Generative Image AI: What It Is and Is Not in HOF
Quick facts
- Definition
- An artificial intelligence model that creates or alters images from an input, such as a photo and an instruction.
- Use in HOF
- Generating an illustrative projection of how an area could look with the planned volume.
- Limit
- The image is a simulation; it does not predict the result or guarantee the same appearance.
- Caution
- It involves facial data: it requires consent, a clear purpose and attention to the LGPD.
Generative image AI is a type of artificial intelligence that creates or modifies images from an input. In HOF, the input is the patient's photo with the areas to adjust indicated, and the output is an image that illustrates a possible result. The model does not "see" anatomy the way the professional does: it generates a plausible image from patterns it has learned.
What does it do well, and what does it not do?
It does visual communication well: the patient quickly understands what is being discussed. It does not measure, does not predict tissue reaction and knows nothing of product, technique or safety planning. It can sometimes get details wrong, a topic covered in hallucination in image AI. The feature that alters only part of an image is inpainting.
What changes for ethics and privacy?
A face photo is sensitive data. Before sending the photo to any AI service, the patient must know, and the professional must record the consent. The topic ties in with the LGPD (Brazil's General Data Protection Law) and informed consent. The technology can also be used for forgeries, such as the deepfake, which reinforces care in storing images.
Frequently asked questions
Does the AI projection show the real result of the procedure?
No. It is an illustrative simulation. The real result depends on anatomy, product, technique and each patient's response, and may differ from the image.