Stimulation Initiative for European Neural Applications
Esprit Project 9811
A New Method for Computer Assisted Prediciton of
Lymphnode-Metastasis in Gastric Cancer
Company background
No information available
The problem
Preoperative staging especially of lymphnode-metastasis (LN-M) in
patients with gastric cancer is very important for therapy decision.
Neural nets, that are tested at present for industrial assignment such
as robotics, language recognition, expert systems, etc., may offer new
possibilities to this problem. The intention of this neural network
application has been to evaluate if the accuracy of predicted frequency
of lymphnode-metastasis is improvable.
Neural network application
For organization and training of the neural net were about 4000
datasettings at disposal.
The simple neural net consisted of one layer of input neurons,
representing values of predicting variables, one layer of output
neurons, reflecting LN-M or no LN-M of one group of lymphnodes, and
weighted junctions between these layers.. The weights of the junctions
were learned by training the net with the given data.
Benefits
Neural networks are very tolerant related to inconsistence of data. The
weights will be given regarding to discrepancy in data, missing values
may be supplied. From this facts an opportunity for advancement of
preoperative prediciton of lymphnode-metastasis may afford.
Generalization
The application of neural networks in cancer treatment is capable of
simplifying therapy decisions and helps to avoid unnecessary surgery.
Contact person
Karin Droste -
Dept. of Surgery,
Technical University of Munich.
Phone: + 49 89 4140-2147
Fax: +49 89 4140-4738
![[ ]](../logo_small.gif){kind=logo}
SIENA