Registration is open

ARISE-2025

Automated rheumatoid arthritis joint assessment in hackathon

Description

Rheumatoid arthritis (RA) is a chronic autoimmune disease characterized by inflammation, joint destruction, and extra-articular manifestations. Radiography is the standard imaging method for diagnosing and monitoring joint damage in RA. However, traditional methods for assessing radiographic changes, such as the Sharp method and its variants, are labor-intensive and subjective.

This hackathon aims to develop automated solutions for joint assessment in RA using computer vision methods.

Competitors are asked to create models for the automatic assessment of hand joint damage in RA. The task includes two key components:

1. Joint localization: Accurately identify the location of joints on radiographs.

2. Pathology assessment: Assess the extent of joint damage, with an emphasis on erosions and joint space narrowing (JSN, Joint Space Narrowing).

For whom

This hackathon is open to undergraduate, master's, speciality degree students and PhD students from technical universities in CIS and BRICS countries, Cuba and Vietnam.

Teams of 1–4 people

Age of participants — 18 years and over

Apply

Platform

Awards

Invitation to an internship at the MTUCI-AIRI SAIL laboratory and merch from the organizers.

Program

Until February 27

Registration

March 10

Start

March 13

Submitting solutions

March 14

Finalist Announcement

FAQ

How many people can be on the same team?

Teams of 1-4 people

Is personal attendance required?

The hackathon takes place online, but the awards ceremony will be held offline at 8A Aviamotornaya St., Moscow.

I have registered and created a team. What are the next steps?

The task release date is March 10. You have to check your equipment and be ready to begin.

Dataset

Participants will be provided with a clinical dataset of X-Ray images with annotations:

• Images: High-quality radiographic images of the hand joints.

• Annotations: Joint bounding box coordinates and joint space erosion and narrowing (JSN) scores.

Deliverables

Participants are expected to create a pipeline that:

1. Accurately localizes joints on X-ray images.

2. Correctly assesses the degree of erosion and joint space narrowing (JSN).

3. Demonstrates high generalizability and reliability across a variety of clinical data.

Metrics

The models performance will be evaluated by the following metrics:

1. Intersection over Union (IoU)

IoU measures the degree of agreement between the predicted bounding box and the ground truth.

• Area of Overlap: The total area of intersection of the predicted and ground truth boxes.

• Area of Union: The total area covered by the predicted and ground truth boxes.

The higher the IoU value, the more accurate the joint localization.

2. Accuracy

Accuracy is the proportion of correctly classified examples among all examples.

• TP (True Positives): True positive cases (e.g., correct classification of erosions).

• TN (True Negatives): True negative cases.

• FP (False Positives): False positive cases.

• FN (False Negatives): False negative cases.

Accuracy will be used to evaluate the accuracy of predictions of the degree of erosion and joint space narrowing (JSN).

3. Outcome metric: IoU × Accuracy

For the comprehensive evaluation of the models, the product of IoU and Accuracy will be used:

The final score ranges from 0 to 1, where higher values indicate better performance of the model, providing both accurate localization and correct classification of lesions.