UROLOGY

Detects prostate cancer with AI in Magnetic Resonance Imaging

From T2W, ADC and DWI sequences, quantusPR analyzes biparametric MRI studies to automatically detect lesions suspicious for malignancy.

ACCESS

DESCRIPTION

quantusPR is a test for prostate cancer prediction from biparametric MRIs.

quantusPR is a software medical device (MDSW) designed for use by radiologists and urologists to provide a prediction (expressed as a percentage) of prostate cancer risk from a biparametric MRI. The device provides medical information to support decision-making related to the identification and management of patients at risk of prostate cancer.

The device is designed as a clinical decision support tool, facilitating screening, case prioritization, and optimization of the radiologist’s workflow.

USE CASES

CU-PR-01: Mass Prostate Cancer Screening

In prostate cancer screening programs, quantusPR automatically analyzes MRIs and classifies risk, allowing the radiologist to prioritize the review of cases with a higher probability of malignancy and optimize workflow.

CU-PR-02: Follow-up of patients under active surveillance

In each follow-up MRI scan within an active surveillance program, quantusPR analyzes the image and provides an updated assessment of the detected lesions, which the specialist compares with the patient’s previous studies to assess disease stability or progression.

CLINICAL EVIDENCE

Tool for prostate cancer prediction from biparametric MRIs (quantusPR).

quantusPR is composed of several sequential preprocessing and classification modules, trained using public databases.

Generation of axial slices

From the three-dimensional volume obtained by prostate MRI, 2D axial slices are generated.

Orientation standardization

An algorithm was developed for automatic detection of bladder orientation in the images, achieving 100% accuracy in classifying the four possible orientations. The algorithm detects the original position of the bladder and automatically reorients the image so it appears on the left.

Segmentation of the prostate gland

The prostate gland is delineated using a Deep Learning (DL) segmentation model, applied to the T2W (T2-weighted) sequence. To train it, 568 MRIs were collected from various public databases, all with gland segmentation performed by experts. The model achieved a Jaccard index of 89.9%.

Image preparation for classification

Based on the bounding box derived from the previous segmentation, the following changes are made for each sequence:

T2W model: the image is cropped using the bounding box.

ADC and DWI model: the same cropping is applied and both sequences are combined into a single three-channel image (RGB format).

Classification

Two DL-based binary classification models are trained, responsible for distinguishing between healthy images and those with suspected cancer. The public dataset from the PI-CAI challenge was used, comprising 1,476 biparametric MRI studies, from which 16,389 2D images were generated (with a cancer prevalence of 7.1%). 70% of the images were used to train both classification models and the remaining 30% to evaluate them.

The evaluation results for the two models were:

Sequences	Sensitivity	Specificity	PPV *	NPV *
T2W	81.6%	71.0%	7.2%	99.3%
ADC+DWI	100.0%	73.9%	11.1%	100.0%

* PPV and NPV: Positive Predictive Value and Negative Predictive Value

SOLUTION INTEGRATION

Our products allow integration with client systems through the DICOM protocol and the HL7 FHIR interface
We ensure data privacy in SaaS installations by establishing IPsec VPNs with our clients.
All communications outside the client environment are secured with SSL TLS 1.3

LICENSING BY NUMBER OF TESTS

LICENSING BY ANALYSES PERFORMED

CUSTOM OFFERS

REQUEST
INFORMATION / QUOTE

DOWNLOAD ITEMS

CATALOG

ACQUISITION GUIDE

SAMPLE REPORT