Table 1 Processing time for muscle cross-section analysis.

From: Impact of functional electrical stimulation on nerve-damaged muscles by quantifying fat infiltration using deep learning

Method

Processing time per image (minutes)

Processing time for 326 images (days)

Turnaround time per patient (days)

[a] Qualitative grading

0.5–5

0.2–1.2

3–5

[b] Manual annotation

60–480

13–109

 > 7

[c] k-means

6.5

 < 1.5

 < 2

[d] IMFSegNet

5

 < 1.2

 < 2

  1. The time per image corresponds to the processing time for analyzing one muscle cross-section. The processing time refers to the extrapolated processing time for 326 muscle cross-sections. Note that accounting for eight workhours per day increases the processing time for qualitative grading and manual annotation by a factor three, respectively, to 0.5–3.6 working days and 42–327 working days. The turnaround time refers to the calculated time from taking a muscle biopsy to the final clinical report, with taking in to account the breaking time required by a human and it is given for one patient. For qualitative grading and manual annotation processing times may vary due to the individual time of the pathologist and, more importantly, the total amount of fat in the muscle cross-sections. The k-means clustering and IMFSegNet were performed on an average consumer workstation (Intel Core i7-8700 K, 32 GB RAM).
Back to article page