Truace
Subscribe

Automated Bergmann-Kliesch Score assessment in testicular biopsies using artificial intelligence-based whole-slide image analysis: a retrospective monocentric development and validation study

Background The Bergmann-Kliesch Score is a histopathological grading system used to evaluate spermatogenesis in testicular biopsies, quantifying the fraction of seminiferous tubules containing elongated spermatids. It serves as a key parameter in the clinical process of men with non-obstructive azoospermia (NOA) undergoing testicular sperm extraction (TESE) for subsequent intracytoplasmic sperm injection (ICSI). Manual Bergmann-Kliesch Score assessment is time-consuming and subject to inter- and intraobserver va…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1796· cite & verify
Automated Bergmann-Kliesch Score assessment in testicular biopsies using artificial intelligence-based whole-slide image analysis: a retrospective monocentric development and validation study
"Capillaritis-Acute hemorrhage - Milk allergy Case 107" by Pulmonary Pathology is licensed under CC BY-SA 2.0. To view a copy of this license, visit https://creativecommons.org/licenses/by-sa/2.0/.

In brief

Researchers at the University Medical Center Frankfurt built a two-model machine-learning pipeline to grade spermatogenesis in testicular biopsies. It analyzes digitized slides stained with haematoxylin-eosin and with OCT3/4 and produces a Bergmann-Kliesch Score automatically. The cohort was 74 retrospective patients who had testicular sperm extraction.

The work matters because manual scoring is slow and varies between and within observers, and the score guides care for men with non-obstructive azoospermia. The evidence is still early. It comes from one center and a small test set, and it has not yet been linked to fertility treatment outcomes or checked externally.

Main points

  1. The cohort was 74 retrospective patients who underwent testicular sperm extraction at one German university medical center.
  2. One U-Net model segmented seminiferous tubules on HE slides, with a multi-class Dice coefficient of 0.897 on the validation set.
  3. A second model detected elongated spermatids on OCT3/4 slides, trained on 14,226 manually annotated cells.
  4. On the 15-patient test set, Pearson r was 0.957 and mean absolute error was 0.93.
  5. Across all 74 patients, agreement was lower but still robust: Pearson r 0.851 and mean absolute error 1.52.

The gain

An AI pipeline automated Bergmann-Kliesch Score assessment from biopsy slide images and agreed strongly with pathologist reference scores on a held-out test set of 15 patients.

The rundown

The pipeline segments tubules in HE-stained slides with a U-Net using a ResNet-34 backbone, trained on 60 slides. A second U-Net with the same backbone was trained on 15 OCT3/4-stained slides to find elongated spermatids.

Rigid alignment and B-spline warping register the two stain types so spermatid detections can be projected onto the tubular segmentation. This lets the system estimate the fraction of tubules containing elongated spermatids, which is what the score measures.

On the held-out test set, Spearman ρ was 0.856, R2 was 0.916 and root mean squared error was 1.28. At the clinical cutoff of a score of 1 or higher, all test cases were classified correctly. Across the full cohort, root mean squared error rose to 2.20.

Sources

  1. Peer-reviewedBasic and Clinical Andrology2026-10-06

The debate