AI-Based Physical Function Assessment System Using In-Clinic Movements
Author Information
Author(s): Kouno Nobuji, Takahashi Satoshi, Takasawa Ken, Komatsu Masaaki, Ishiguro Naoaki, Takeda Katsuji, Matsuoka Ayumu, Fujimori Maiko, Yokoyama Kazuki, Yamamoto Shun, Honma Yoshitaka, Kato Ken, Obama Kazutaka, Hamamoto Ryuji, Giansanti Daniele
Primary Institution: National Cancer Center Research Institute
Hypothesis
Can inertial measurement unit data collected during in-clinic movements predict physical function assessment outcomes in cancer patients?
Conclusion
The study found that higher pelvic velocity variability is associated with shorter timed up and go test results, suggesting a potential for AI-based physical function assessments.
Supporting Evidence
- Pelvic velocity variability was identified as a key feature for predicting TUG test outcomes.
- A linear regression model using pelvic velocity variability achieved a mean absolute error of 1.322 seconds.
- Higher pelvic velocity variability correlated with shorter TUG test results.
- The study included 61 patients with cancer who performed movements in a consultation room.
- Correlation analyses were conducted to identify key features from the inertial measurement unit data.
- Machine learning models were developed to predict physical function assessments.
- The study highlights the potential for AI-based assessments in clinical settings.
- Limitations include a small sample size and focus on patients with performance status of 0 or 1.
Takeaway
This study shows that we can use data from movements in a doctor's office to help understand how well cancer patients can move, which can help doctors make better treatment decisions.
Methodology
Inertial measurement unit data were collected from 61 cancer patients performing movements in a consultation room, followed by correlation analyses and machine learning model development to predict TUG test outcomes.
Potential Biases
The study may have selection bias due to the exclusion of patients with poorer physical function.
Limitations
The study had a small sample size and focused only on patients with a performance status of 0 or 1, which may limit the generalizability of the findings.
Participant Demographics
The median age of participants was 71 years, with 48 males and 13 females, primarily diagnosed with esophageal cancer.
Statistical Information
P-Value
0.713
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website