|
การพยากรณ์น้ำหนักผู้ป่วยติดเตียงผ่านอุปกรณ์ไคเนคโดยใช้กระบวนการปรับปรุงภาพด้วยตัวกรองแบบมัธยฐานร่วมกับโครงข่ายประสาทเทียม |
|---|---|
| รหัสดีโอไอ | |
| Creator | อภิศักดิ์ พรหมฝาย |
| Title | การพยากรณ์น้ำหนักผู้ป่วยติดเตียงผ่านอุปกรณ์ไคเนคโดยใช้กระบวนการปรับปรุงภาพด้วยตัวกรองแบบมัธยฐานร่วมกับโครงข่ายประสาทเทียม |
| Contributor | วรพล มะโนสร้อย ดุษฎี บุญธรรม |
| Publisher | King Mongkut’s University of Technology Thonburi |
| Publication Year | 2565 |
| Journal Title | KMUTT Research and Development Journal |
| Journal Vol. | 45 |
| Journal No. | 3 |
| Page no. | 341-352 |
| Keyword | Kinect, Neural Network, Image Processing |
| URL Website | https://journal.kmutt.ac.th/ |
| Website title | เว็บไซต์วารสารวิจัยและพัฒนา มจธ. |
| ISSN | 2697-5521 |
| Abstract | Drug dosage for treatment of bedridden patients need to calculated based on patient body weight. Currently, the methods for calculating such a weight exhibit many limitations, including the need to move a patient to weigh, which may worsen the symptoms of a patient. Weighing bed, which yields high accuracy, is nevertheless very expensive, making the treatment cost higher. Prediction of the weight from the patient size is also normally not accurate. The present research therefore proposed a method for predicting the weight of a bedridden patient by Kinect, with image enhancement using median filter and artificial neural network. One hundred weight data of normal people, which were used instead of those belonging to real patients, were divided into 3 groups for training, testing, and validation at 70%, 15%, and 15%, respectively. Eighteen neural network architectures were designed. The results showed that the suitable artificial neural network that could be used to predict the body weight consisted of 3 nodes in the hidden layers; Lavenberg-Maqurdt algorithm should be used as the training function, while tan-sigmoid transfer function should be used as the activation function; R and RMSE values were noted to be 0.93393 and 5.92, respectively. |
