Magnetic resonance imaging with k-means clustering objectively measures whole muscle volume compartments in sarcopenia/cancer cachexia

Calum Gray, Thomas J MacGillivray, Clare Eeley, Nathan A Stephens, Ian Beggs, Kenneth C Fearon, Carolyn A Greig

Research output: Contribution to journalArticlepeer-review

Abstract

Background & aims: Sarcopenia and cachexia are characterized by infiltration of non-contractile tissue within muscle which influences area and volume measurements. We applied a statistical clustering (k-means) technique to magnetic resonance (MR) images of the quadriceps of young and elderly healthy women and women with cancer to objectively separate the contractile and non-contractile tissue compartments.

Methods: MR scans of the thigh were obtained for 34 women (n = 16 young, (median) age 26 y: n = 9 older, age 80 y; n = 9 upper gastrointestinal cancer patients, age 65 y). Segmented regions of consecutive axial images were used to calculate cross-sectional area and (gross) volume. The k-means unsupervised algorithm was subsequently applied to the MR binary mask image array data with resultant volumes compared between groups.

Results: Older women and women with cancer had 37% and 48% less quadriceps muscle respectively than young women (p < 0.001). Application of k-means subtracted a significant 9%, 14% and 20% non-contractile tissue from the quadriceps of young, older and patient groups respectively (p < 0.001). There was a significant effect of group (i.e., cancer vs healthy) when controlling for age as a covariate (p = 0.003).

Conclusions: K-means objectively separates contractile and non-contractile tissue components. Women with upper GI cancer have significant fatty infiltration throughout whole muscle groups which is maintained when controlling for age.

Original languageEnglish
Pages (from-to)106-111
Number of pages6
JournalClinical Nutrition
Volume30
Issue number1
DOIs
Publication statusPublished - Feb 2011

Keywords

  • Magnetic resonance imaging
  • Image processing
  • Cluster analysis
  • Skeletal muscle
  • Ageing
  • Cachexia

Fingerprint

Dive into the research topics of 'Magnetic resonance imaging with k-means clustering objectively measures whole muscle volume compartments in sarcopenia/cancer cachexia'. Together they form a unique fingerprint.

Cite this