Integrating Learning from Examples into the Search for Diagnostic Policies

AAAI Conferences

This paper studies the problem of learning diagnostic policies from training examples. A diagnostic policy is a complete description of the decision-making actions of a diagnostician (i.e., tests followed by a diagnostic decision) for all possible combinations of test results. An optimal diagnostic policy is one that minimizes the expected total cost, which is the sum of measurement costs and misdiagnosis costs. In most diagnostic settings, there is a tradeoff between these two kinds of costs. This paper formalizes diagnostic decision making as a Markov Decision Process (MDP).


Integrating Learning from Examples into the Search for Diagnostic Policies

arXiv.org Artificial Intelligence

This paper studies the problem of learning diagnostic policies from training examples. A diagnostic policy is a complete description of the decision-making actions of a diagnostician (i.e., tests followed by a diagnostic decision) for all possible combinations of test results. An optimal diagnostic policy is one that minimizes the expected total cost, which is the sum of measurement costs and misdiagnosis costs. In most diagnostic settings, there is a tradeoff between these two kinds of costs. This paper formalizes diagnostic decision making as a Markov Decision Process (MDP). The paper introduces a new family of systematic search algorithms based on the AO* algorithm to solve this MDP. To make AO* efficient, the paper describes an admissible heuristic that enables AO* to prune large parts of the search space. The paper also introduces several greedy algorithms including some improvements over previously-published methods. The paper then addresses the question of learning diagnostic policies from examples. When the probabilities of diseases and test results are computed from training data, there is a great danger of overfitting. To reduce overfitting, regularizers are integrated into the search algorithms. Finally, the paper compares the proposed methods on five benchmark diagnostic data sets. The studies show that in most cases the systematic search methods produce better diagnostic policies than the greedy methods. In addition, the studies show that for training sets of realistic size, the systematic search algorithms are practical on todays desktop computers.


Third 'given wrong initial heart attack diagnosis'

BBC News

Almost a third of patients in England and Wales are being given the wrong initial diagnosis after a heart attack - with women having a far higher chance of being affected, a study suggests. University of Leeds research examined NHS data on about 600,000 heart attack cases over a period of nine years. Women were 50% more likely than men to have an initial diagnosis different from their final diagnosis, it said. NHS England said it was working to improve the diagnosis of heart attacks. The British Heart Foundation is urging people to be aware of the symptoms of a heart attack.


New AI tool increases accuracy of schizophrenia diagnosis

#artificialintelligence

A novel artificial intelligence (AI) enabled tool can help diagnose schizophrenia more accurately than other such systems, according to a study led by an Indian origin scientist. The tool called EMPaSchiz, developed by researchers at the University of Alberta in Canada, examined brain scans from patients who were diagnosed with schizophrenia and predicted the diagnosis with 87 per cent accuracy. The finding, published in the journal NPJ Schizophrenia, follows on a previous study in 2017 in which researchers at IBM and Alberta developed a tool capable of predicting schizophrenia with 74 per cent accuracy. "Schizophrenia is characterised by a constellation of symptoms that might co-occur in patients. Two individuals with the same diagnosis might still present different symptoms. This often leads to misdiagnosis," said Sunil Kalmady, a post-doctoral fellow at the University of Alberta.


Artificial Intelligence Better Than Doctors At Diagnosing Skin Cancer

#artificialintelligence

Skin cancer was found to be diagnosed more accurately by artificial intelligence than experienced dermatologists in a new international study. Researchers tested a form of machine learning known as a deep learning convolutional neural network (CNN) to reach this conclusion. The study titled "Artificial intelligence for melanoma diagnosis: How can we deliver on the promise?" was published in the cancer journal Annals of Oncology on May 28. Malignant melanoma accounts for 1 percent of all skin cancers but causes a majority of skin cancer-related deaths. The American Cancer Society estimates 9,320 people will die from melanoma in 2018 while 91,270 new cases will be diagnosed.