S-4-03-04
History of Rehabilitation in Japan
Ken Akashi (Kawasaki Medical School, Kurashiki, Japan)
Since the Muromachi era of 15th century when oriental medicine were permitted to practice only for blind people by Shogun. However, regarding the rehabilitation of physically disabled, we have to come back to Kenji Takagi, Professor of Orthopedic Surgery, Tokyo University, who established a crippled children's hospital at Tokyo in 1943, and then all over Japan. He also made the welfare law of disabled. After the World War II, GHQ of United States of America sent Drs. Koike, Mizuno and Hieda for studying rehabilitation in U.S.A. They made a big flow of rehabilitation medicine in Japan. Many thanks to Dr. Rusk. Many Japanese physicians and surgeons were trained under him. So, you will see many members of Takagi family, those of NYU family, besides, those of Kottke family raised under Dr. Chino of Keio University. And naturally, they mixed, are growing and are developing a new Japanese Rehabilitation Medicine.
S-4-04-01
PREDICTION OF STROKE OUTCOME USING AN ARTIFICIAL NEURAL NETWORK; VALIDATION STUDY PERFORMED AFTER A 3-YEAR INTERVAL
Shigeru Sonoda (Department of Rehabilitation Medicine, Keio University School of Medicine, Tokyo, Japan)
Predictive validity of our previous neural network results was evaluated. In 1993, a neural network was created for 200 stroke patients admitted to our affiliated rehabilitation hospital within 4 months of the onset of stroke (creation group). One hundred patients had cerebral infarction and 100 cerebral bleeding. The obtained network equation was then applied to 60 patients who were admitted in 1996 or 1997 to validate the usefulness of the neural network method.
The neural network was calculated by the software "Skiltran," which is based on a back-propagation algorithm and also extracts explicit knowledge. The neural network has 3 layers. The input layer is composed of the items on the Stroke Impairment Assessment Set (SIAS) and items of the FIM (Functional Independence Measure) on admission. Output layer was the motor subscore of the FIM at discharge. In addition, there is a so-called hidden layer.
Correlation coefficient between the predicted value of the FIM at discharge and the actual value was 0.93 in the creation group and 0.88 in the validation group. The neural network method is, therefore, useful in predicting outcome of stroke.