Introduction 1 2 3 5 6 3 4 2 4 2 7 13 4 5 14 18 19 20 Subjects and methods Participants 2 n n 21 1c 22 2 z z z MR imaging Image processing 23 k k 20 20 1 n n Fig. 1 a b white areas c colours yellow dark blue green red light blue  Statistical analysis 1c Results 1 2 3 p F p F p F F p F 2 χ 2 p 19 3 Table 1 Demographics of participants and risk factors   Control participants Type 2 diabetes patients a a Men/women 20/26 49/50 – 1.3 (0.6, 2.6) Age (years)   Men 66.4 (6.3) 65.9 (6.0) −0.5 (−2.8, 3.7) –   Women 63.8 (5.0) 65.9 (5.2) 2.1 (−0.4, 4.5) – b 4 (3–5) 4 (3–5) – – Diabetes duration (years) – 8.7 (6.1) – – 1c 5.5 (0.3) 6.8 (1.2) 1.4 (1.0, 1.7)*** – Use of insulin (%) – 29.3 – – c 28.3 70.7  6.2 (2.8, 13.5)*** Mean arterial pressure (mmHg) 97.5 (10.6) 102.6 (11.5) 4.8 (0.9, 8.8)* – History of macrovascular disease (%) 4.3 27.3 – 8.3 (1.8, 37.0)** Hypercholesterolaemia (%) 42.2 69.4 – 3.7 (1.7, 8.1)** Smoking ever (%) 47.8 66.3 – 2.2 (1.0, 4.7)* 2 27.2 (4.4) 28.0 (4.4) 0.8 (−0.8, 2.4) – Cognition (composite z-score) 0.13 (0.45) −0.10 (0.63) d – Data in first two columns are mean (SD) except for level of education, which is given as median (interquartile range) p p p a b c d Table 2 Volumes of cerebral compartments (ml) unadjusted for age or intracranial volume   Control participants Type 2 diabetes patients Men Women Men Women White matter 688.0 (49.3) 621.8 (46.9) 687.2 (56.5) 601.0 (56.5) Grey matter 401.7 (47.8) 404.4 (40.8) 388.3 (48.8) 360.1 (35.6) Total brain 1,090 (81.0) 1,026 (72.9) 1,076 (91.0) 961.1 (71.0) Lateral ventricle 33.9 (18.7) 22.1 (8.66) 36.9 (17.4) 29.8 (13.9) a 30.1 (23.9–36.6) 20.3 (16.6–28.0) 33.8 (24.9–42.6) 26.7 (20.6–35.3) CSF excluding lateral ventricles 273.3 (27.1) 222.1 (25.1) 279.7 (37.8) 229.4 (35.9) Total CSF 307.2 (32.3) 244.2 (26.3) 316.5 (42.4) 259.2 (41.2) WML 3.89 (5.76) 2.98 (4.70) 3.66 (5.37) 6.19 (14.2) a 1.81 (0.47–3.48) 1.00 (0.48–3.20) 2.16 (0.91–3.86) 2.56 (0.86–4.46) Intracranial volume 1,403 (90.7) 1,275 (82.2) 1,398 (104.2) 1,228 (93.7) a 3 Table 3 Adjusted tissue volume differences between participants with type 2 diabetes and control participants   Estimated volume difference (ml) F All  White matter 2.8 (−3.2, 8.8) 0.844  Grey matter −21.8 (−34.2, −9.4)** 12.091  Total brain −19.0 (−29.9, −8.1)** 11.863  Lateral ventricles 7.1 (2.3, 12.0)** 8.441  CSF not including lateral ventricles 9.5 (−0.4, 19.5) 3.594  Total CSF 16.7 (6.8, 26.5)** 11.247  LN WML a 4.684 Men  White matter 1.8 (−7.4, 11.0) 0.157  Grey matter −14.5 (−33.6, 4.6) 2.315  Total brain −12.7 (−30.6, 5.2) 2.026  Lateral ventricles 4.9 (−3.7, 13.5) 1.283  CSF not including lateral ventricles 6.9 (−10.2, 24.0) 0.650  Total CSF 11.8 (−4.9, 28.5) 1.993  LN WML b 1.781 Women  White matter 7.0 (−1.5, 15.6) 2.696  Grey matter −37.1 (−54.1, −20.1)*** 19.062  Total brain −30.1 (−44.9, 15.4)*** 16.695  Lateral ventricles 9.0 (3.2, 14.8)** 9.563  CSF not including lateral ventricles 17.2 (4.9, 29.4)** 7.802  Total CSF 26.1 (13.6, 38.7)*** 17.215  LN WML c 2.745 LN WML Data are mean (95% CI) adjusted for age, sex, intracranial volume and level of education (data missing for one woman with type 2 diabetes) p p p a b c Fig. 2 closed squares closed triangles open squares open triangles p  1c 1c p F p F p F z p F p F p F p F p F p p F p F p Discussion Our MR image-based segmentation paradigm showed that type 2 diabetes was significantly associated with smaller grey matter volumes and larger lateral ventricle and WML volumes, whereas white matter volume was not affected. A smaller grey matter volume suggests cortical atrophy, whereas a larger lateral ventricle volume may indicate subcortical atrophy. Separate analysis for men and women showed significantly smaller grey matter and total brain volumes and significantly larger lateral ventricle, CSF and total CSF volumes in female but not in male diabetic patients. 24 25 26 28 29 30 15 31 33 19 2 5 4 34 35 36 4 5 7 18 12 13 18 37 38 8 11 13 18 39 38 40 41 A limitation of our quantitative automated volumetric segmentation algorithm is slight overestimation of the CSF and slight underestimation of the lateral ventricle volume in subjects with markedly enlarged occipital horns. Some misclassification of interhemispheric CSF as lateral ventricle might also occur. However, the misclassified volumes are small compared with ventricle (about 2% and 6%) and CSF (about 0.3% and 0.4%) volumes and total CSF volumes are not affected. The artefacts were proportionally distributed across men and women and between type 2 diabetes patients and controls. Therefore, it is unlikely that these inaccuracies affected the analyses. Furthermore, all segmentations were inspected visually, so results will not have been affected by other segmentation errors. 42 25 26 27 30 20 Our automated segmentation method offers major advantages over manual methods by enabling precise, objective and reproducible volumetric measurements of cerebral tissues in large numbers of patients. In future studies, this method might be used to quantify the progression of abnormalities on MR images in patients with type 2 diabetes, as well as in patients with other conditions. Such studies using quantitative MR data should focus not only on the differences between groups of patients and controls, but also on differences within a patient group to identify determinants of structural brain changes. Electronic supplementary material Below is the link to the electronic supplementary material.