Even though various biomechanical models have been developed so far to study the properties and behaviour of the foot [14, 18, 19, 21, 22], the present study focuses on developing a methodology for the functional assessment of the foot-ankle complex and for the definition of a functional model of the diabetic neuropathic foot.
A method for capturing forefoot, midfoot and hindfoot motion during different gait tasks have been proposed. The model includes tibia and fibula, hindfoot, midfoot and forefoot, and allows investigation of 3-dimensional foot and ankle kinematics through stereophotogrammetry. A new model has been generated since available foot protocols were not suitable for this type of analysis [18, 19, 21, 22, 26]. One important limitation of the literature was that the 3 planar motion of the midfoot was not evaluated. As the diabetic foot disease accounts for midfoot structural polymorphism which commonly leads to plantar ulceration [34, 35], the authors believe that a suitable model to describe the diabetic foot biomechanics should perform 3D midfoot kinematic analysis. Furthermore, this was confirmed by the results reported in Figure 3 where, the diabetic group has statistically significant differences in midfoot kinematic parameters over a large part of the gait cycle. Nevertheless the forefoot should be considered entirely and not represented by a single toe as the hallux  because it is considered the high risk zone for plantar ulcer formation [4, 36]. This was confirmed by the results reported in Figure 3 where, the diabetic group showed statistically significant differences in forefoot kinematic parameters over the full gait cycle in the sagittal and coronal planes, and in the 50% of gait cycle in the transversal plane. Furthermore in the literature has been reported that in diabetic patients, changes in weight bearing patterns are linked to limited joint mobility that occurs mostly at metatarsophalangeals and subtalar joints. Nevertheless the location of forefoot plantar ulcers in diabetic subjects has been demonstrated to be highly correlated with rearfoot alignment . In addition to the different types of mechanisms of excessive pressure loading, abnormal alignment of the foot also affects pressure loading on the foot. Finally patients with an uncompensated forefoot varus or forefoot valgus (inverted or everted forefoot) had ulcers located at the first or fifth metatarsal head. Similarly, an inverted heel position has been associated with lateral ulcers, whereas an everted heel position has been associated with medial ulcers . So far the authors believe that a technique for the measurement of rearfoot-forefoot-midfoot structures alignment is needed in understanding the aetiology of diabetic foot ulcers. Finally, the triplanar orientation of the joint axis allows for movement in all three body planes and thus provides a mechanism for compensatory motion if there is presence of structural anomalies in the foot , which is indeed the case of the diabetic foot.
Protocols which adopt rigid array of markers and procedures which implies calibration techniques [13–15] where not adopted because they present as major disadvantage the time required for each anatomical landmark calibration trial. Furthermore when errors due to skin artefacts affect protocols using mounting plates, it is difficult to identify the relative contribution of each individual marker, as the errors affect the complete cluster as a whole. Therefore we choose direct skin marker placement on ALs even though these are more subject to errors due to skin artefacts and markers misplacement . An extensive review of the problem is given in Leardini et al. 2005 . This protocol tries to prevent these errors by using the above described algorithm with a static calibration and by controlling for markers occlusion.
The ALs were selected in order to be easily palpated and identified. The location of the ALs was chosen so that BEFs were directly defined with no need of technical frames definition .
As suggested by Baker  a foot model have been applied to a pathologic population, even though in this specific case one of the existing models could not be adopted as was previously done by Woodburn  for the reasons reported above. In the present work, ten neuropathic subjects have been evaluated and results of this analysis showed major statistically significant differences between the two populations both in the forefoot vs midfoot and midfoot vs hindfoot dorsi-plantarflexion (100% of frame of the gait cycle). Also important statistically significant differences were observed in midfoot vs hindfoot internal-external rotation (90% of frame of the gait cycle), in forefoot vs midfoot inversion-eversion, in ankle internal-external rotation (96% of frame of the gait cycle) and inversion-eversion (92% of frame of the gait cycle). Thus to confirm the validity of a similar approach in order to assess diabetic neuropathics' biomechanics impairment.
An important step in assessing the effectiveness of gait analysis is to establish the precision of the data collection  and the accuracy in determining the anatomical landmarks and joint embedded frames definition. An effort in this direction is documented in Table 2 and 3 were a detailed description of ALs, and BEFs together with instruction for marker placement can be found. It is, of course, important to evaluate the major sources of variation in gait analysis (true variation in the subject's gait and artefact from the measurement procedure) . We therefore need an estimate of the expected variability in joint rotation angles estimation. This is important when, for example, comparing a patient data against normative standards - we need to know how much difference is significant. Normal biological variation affects kinematics data since subjects never walk in the exact same way in every trial, therefore variability introduced by the subject within a test session were examined. Variability is also introduced by the measurement procedure by means of anatomical landmarks identification and skin movement artefact. Therefore joint rotation angles variability due to differences between clinicians, and by the subject within and between test days were examined [30–32]. Three walking trials per subject were acquired together with a static acquisition. The same procedure was applied both to normal and pathological subject, in order to check the feasibility of this approach onto diabetic subjects.
Repeatability has been assessed by the mean, range and SD values of model segments and joint rotation angles, together with Vabs coefficient of Noonan , following the methodology proposed by Schwartz . The results showed the suitability of this method as they were found comparable with similar studies [19, 21, 29, 21, 22, 30, 31] and this gives strength to the present work. The repeatability analysis on the pathological subject shows results comparable to the normal one in terms of SD and Vabs in most of the rotation angles, which asses the suitability of this protocol to this type of patients.
Based on the results reported in Table 5 and [see Additional file 2], we can assess that the model has been tested for repeatability therefore anatomical landmark identification can be considered feasible.
The elementary movements were used in order to check the ability of the model to measure sub-segments rotations. The range, mean and SD values of the angles obtained by executing passive movements of the foot allowed us to test the suitability of the chosen reference systems and angles definition. We think this is in fact the only possible way to quantitatively assess the capability of the model of measuring correctly model segments rotations. Since the possibility of executing elementary movements of each model segment is still under study in our laboratory, model segments rotations were obtained by performing full foot elementary passive movements. Then the movement of each segment component was obtained by the model. The rotations relative to model segments are considered clinically acceptable .