AI assessment of sarcopenia in cancer
Sarcopenia, which is the loss of lean muscle mass, is a known marker of malnutrition, frailty, and functional decline in cancer patients. Up until recently, there had been no way to practically and directly assess sarcopenia in head and neck cancer patients. Instead, patient weight and body mass index (BMI) have been used as rough surrogates, though these measures have limitations and do not directly measure muscle.
We developed and externally validated a fully automated image-based deep learning platform for cervical vertebral muscle segmentation, skeletal muscle index calculation, and diagnosis of sarcopenia. The algorithm uses routine head and neck cancer CT scans and was developed and validated via a total of 899 patients undergoing primary radiation for HNSCC. The algorithm demonstrated high accuracy in the ability to segment skeletal muscle and measure muscle area. Using the algorithm, patients could be divided into those with and without sarcopenia.
We found that those with AI-diagnosed sarcopenia had lower survival and a higher risk of feeding tube dependency than those without sarcopenia. This algorithm could be used to help triage patients with sarcopenia for supportive interventions to help prevent feeding tube dependence and improve their cancer care. Further testing of this algorithm is ongoing, and we plan to implement it in a clinical trial in the future.
AIM Investigators
