A Tractable Inference Algorithm for Diagnosing Multiple Diseases

By David E. Heckerman. 2013

180 citations recorded by the source corpus.

We examine a probabilistic model for the diagnosis of multiple diseases. In the model, diseases and findings are represented as binary variables. Also, diseases are marginally independent, features are conditionally independent given disease instances, and diseases interact to produce findings via a noisy OR-gate. An algorithm for computing the posterior probability of each disease, given a set of observed findings, called quickscore, is presented. The time complexity of the algorithm is 0 (nm - 2 m+ ), where n is the number of diseases, M + is the number of positive findings and m - is the number of negative findings. Although the time complexity of quickscore is exponential in the number of positive findings, the algorithm is useful in practice because the number of observed positive findings is usually far less than the number of diseases under consideration. Performance results for quickscore applied to a probabilistic version of Quick Medical Reference (QMR) are provided.

Subfields: Artificial Intelligence, Molecular Biology, Probabilistic & Bayesian Methods, Medical & Health AI

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Molecular Biology

In Probabilistic & Bayesian Methods

In Medical & Health AI

Citations in the corpus

Fully Convolutional Multi-Class Multiple Instance Learning (2014, methodology): “Most MIL problems are framed as max-margin learning (Andrews et al., 2002; Felzenszwalb et al., 2010), while other approaches use boosting (Ali & Saenko, 2014) or Noisy-OR models (Heckerman, 2013).”
Improving the accuracy of medical diagnosis with causal machine learning (2020, background): “In addition, they support ef ﬁ cient inference 67 and learning 68,69 , and allow for large BNs to be described by a number of parameters that grows linearly with the size of the network 68,70 .”
Doctor AI: Predicting Clinical Events via Recurrent Neural Networks (2015, background): “Although related problems such as disease progression modeling have been studied by many researchers over several decades, e.g. [16, 6, 27], most works do not achieve required accuracy and scalability, or need excessive expert domain knowledge, partly due to the lack of rich longitudinal EHR data…”
Constrained Convolutional Neural Networks for Weakly Supervised Segmentation (2015, background): “It is most frequently formulated as a maximum margin problem, although boosting [1,36] and NoisyOR models [15] have been explored as well.”
Evaluate the Malignancy of Pulmonary Nodules Using the 3-D Deep Leaky Noisy-OR Network (2017, background): “Heckerman [32] built a multifeatures and multidisease diagnosing system based on the noisy-OR gate.”
Probabilistic Diagnosis Using a Reformulation of the INTERNIST-1/QMR Knowledge Base (1991, methodology): “For small test cases, we can use the quickscore algorithm developed by Heckerman [21] to compute the exact posterior marginal probabilities of dis eases.”
On probabilistic inference by weighted model counting (2008, methodology): “Moreover, the paper showed that the performance of this general technique subsumed the performance of the specialized quickscore algorithm [13], which capitalizes on evidence in certain types of diagnostic networks.”
Confidence-Rated Multiple Instance Boosting for Object Detection (2014, background): “There, the probability of a bag being positive is formulated as a Noisy-OR [13] over the probabilities of its instances.”
Detector discovery in the wild: Joint multiple instance and representation learning (2014, background): “Multiple Instance Learning Since its inception, the MIL [11] problem has been attempted in several frameworks including Noisy-OR [18], boosting [2, 40] etc.”
Deep Multi-patch Aggregation Network for Image Style, Aesthetics, and Quality Estimation (2015, background): “According to the assumption, probability measures such as Noisy-OR [13, 25] and ISR [19] are designed for multiple instance aggregation, following which various cost functions were proposed to learn a function that maps the multiple-instances as a whole to the label.”
Multiple Instance Boosting for Object Detection (2005, methodology): “Diverse Density uses the Noisy OR generative model [6] to explain the bag labels.”