Medical Imaging · GPU Visualization · AI Integration

AI-Powered Medical Imaging Workstation

Native C++ workstation for DICOM CT, MPR, GPU volume rendering, image processing and MONAI / ONNX segmentation.

DICOM CT → MPR / 3D → Processing → AI Inference → MPR / 3D Results

C++17Qt 6VTK 9.3DICOMGPU Volume RenderingONNX RuntimeMONAI

Watch 75-sec Demo ▶View GitHub ↗

AI Medical Imaging Workstation showing completed aorta segmentation, MPR views, 3D visualization and quantitative results

Overview

DICOM to AI Results

Medical Visualization

MPR · W/L · CT presets · GPU 3D volume rendering

Processing & Analysis

Measure · crop · segment · surface · volume

Medical imaging workstation overview with synchronized MPR and 3D CT visualization

Demo

75-Second Workflow

CT → MPR / 3D → AI → results.

Native AI Integration

MONAI runs through ONNX Runtime on a background QThread; the viewer stays interactive.

ONNX RuntimeBackground QThread105 Classes96³ Sliding Window

System Architecture

Medical Imaging + AI Pipeline

01 · InputDICOM CTLoad + validate

02 · ImagingMPR + GPU 3DVisualize + process

03 · PreprocessMONAIRAS · 3 mm · 96³

04 · InferenceONNX Runtime105 classes · CPU

05 · MappingAorta · Class 7Map to DICOM

06 · OutputMPR + 3DOverlay · surface · metrics

Native C++17 • Qt 6 • VTK • DICOM • ONNX Runtime • MONAI

Core Capabilities

Visualization, Processing & AI

Aorta segmentation shown with dimmed CT context

DICOM + MPRLoad · validate · navigate

GPU Volume RenderingInteractive 3D CT

Image ProcessingMeasure · crop · segment

Native AIMONAI · ONNX Runtime

DICOM MappingAI → source geometry

MPR + 3D ResultsOverlay · surface · metrics

AI Pipeline

Background Inference

AI inference running while the viewer remains interactive

MONAI Preprocessing + ONNX

RAS orientation + 3 mm isotropic preprocessing
96³ sliding-window inference with Gaussian blending
105-class segmentation through ONNX Runtime
Multiclass argmax → Aorta (Class 7)
Background QThread execution
Original DICOM geometry restoration

Reference Result

Quantitative Output

Completed AI aorta segmentation with quantitative results

53.70 cm³Aorta volume

1,971AI-grid voxels

27,805Source voxels

105Output classes

4Patches

~10 sInference

~30 sTotal pipeline

CPU + QThreadBackend

Visualization

MPR + 3D Result Modes

Aorta segmentation overlay in synchronized MPR views — MPR AI overlay mapped to the original DICOM geometry.

Aorta-only 3D AI surface — Aorta-only 3D surface mode.

Aorta AI surface with full CT context — Aorta + full CT anatomical context.

Technology Stack

Core Technologies

Native C++ medical imaging workstation running ONNX inference in a background QThread

C++17Application + AI

Qt 6UI + QThread

VTK 9.3MPR + volume rendering

vtk-dicomDICOM geometry

OpenGLGPU visualization

ONNX RuntimeNative inference

MONAICT segmentation

CMake / VS2022Build toolchain

Demo CT data: HEART_CT sample DICOM dataset from Fovia F.A.S.T. SDK demo cases. Imaging data is not redistributed.

Medical Imaging & AI Engineering

Need a DICOM / AI Integration?

DICOM Viewers GPU Volume Rendering AI Integration C++ / Qt / VTK

Discuss a Project ✉ GitHub ↗