DeepDx Prostate (RUO)

DeepDx Prostate automates the screening of prostate biopsies using deep learning to standardize grading workflows:

• Gland-Level Analysis: The algorithm utilizes advanced neural networks to perform segmentation at the gland level, automatically differentiating between benign tissue structures and adenocarcinoma.
• Instant Scoring: It instantly calculates the Gleason Score (e.g., 3+4) and ISUP Grade Group for the slide, providing a consistent, automated baseline for every case.
• Visual Verification: Results are visualized via gland-level segmentation masks that overlay the tissue.
• Gleason Pattern 3: Yellow overlay
• Gleason Pattern 4: Orange overlay
• Gleason Pattern 5: Red overlay

The algorithm calculates quantitative metrics for the specific Gleason Patterns present in the biopsy core. Rather than providing only a summary Grade Group, DeepDx Prostate computes the precise area occupied by each pattern relative to the total tumor area.

The following pattern-specific metrics are reported as percentages:

• % Pattern 3: (Pattern 3 Area / Total Tumor Area)
• % Pattern 4: (Pattern 4 Area / Total Tumor Area)
• % Pattern 5: (Pattern 5 Area / Total Tumor Area)

These continuous variables provide granular data on tumor composition, supporting precise clinical risk stratification and therapeutic decision-making for intermediate-risk cases.

The algorithm automatically measures the linear extent of disease across the biopsy core, including fragmented or curved samples. Linear metrics are derived from the segmentation mask and reported in millimeters.

The following length-based metrics are calculated as outputs for each slide:

• Length (Tumor) (mm): The cumulative linear length of all detected tumor foci.
• Length (Tissue) (mm): The cumulative length of all benign and malignant tissue fragments.
• % Tumor Involvement (Length): (Length (Tumor) / Length (Tissue)) × 100.

These standardized outputs replace manual measurements, providing consistent data for assessing linear tumor burden.

DeepDx Prostate calculates volumetric tumor burden metrics based on the two-dimensional surface area of the detected tissue, offering granular data essential for precision research. These metrics are derived from the total pixel count of the segmentation masks.

The following area-based metrics are calculated as outputs for each slide:

• Area (Tissue) (mm²): The total surface area of the biopsy core.
• Area (Tumor) (mm²): The total surface area classified as adenocarcinoma.
• % Tumor Involvement (Area): (Area (Tumor) / Area (Tissue)) × 100.