MedVision-DiagnosisAI-Release

MedVision-DiagnosisAI

<!-- markdownlint-disable first-line-h1 --> <!-- markdownlint-disable html --> <!-- markdownlint-disable no-duplicate-header --> <div align="center"> <img src="figures/fig1.png" width="60%" alt="MedVision-DiagnosisAI" /> </div> <hr> <div align="center" style="line-height: 1;"> <a href="LICENSE" style="margin: 2px;"> <img alt="License" src="figures/fig2.png" style="display: inline-block; vertical-align: middle;"/> </a> </div>

1. Introduction

MedVision-DiagnosisAI represents a breakthrough in medical imaging analysis. This model has been specifically trained on diverse medical imaging datasets including X-rays, CT scans, MRIs, and pathology slides. The latest version incorporates advanced attention mechanisms and multi-scale feature extraction for improved diagnostic accuracy.

<p align="center"> <img width="80%" src="figures/fig3.png"> </p>

Compared to the previous version, this upgrade shows significant improvements in detecting subtle anomalies. For instance, in the RadBench 2025 evaluation, the model's sensitivity has increased from 82% to 94.5%. This advancement stems from enhanced multi-resolution feature fusion: the previous model processed images at 3 scales, whereas the new version uses 7 scales with dynamic attention.

Beyond improved detection rates, this version also features reduced false positive rates and better generalization across different imaging equipment manufacturers.

2. Evaluation Results

Comprehensive Benchmark Results

<div align="center"> </div>

Overall Performance Summary

MedVision-DiagnosisAI demonstrates superior performance across all evaluated medical imaging benchmark categories, with particularly notable results in early cancer detection and fracture identification tasks.

3. Clinical Integration & API Platform

We offer a HIPAA-compliant API for clinical integration. Please check our official medical portal for more details on deployment options.

4. How to Run Locally

Please refer to our clinical deployment guide for information about running MedVision-DiagnosisAI in your healthcare environment.

Key requirements for deployment:

1. DICOM image preprocessing pipeline is required.
2. GPU with minimum 16GB VRAM recommended for real-time inference.

The model architecture of MedVision-DiagnosisAI is based on a Vision Transformer backbone with medical-specific preprocessing layers.

Inference Configuration

We recommend using the following configuration for clinical deployment:

confidence_threshold = 0.85
max_resolution = 2048
enable_attention_maps = True

Input Format Requirements

For medical image analysis, please follow the DICOM standard input format. The model accepts:

• DICOM files directly
• PNG/JPEG with proper DICOM metadata headers
• NIfTI format for volumetric data (CT/MRI)

from medvision import DiagnosisAI

model = DiagnosisAI.from_pretrained("medvision/diagnosis-ai")
result = model.analyze(dicom_image, return_attention_map=True)

5. License

This model is licensed under the Apache 2.0 License. Clinical use requires additional validation and regulatory approval in your jurisdiction.

6. Contact

For clinical partnerships and research collaborations, please contact us at [email protected].