On 2015 Dec 08, Michael Doube commented:

Structure model index does not measure rods and plates in trabecular bone

Structure model index (SMI) works by dilating a surface mesh a very small distance and measuring the change in surface area that occurs, then forming a ratio with volume and surface area squared. On convex surfaces, area increases during dilation, however, on concave surfaces area decreases during dilation. The negative change in surface area from the dilation of concave surfaces leads to a negative contribution to SMI, which is an aberration from its theoretical formulation. The authors recognise this and state that "the intersections between structure elements, e.g. between rods and plates, are not accounted for", but do not attempt to assess the degree of distortion such "intersections" would have on the final SMI value.

In recent work (Salmon PL, 2015) my colleagues and I quantified the amount of concave surface in trabecular bone and its contribution to SMI. We found that the assumption that the negative contribution from concave surfaces is negligible, is almost never correct in real bone samples, typically being ~20-70% of the surface. The negative contribution to SMI is substantial and renders the final SMI measurement very difficult to interpret at best, and misleading at worst. SMI's correlation with bone volume fraction has led to the false observation of a rod-like transition during bone loss, but this is little more than an artefact relating to a strong correlation between bone volume fraction and the fraction of the surface that is concave, which itself artificially depresses SMI.

We conclude that SMI should not be used for the measurement of rods and plates in trabecular bone, and that research which relies on SMI should be treated with caution.

On 2015 Dec 08, Michael Doube commented:

Structure model index does not measure rods and plates in trabecular bone

Structure model index (SMI) works by dilating a surface mesh a very small distance and measuring the change in surface area that occurs, then forming a ratio with volume and surface area squared. On convex surfaces, area increases during dilation, however, on concave surfaces area decreases during dilation. The negative change in surface area from the dilation of concave surfaces leads to a negative contribution to SMI, which is an aberration from its theoretical formulation. The authors recognise this and state that "the intersections between structure elements, e.g. between rods and plates, are not accounted for", but do not attempt to assess the degree of distortion such "intersections" would have on the final SMI value.

In recent work (Salmon PL, 2015) my colleagues and I quantified the amount of concave surface in trabecular bone and its contribution to SMI. We found that the assumption that the negative contribution from concave surfaces is negligible, is almost never correct in real bone samples, typically being ~20-70% of the surface. The negative contribution to SMI is substantial and renders the final SMI measurement very difficult to interpret at best, and misleading at worst. SMI's correlation with bone volume fraction has led to the false observation of a rod-like transition during bone loss, but this is little more than an artefact relating to a strong correlation between bone volume fraction and the fraction of the surface that is concave, which itself artificially depresses SMI.

We conclude that SMI should not be used for the measurement of rods and plates in trabecular bone, and that research which relies on SMI should be treated with caution.