Defining the loss function is an important part of neural network design and critically determines the success of deep learning modeling. A significant shortcoming of the conventional loss functions is that they weight all regions in the input image volume equally, despite the fact that the system is known to be heterogeneous (i.e., some regions can achieve high prediction performance more easily than others). Here, we introduce a region-specific loss to lift the implicit assumption of homogeneous weighting for better learning. We divide the entire volume into multiple sub-regions, each with an individualized loss constructed for optimal local performance. Effectively, this scheme imposes higher weightings on the sub-regions that are more difficult to segment, and vice versa. Furthermore, the regional false positive and false negative errors are computed for each input image during a training step and the regional penalty is adjusted accordingly to enhance the overall accuracy of the prediction. Using different public and in-house medical image datasets, we demonstrate that the proposed regionally adaptive loss paradigm outperforms conventional methods in the multi-organ segmentations, without any modification to the neural network architecture or additional data preparation.

OBJECTIVE: Being able to predict a patient's life expectancy can help doctors and patients prioritize treatments and supportive care. For predicting life expectancy, physicians have been shown to outperform traditional models that use only a few predictor variables. It is possible that a machine learning model that uses many predictor variables and diverse data sources from the electronic medical record can improve on physicians' performance. For patients with metastatic cancer, we compared accuracy of life expectancy predictions by the treating physician, a machine learning model, and a traditional model. MATERIALS AND METHODS: A machine learning model was trained using 14 600 metastatic cancer patients' data to predict each patient's distribution of survival time. Data sources included note text, laboratory values, and vital signs. From 2015-2016, 899 patients receiving radiotherapy for metastatic cancer were enrolled in a study in which their radiation oncologist estimated life expectancy. Survival predictions were also made by the machine learning model and a traditional model using only performance status. Performance was assessed with area under the curve for 1-year survival and calibration plots. RESULTS: The radiotherapy study included 1190 treatment courses in 899 patients. A total of 879 treatment courses in 685 patients were included in this analysis. Median overall survival was 11.7 months. Physicians, machine learning model, and traditional model had area under the curve for 1-year survival of 0.72 (95% CI 0.63-0.81), 0.77 (0.73-0.81), and 0.68 (0.65-0.71), respectively. CONCLUSIONS: The machine learning model's predictions were more accurate than those of the treating physician or a traditional model.

Stereotactic body radiation therapy (SBRT) is a relatively novel treatment modality, with little post-treatment prognostic information reported. This study proposes a novel neural network based paradigm for accurate prediction of liver SBRT outcomes. We assembled a database of patients treated with liver SBRT at our institution. Together with a three-dimensional (3-D) dose delivery plans for each SBRT treatment, other variables such as patients' demographics, quantified abdominal anatomy, history of liver comorbidities, other liver-directed therapies, and liver function tests were collected. We developed a multi-path neural network with the convolutional path for 3-D dose plan analysis and fully connected path for other variables analysis, where the network was trained to predict post-SBRT survival and local cancer progression. To enhance the network robustness, it was initially pre-trained on a large database of computed tomography images. Following n-fold cross-validation, the network automatically identified patients that are likely to have longer survival or late cancer recurrence, i.e., patients with the positive predicted outcome (PPO) of SBRT, and vice versa, i.e., negative predicted outcome (NPO). The predicted results agreed with actual SBRT outcomes with 56% of PPO patients and 0% NPO patients with primary liver cancer survived more than two years after SBRT. Similarly, 82% of PPO patients and 0% of NPO patients with metastatic liver cancer survived two-year threshold. The obtained results were superior to the performance of support vector machine and random forest classifiers. Furthermore, the network was able to identify the critical-to-spare liver regions, and the critical clinical features associated with the highest risks of negative SBRT outcomes.

Developing physiologically-based pharmacokinetic (PBPK) models for chemicals can be resource-intensive, as neither chemical-specific parameters nor in vivo pharmacokinetic data are easily available for model construction. Previously developed, well-parameterized, and thoroughly-vetted models can be a great resource for the construction of models pertaining to new chemicals. A PBPK knowledgebase was compiled and developed from existing PBPK-related articles and used to develop new models. From 2,039 PBPK-related articles published between 1977 and 2013, 307 unique chemicals were identified for use as the basis of our knowledgebase. Keywords related to species, gender, developmental stages, and organs were analyzed from the articles within the PBPK knowledgebase. A correlation matrix of the 307 chemicals in the PBPK knowledgebase was calculated based on pharmacokinetic-relevant molecular descriptors. Chemicals in the PBPK knowledgebase were ranked based on their correlation toward ethylbenzene and gefitinib. Next, multiple chemicals were selected to represent exact matches, close analogues, or non-analogues of the target case study chemicals. Parameters, equations, or experimental data relevant to existing models for these chemicals and their analogues were used to construct new models, and model predictions were compared to observed values. This compiled knowledgebase provides a chemical structure-based approach for identifying PBPK models relevant to other chemical entities. Using suitable correlation metrics, we demonstrated that models of chemical analogues in the PBPK knowledgebase can guide the construction of PBPK models for other chemicals.

HTML has so far been the standard format for delivering information on the World Wide Web. However, automated information processing of these documents for data exchange and interoperability has been difficult. XML, a subset of SGML, has been proposed to be the next standard format, that allows user-defined tags for better describing nested document structures and associated semantics. Operations on structured documents, such as searching in nested document structures, require new functions that are not currently available on most systems today. We describe a general framework for manipulating structured documents based on document abstractions. An abstraction is an approximation of an actual document, while possessing useful properties for analyses of interest. The framework provides a wide design space for the tradeoff between cost and capability. This general framework can be applied to index design, document searching and categorization. We present this framework by focusing on the indexing and searching of structured documents in the XML domain, and prove their soundness. We also address the issue of rich data types in XML documents.