From explanation to intervention: Interactive knowledge extraction from Convolutional Neural Networks used in radiology.

Deep Learning models such as Convolutional Neural Networks (CNNs) are very effective at extracting complex image features from medical X-rays. However, the limited interpretability of CNNs has hampered their deployment in medical settings as they failed to gain trust among clinicians. In this work, we propose an interactive framework to allow clinicians to ask what-if questions and intervene in the decisions of a CNN, with the aim of increasing trust in the system. The framework translates a layer of a trained CNN into a measurable and compact set of symbolic rules. Expert interactions with visualizations of the rules promote the use of clinically-relevant CNN kernels and attach meaning to the rules. The definition and relevance of the kernels are supported by radiomics analyses and permutation evaluations, respectively. CNN kernels that do not have a clinically-meaningful interpretation are removed without affecting model performance. By allowing clinicians to evaluate the impact of adding or removing kernels from the rule set, our approach produces an interpretable refinement of the data-driven CNN in alignment with medical best practice.

Sequence Classification Restricted Boltzmann Machines With Gated Units.

For the classification of sequential data, dynamic Bayesian networks and recurrent neural networks (RNNs) are the preferred models. While the former can explicitly model the temporal dependences between the variables, and the latter have the capability of learning representations. The recurrent temporal restricted Boltzmann machine (RTRBM) is a model that combines these two features. However, learning and inference in RTRBMs can be difficult because of the exponential nature of its gradient computations when maximizing log likelihoods. In this article, first, we address this intractability by optimizing a conditional rather than a joint probability distribution when performing sequence classification. This results in the "sequence classification restricted Boltzmann machine" (SCRBM). Second, we introduce gated SCRBMs (gSCRBMs), which use an information processing gate, as an integration of SCRBMs with long short-term memory (LSTM) models. In the experiments reported in this article, we evaluate the proposed models on optical character recognition, chunking, and multiresident activity recognition in smart homes. The experimental results show that gSCRBMs achieve the performance comparable to that of the state of the art in all three tasks. gSCRBMs require far fewer parameters in comparison with other recurrent networks with memory gates, in particular, LSTMs and gated recurrent units (GRUs).

A system dynamics approach to analyze laboratory test errors.

Although many researches have been carried out to analyze laboratory test errors during the last decade, it still lacks a systemic view of study, especially to trace errors during test process and evaluate potential interventions. This study implements system dynamics modeling into laboratory errors to trace the laboratory error flows and to simulate the system behaviors while changing internal variable values. The change of the variables may reflect a change in demand or a proposed intervention. A review of literature on laboratory test errors was given and provided as the main data source for the system dynamics model. Three "what if" scenarios were selected for testing the model. System behaviors were observed and compared under different scenarios over a period of time. The results suggest system dynamics modeling has potential effectiveness of helping to understand laboratory errors, observe model behaviours, and provide a risk-free simulation experiments for possible strategies.

Modelling clinical diagnostic errors: a system dynamics approach.

The diagnostic process involves a series of stages in the patient pathway. Any errors or misleading information from any stage could lead to errors in the final decision-making. System dynamics modeling maps the diagnostic process as a whole and seeks to provide a quantitative way of analyzing different errors at each stage as well as relevant key factors. This paper provides a framework based on system dynamics for modeling the tracing of errors inside of the system from where errors initially occur, the routes of errors inside of the system, and how errors are delivered out of the system. Also, a detailed illustration of the phase history and physical examinations is provided as an example to explain how relevant factors can be interpreted and how this affects errors according to the framework.

A causal loop approach to the study of diagnostic errors.

Missed, wrong or delayed diagnosis has a direct effect on patient safety. Diagnostic errors have been discussed at length, however it still lacks a systematic approach. This study proposed a more systematic way of studying diagnostic errors by using a causal loop diagram. A systematic review was used to find the key factors which may cause diagnostic errors and their interrelationships. A causal loop diagram, as a qualitative model at the first stage of system dynamics modeling, was produced to map all the factor and interrelationships. The diagram provides not only the direct and indirect factors affecting correct diagnosis, but also a clear view of how the change of one factor in the model triggers changes of other factors and then the change of the number of final diagnostic errors.

Learning and representing temporal knowledge in recurrent networks.

The effective integration of knowledge representation, reasoning, and learning in a robust computational model is one of the key challenges of computer science and artificial intelligence. In particular, temporal knowledge and models have been fundamental in describing the behavior of computational systems. However, knowledge acquisition of correct descriptions of a system's desired behavior is a complex task. In this paper, we present a novel neural-computation model capable of representing and learning temporal knowledge in recurrent networks. The model works in an integrated fashion. It enables the effective representation of temporal knowledge, the adaptation of temporal models given a set of desirable system properties, and effective learning from examples, which in turn can lead to temporal knowledge extraction from the corresponding trained networks. The model is sound from a theoretical standpoint, but it has also been tested on a case study in the area of model verification and adaptation. The results contained in this paper indicate that model verification and learning can be integrated within the neural computation paradigm, contributing to the development of predictive temporal knowledge-based systems and offering interpretable results that allow system researchers and engineers to improve their models and specifications. The model has been implemented and is available as part of a neural-symbolic computational toolkit.