Ellipsoid prostate volume matches segmentation closely enough for PSA density

Design

Prospective single-centre method comparison

Population

328 men with suspected prostate cancer having 3 T multiparametric MRI

Primary outcome

Agreement of ellipsoid and manually segmented prostate volume

Effect

ICC 0.985; mean absolute difference 5.5 cc; PSA density agreement sensitivity 0.86 to 0.95, specificity 0.88 to 0.97

A prospective study in European Radiology (published 28 September) compared prostate volume from the PI-RADS ellipsoid formula (three diameters times 0.52) with manual segmentation of every axial T2-weighted slice in 328 men having 3 T multiparametric MRI for suspected prostate cancer.

Agreement was excellent (intraclass correlation 0.985), with a mean absolute difference of 5.5 cc. The ellipsoid underestimated volume in glands over 80 cc. Adjusting the 0.52 coefficient by gland size did not improve performance. At PSA density cut-offs of 0.10, 0.15 and 0.20, the ellipsoid volume classified patients the same way as segmentation with sensitivity 0.86 to 0.95 and specificity 0.88 to 0.97.

Some departments segment the gland manually or semi-automatically for every case, which takes time. These data support the ellipsoid for routine reporting, with extra care in very large glands, where underestimating volume overestimates PSA density and could tip a man towards biopsy.

• Use the PI-RADS ellipsoid formula for routine prostate volume on MRI.
• Measure the three diameters on the correct planes as PI-RADS specifies to keep agreement high.
• Take extra care in glands over 80 cc, where the ellipsoid underestimates volume and inflates PSA density.
• Consider segmentation when PSA density sits close to a decision threshold in a large gland.
• Report the volume method used so serial measurements are comparable.

Why it matters

It removes the case for routine manual segmentation, saving reporting time without changing biopsy decisions in most men.

Don't overread it

The comparison was against manual segmentation, not against biopsy outcomes, so it shows measurement agreement rather than better cancer detection.

The statistics, in plain English

An intraclass correlation of 0.985 means the two methods agree almost perfectly across the range. The sensitivities and specificities here describe how often the ellipsoid put a patient on the same side of a PSA density cut-off as segmentation, not how well PSA density detects cancer.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.