## Challenge Description

Every year, millions of people suffer from fractures caused by osteoporosis, a silent disease that weakens bones over time. The clinical gold standard for diagnosing osteoporosis is to measure **bone mineral density (BMD) at the hip** using a DXA scan. However, DXA machines are expensive, not portable, and often out of reach for patients in remote or resource-limited settings.

Now imagine a different path: what if, instead of relying on the hip DXA, we could take a **simple X-ray of the hand or wrist**, using standard X-ray machines available in every hospital, and use AI to predict the patient’s hip BMD? Screening for osteoporosis could become cheaper, faster, and more accessible, helping doctors identify at-risk patients before fractures happen.

---

## Task

Your task is to build a model that analyses **hand or wrist X-ray images** and predicts the patient’s **hip BMD** as a continuous value.

* The predicted BMD is the **only required output for submission and scoring**.
* A hidden test set will evaluate your models with unseen X-rays.
* The leaderboard metric is **Mean Absolute Error (MAE)** on predicted BMD.

---

## Dataset

The dataset provides:

* **Hand or wrist X-ray images**
* **Ground-truth hip BMD values**
* Metadata such as **age** and **sex**

---

## Submission Format

Submit a CSV file named **submission.csv** with the following structure:

```
image	                                  Predicted BMD value
6.E.1_9095839_20090409_001.png	0.891300089
6.C.1_9100862_20090219_001.png	0.877347851
```

---

## Optional Analysis (Not for Submission)

From your predicted BMD values, you may compute **T-scores** and estimate the number of patients who are **Normal** vs **Low BMD** (optionally splitting Low BMD into osteopenia and osteoporosis).

Reference values for T-score calculation:

* Reference BMD: 0.86
* Standard Deviation: 0.12
* Formula: **T = (BMD\_pred − 0.86) / 0.12**

Label rule:

* “Normal” if T ≥ −1.0
* “Low BMD” otherwise

---

## Why It Matters

This challenge asks a simple but powerful question:
**Can AI transform an ordinary hand X-ray into a window on hip bone health?**

By participating, you may help advance medical AI toward tools that are **accessible, affordable, and trustworthy**, bringing earlier osteoporosis screening to the people who need it most.