, and for each individual object. Our method considers the object scale, the spatial texture, and the changes of the energy functional over iterations. Global and local statistics are considered by calculating several gray level co-occurrence matrices. We demonstrate the capabilities of the method in the domain of medical imaging for segmenting 233 images with liver lesions. To illustrate the strength of our method, those lesions were screened by either Computed Tomography or Magnetic Resonance Imaging. Moreover, we analyzed images using three different energy models. We compared our method to a global level set segmentation, to a local framework that uses predefined fixed-size square windows and to a local region-scalable fitting model. The results indicate that our proposed method outperforms the other methods in terms of agreement with the manual marking and dependence on contour initialization or the energy model used. In case of complex lesions, such as low contrast lesions, heterogeneous lesions, or lesions with a noisy background, our method shows significantly better segmentation with an improvement of 0.25 ± 0.13 in Dice similarity coefficient, compared with state of the art fixed-size local windows (Wilcoxon, p < 0.001).

Objective:

Stereotactic ablative radiotherapy (SABR) is being increasingly used as a non-invasive treatment for early-stage non-small cell lung cancer (NSCLC). A non-invasive method to estimate treatment outcomes in these patients would be valuable, especially since access to tissue specimens is often difficult in these cases.

Methods:

We developed a method to predict survival following SABR in NSCLC patients using analysis of quantitative image features on pre-treatment CT images. We developed a Cox Lasso model based on two-dimensional Riesz wavelet quantitative texture features on CT scans with the goal of separating patients based on survival.

Results:

The median log-rank p-value for 1000 cross-validations was 0.030. Our model was able to separate patients based upon predicted survival. When we added tumor size into the model, the p-value lost its significance, demonstrating that tumor size is not a key feature in the model but rather decreases significance likely due to the relatively small number of events in the dataset. Furthermore, running the model using Riesz features extracted either from the solid component of the tumor or from the ground glass opacity (GGO) component of the tumor maintained statistical significance. However, the p-value improved when combining features from the solid and the GGO components, demonstrating that there are important data that can be extracted from the entire tumor.

Conclusions:

The model predicting patient survival following SABR in NSCLC may be useful in future studies by enabling prediction of survival-based outcomes using radiomics features in CT images.

Advances in knowledge:

Quantitative image features from NSCLC nodules on CT images have been found to significantly separate patient populations based on overall survival (p = 0.04). In the long term, a non-invasive method to estimate treatment outcomes in patients undergoing SABR would be valuable, especially since access to tissue specimens is often difficult in these cases.

Background

Patient stratification to identify subtypes with different disease manifestations, severity, and expected survival time is a critical task in cancer diagnosis and treatment. While stratification approaches using various biomarkers (including high-throughput gene expression measurements) for patient-to-patient comparisons have been successful in elucidating previously unseen subtypes, there remains an untapped potential of incorporating various genotypic and phenotypic data to discover novel or improved groupings.

Methods

Here, we present HOCUS, a unified analytical framework for patient stratification that uses a community detection technique to extract subtypes out of sparse patient measurements. HOCUS constructs a patient-to-patient network from similarities in the data and iteratively groups and reconstructs the network into higher order clusters. We investigate the merits of using higher-order correlations to cluster samples of cancer patients in terms of their associations with survival outcomes.

Results

In an initial test of the method, the approach identifies cancer subtypes in mutation data of glioblastoma, ovarian, breast, prostate, and bladder cancers. In several cases, HOCUS provides an improvement over using the molecular features directly to compare samples. Application of HOCUS to glioblastoma images reveals a size and location classification of tumors that improves over human expert-based stratification.

Conclusions

Subtypes based on higher order features can reveal comparable or distinct groupings. The distinct solutions can provide biologically- and treatment-relevant solutions that are just as significant as solutions based on the original data.

A major bottleneck in developing clinically impactful machine learning models is a lack of labeled training data for model supervision. Thus, medical researchers increasingly turn to weaker, noisier sources of supervision, such as leveraging extractions from unstructured text reports to supervise image classification. A key challenge in weak supervision is combining sources of information that may differ in quality and have correlated errors. Recently, a statistical theory of weak supervision called data programming has shown promise in addressing this challenge. Data programming now underpins many deployed machine-learning systems in the technology industry, even for critical applications. We propose a new technique for applying data programming to the problem of cross-modal weak supervision in medicine, wherein weak labels derived from an auxiliary modality (e.g., text) are used to train models over a different target modality (e.g., images). We evaluate our approach on diverse clinical tasks via direct comparison to institution-scale, hand-labeled datasets. We find that our supervision technique increases model performance by up to 6 points area under the receiver operating characteristic curve (ROC-AUC) over baseline methods by improving both coverage and quality of the weak labels. Our approach yields models that on average perform within 1.75 points ROC-AUC of those supervised with physician-years of hand labeling and outperform those supervised with physician-months of hand labeling by 10.25 points ROC-AUC, while using only person-days of developer time and clinician work-a time saving of 96%. Our results suggest that modern weak supervision techniques such as data programming may enable more rapid development and deployment of clinically useful machine-learning models.

Purpose

The response to the COVID-19 pandemic has affected the management of patients with cancer. In this pooled retrospective analysis, we describe changes in management patterns for patients with cancer diagnosed with COVID-19 in two academic institutions in the San Francisco Bay Area.

Materials and methods

Adult and pediatric patients diagnosed with COVID-19 with a current or historical diagnosis of malignancy were identified from the electronic medical record at the University of California, San Francisco, and Stanford University. The proportion of patients undergoing active cancer management whose care was affected was quantified and analyzed for significant differences with regard to management type, treatment intent, and the time of COVID-19 diagnosis. The duration and characteristics of such changes were compared across subgroups.

Results

A total of 131 patients were included, of whom 55 were undergoing active cancer management. Of these, 35 of 55 (64%) had significant changes in management that consisted primarily of delays. An additional three patients not undergoing active cancer management experienced a delay in management after being diagnosed with COVID-19. The decision to change management was correlated with the time of COVID-19 diagnosis, with more delays identified in patients treated with palliative intent earlier in the course of the pandemic (March/April 2020) compared with later (May/June 2020) (OR, 4.2; 95% CI, 1.03 to 17.3; P = .0497). This difference was not seen among patients treated with curative intent during the same timeframe.

Conclusion

We found significant changes in the management of cancer patients with COVID-19 treated with curative and palliative intent that evolved over time. Future studies are needed to determine the impact of changes in management and treatment on cancer outcomes for patients with cancer and COVID-19.

Background

The diverse set of human brain structure and function analysis methods represents a difficult challenge for reconciling multiple views of neuroanatomical organization. While different views of organization are expected and valid, no widely adopted approach exists to harmonize different brain labeling protocols and terminologies. Our approach uses the natural organizing framework provided by anatomical structure to correlate terminologies commonly used in neuroimaging.

Description

The Foundational Model of Anatomy (FMA) Ontology provides a semantic framework for representing the anatomical entities and relationships that constitute the phenotypic organization of the human body. In this paper we describe recent enhancements to the neuroanatomical content of the FMA that models cytoarchitectural and morphological regions of the cerebral cortex, as well as white matter structure and connectivity. This modeling effort is driven by the need to correlate and reconcile the terms used in neuroanatomical labeling protocols. By providing an ontological framework that harmonizes multiple views of neuroanatomical organization, the FMA provides developers with reusable and computable knowledge for a range of biomedical applications.

Conclusions

A requirement for facilitating the integration of basic and clinical neuroscience data from diverse sources is a well-structured ontology that can incorporate, organize, and associate neuroanatomical data. We applied the ontological framework of the FMA to align the vocabularies used by several human brain atlases, and to encode emerging knowledge about structural connectivity in the brain. We highlighted several use cases of these extensions, including ontology reuse, neuroimaging data annotation, and organizing 3D brain models.

Objective:

Stereotactic ablative radiotherapy (SABR) is being increasingly used as a non-invasive treatment for early-stage non-small cell lung cancer (NSCLC). A non-invasive method to estimate treatment outcomes in these patients would be valuable, especially since access to tissue specimens is often difficult in these cases.

Methods:

We developed a method to predict survival following SABR in NSCLC patients using analysis of quantitative image features on pre-treatment CT images. We developed a Cox Lasso model based on two-dimensional Riesz wavelet quantitative texture features on CT scans with the goal of separating patients based on survival.

Results:

The median log-rank p-value for 1000 cross-validations was 0.030. Our model was able to separate patients based upon predicted survival. When we added tumor size into the model, the p-value lost its significance, demonstrating that tumor size is not a key feature in the model but rather decreases significance likely due to the relatively small number of events in the dataset. Furthermore, running the model using Riesz features extracted either from the solid component of the tumor or from the ground glass opacity (GGO) component of the tumor maintained statistical significance. However, the p-value improved when combining features from the solid and the GGO components, demonstrating that there are important data that can be extracted from the entire tumor.

Conclusions:

The model predicting patient survival following SABR in NSCLC may be useful in future studies by enabling prediction of survival-based outcomes using radiomics features in CT images.

Advances in knowledge:

Quantitative image features from NSCLC nodules on CT images have been found to significantly separate patient populations based on overall survival (p = 0.04). In the long term, a non-invasive method to estimate treatment outcomes in patients undergoing SABR would be valuable, especially since access to tissue specimens is often difficult in these cases.