- مبلغ: ۸۶,۰۰۰ تومان
- مبلغ: ۹۱,۰۰۰ تومان
Main aim of this study is to develop a software prototype tool for grading and diagnosing depression that will help general physicians for first hand applications. Identification of key symptoms responsible for depression is also another important issue considered in this study. It involves collection of data taken from patients through doctors. Due to several reasons, collection of data in Indian scenario is extremely difficult and thus this tool will be very handy and useful for general physicians working at remote locations. Also, it is possible to collect a data pool through this software model. An intelligent Neuro-Fuzzy model is developed for this purpose. Performance of the said model has been optimized through two approaches. In Approach 1, where a back-propagation algorithm has been considered and in Approach 2, Genetic Algorithm has been used. The model is trained with 78 data and validated with 10 data. Approach 2 superseded Approach 1 in terms of diagnostic accuracy. Therefore, it can be said that the soft computing-based diagnostic models could assist the doctors to make informed decisions. Data for training and validation for this purpose has been collected during 2004–2005 from a Government mental hospital in India.
4. Conclusions and future work
In this study, an attempt has been made to develop a NNFL system for diagnosing depression based on seven different symptoms. Performance of NNFL-based diagnosing tool is improved through GA-based optimization. Both rule base as well as data base has been optimized using GA. Unfortunately, GA took huge time to evolve the good solution. There are reported works which have brought down the time consumed while running a GA-based optimizer in machine automation. However, it is important to note here that diagnosis of depression consumes months to get diagnosed even by an expert psychiatrist and therefore speed is out of the scope of the paper. Still, to reduce the time the paper attempts to identify the redundant rules and analysis has been made to remove them by identifying the load of individual rules on the morbidity. Finally, performance of the optimized GAtuned NNFL system is compared with that of back-propagation based NNFL system for ten different cases. It was found that the performance of GA-tuned NNFL system is better compared to the other one. Moreover, it has been observed that GA-tuned NNFL system has outperformed to a NN-based system. Therefore, Approach 2 can be of great help to the doctors for diagnosing the patients suffering from depression.
Presently forward mapping has been between the symptoms and the disease. However, it will be much more interesting if the reverse mapping is made, where it will be possible to identify the impact of a symptom on the disease. The authors are presently working with this problem. It is also worth noting while reviewing literature we have come across few works on the said problems.