Posted in

Breakthrough: AI Classifier Uses Ultrasound to Detect PCa

•Posted inRadiology

Breakthrough: AI Classifier Uses Ultrasound to Detect PCa

Prostate cancer (PCa) diagnosis increasingly relies on advanced imaging. However, cost and accessibility remain significant barriers for many patients. Consequently, researchers developed a deep learning classifier to predict clinically significant PCa (csPCa) presence using quantitative features from 3D multiparametric ultrasound (mpUS). This development provides a cost-effective and widely available solution for mpUS prostate cancer detection. The classifier uses dynamic contrast-enhanced US and shear-wave elastography data for enhanced diagnostic capability.

Study Design and Methodology

Investigators conducted a multicenter prospective cohort study. They enrolled 327 patients with suspected PCa who underwent a transrectal 3D mpUS scan. The team carefully registered the acquisitions to 3D histology from subsequent radical prostatectomy, which served as the gold standard for csPCa presence. Specifically, voxels within lesions designated as International Society of Urological Pathology (ISUP) Grade Group $\ge$ 2 were classified as malignant; all others were benign. Researchers then trained a 3D deep learning classifier using these quantitative mpUS features. They performed internal evaluation on 250 patients and a later external evaluation on 77 patients, ensuring the model’s robust generalizability. They used the area under the receiver operating characteristic curve (ROC AUC) to evaluate performance per voxel.

High Accuracy for mpUS Prostate Cancer Detection

The classifier demonstrated impressive accuracy in both internal and external evaluations. First, the model achieved a high ROC AUC of 0.87 (95% CI: 0.85-0.89) on the internal evaluation set. This result utilized 7-fold cross-validation across the 327 patient dataset. Furthermore, the external evaluation cohort showed even better performance. The classifier obtained a ROC AUC of 0.88 (95% CI: 0.87-0.89) on this later, independent dataset. Therefore, the classifier successfully detects csPCa using quantitative 3D mpUS features. The study, validated on one of the largest mpUS prostate datasets globally, confirms the potential of ultrasound technology to offer an accurate, accessible alternative to expensive MRI for PCa diagnosis.

Frequently Asked Questions

Q1: What imaging techniques did the 3D mpUS include?

The 3D multiparametric ultrasound acquisitions included dynamic contrast-enhanced ultrasound (DCE-US) and shear-wave elastography (SWE). The combination provides both functional and anatomical information for the deep learning classifier.

Q2: How accurate was the deep learning classifier for csPCa?

The classifier achieved a high diagnostic performance, with an Area Under the Receiver Operating Characteristic Curve (ROC AUC) of 0.87 on the internal dataset, which improved slightly to 0.88 during the external evaluation.

Q3: Why is this technology important for global healthcare?

The research supports mpUS as a promising, cost-effective, and widely accessible tool for csPCa diagnosis. This is especially relevant in resource-constrained settings where multiparametric MRI (mpMRI) access is limited.

References

  1. Delberghe F et al. Development of a quantitative multiparametric ultrasound and deep learning classifier for the detection of prostate cancer. Eur Radiol. 2026 Jan 30. doi: 10.1007/s00330-026-12323-y. PMID: 41612079.
  2. Gorsi U, et al. Diagnostic Value of Multiparametric Transrectal Ultrasound in Patients with Suspected Carcinoma of Prostate: A Tertiary Care Centre Experience. Indian J Radiol Imaging. 22 Dec 2025.
  3. Ahmed HU, et al. The CADMUS trial: multiparametric ultrasound in prostate cancer diagnosis. VJOncology. 2021 Jun 1.