|
|
|
|
|
|
 |
Applications of Neural NetworksNeural Networks in HealthNeural networks have found many uses in medicine. Neural networks are particularly useful in recognition, aiding in medical diagnosis. Specific examples of neural networks within the health industry follow. Breast Cancer Cell Analysis by Neural Networks Andrea Dawson et al. developed a neural network that recognise the "grade" of breast cancer cells under microscopy. Recognition of grades has traditionally been quite difficult with differences noted between observers. Initial comparisons showed the neural network in good agreement with human observer cancer classifications. Correct grade assignment as made between 50 and 90% of the time on cases not seen during training. Predicting Length of Stay with Neural Networks The Anderson Memorial Hospital in South Carolina developed a neural network to predict the severity of illnesses. The goal was to provide healthcare workers with information to improve patient care. The networks were trained to use variables about the diagnosis, body system involved, co-morbidities etc. to predict the length of stay of patients. The Augusta Mental Health Center also used a neural network
to help determine the length of stay of psychiatric inpatients. The network
uses a wide variety of inputs are used including the patient's demographics
and reason of admission. Diagnosing Heart Attacks with Neural Networks Doctors working at St. Joseph Mercy Hospital in Michigan created a neural network that recognised cases of acute myocardial infarction (heart attack) using serial blood levels of the marker creatinine kinase (CK). The neural network was trained on almost 200 examples from about 50 patients. Subsequent tests revealed the network to be in 95% agreement with specialists interpreting blood cardiac enzyme levels. Ordering Investigations with Neural Networks Dr Steven Berkov of Kaiser Hospital in California created a neural network that ordered lab tests as soon as a patient was registered. The network was trained on data from 250 patients past medical records. The network uses a patients demographics and symptoms to select from a range of almost 40 tests. According to Dr Berkov, the system has proven to be about 95% accurate. Diagnosing Giant Cell Arteritis with Neural Networks Several doctors have created a neural network to help diagnose cases of giant cell arteritis amongst other similar vasculitic diseases. The network was trained on the American College of Rheumatology database of cases using different variables such as age, headache, scalp tenderness etc. that were identified by the college. After training the network correctly classified about 95% of test cases with the disease, and 92% of those without the disease. Previous | Next | Page 1 2 3 4 Index | Applications | Health References: © 2008 Marcus bros
|
|
|
|
