A core feature of anorexia nervosa is an over-estimation of body size. However, quantifying this over-estimation has been problematic as existing methodologies introduce a series of artefacts and inaccuracies in the stimuli used for judgements of body size. To overcome these problems, we have: (i) taken 3D scans of 15 women who have symptoms of anorexia (referred to henceforth as anorexia spectrum disorders, ANSD) and 15 healthy control women, (ii) used a 3D modelling package to build avatars from the scans, (iii) manipulated the body shapes of these avatars to reflect biometrically accurate, continuous changes in body mass index (BMI), (iv) used these personalized avatars as stimuli to allow the women to estimate their body size. The results show that women who are currently receiving treatment for ANSD show an over-estimation of body size which rapidly increases as their own BMI increases. By contrast, the women acting as healthy controls can accurately estimate their body size irrespective of their own BMI. This study demonstrates the viability of combining 3D scanning and CGI techniques to create personalised realistic avatars of individual patients to directly assess their body image perception.