Neural Bone Density Estimation From Plain X-Ray Images

The patent creates a virtual copy of the complex DEXA measurement system by training a neural network on DEXA measurement data. The trained model then processes simple plain X-ray images to predict bone density, effectively copying the measurement capability without requiring the expensive physical equipment. This allows bone density estimation using readily available plain X-ray images instead of specialized DEXA apparatuses.

The patent replaces expensive, complex DEXA equipment with a computationally-based solution that processes inexpensive plain X-ray images. The neural network model, once trained, can be deployed on standard computing hardware, eliminating the need for costly specialized measurement devices. This makes bone density measurement accessible through routine, low-cost X-ray imaging.

The patent transforms the relationship between plain X-ray images and bone density measurement by applying a neural network model with learned parameters. The model learns optimal parameter transformations during training on DEXA data, enabling accurate bone density prediction from plain X-ray images that would otherwise be insufficient for precise measurement. This parameter transformation bridges the gap between simple imaging and accurate measurement.