Informed consent was waived by the Medical Ethics Review Committee of the Amsterdam UMC. Inclusion criteria were a diagnosis of diabetes mellitus type 1 or 2, clinically suspected active CN and diabetic peripheral neuropathy. Individuals were classified as having clinically suspected active CN if they presented with an edematous and erythematous foot and an elevated foot temperature of > 2 degrees Celsius compared to the unaffected foot [22]. Exclusion criteria were concomitant foot diseases (e.g. osteomyelitis), recent trauma and unavailable digital patient files. Two clinical study groups were formed: active CN (CN group) and no active CN (no-CN group) depending on the final clinical diagnosis of a multidisciplinary team of medical specialists. This diagnosis was based on clinical data, presence of osseous abnormalities (e.g. fractures, joint subluxation or dislocation) on plain radiographs or DECT (the presence of BME on VNCa was not used in this diagnosis) assessed by a musculoskeletal radiologist, the need for a total contact cast, and the course of the disease during follow-up with a minimum of 3 months. All relevant diagnostics were performed on the same day that a patient presented with clinical signs of active CN. We adhered to the recommendations for diagnosis that are listed in the IWGDF guideline regarding CN [10]. MRI was not available, due to long waiting times in clinical practice. All DECT scans of the feet performed between January 2018 and September 2022 of people being treated at the Amsterdam UMC were retrospectively screened. A sub-analysis was performed in people without a previous diagnosis of CN, thus CN de novo. People with a history of CN might have fragmentation of the midfoot which might influence the results. It is therefore important to know if there is a significant difference CT value in this sub-group as well.

A dual-source CT scanner (SOMATOM Force; Siemens Healthcare, Forchheim, Germany) was used (80kVp (tube A) and Sn150kVp (tube B)). No contrast agents were used. CT parameters included: collimation of 0.6 mm, rotation time of 0.5 s, slice thickness and increment of 1.5 mm, mean dose length product (DLP) of 164.1 mGy*cm (range 113.7—246.0 mGy*cm), mean CT dose index volume (CTDIvol) of 6.2 mGy (range 3.9—8.5 mGy), and a medium smooth kernel (Qr54d or Qr40d). Automatic attenuation-based tube current modulation was applied on both tubes (150 reference mAs (tube A) and 380 reference mAs (tube B)).

Three image data sets were reconstructed from each DECT scan: a weighted average data set, a 80kVp dataset and a Sn150kVp dataset. Subsequently, all scans were processed using a three-material decomposition algorithm, differentiating between yellow bone marrow, red bone marrow, and bone mineral (SyngoVia VB40; Siemens Healthcare, Forchheim, Germany), a lower threshold of 50 Hounsfield unitis (HU) and a maximum of 800 HU. This allowed for the creation of VNCa-images by subtracting calcium. Densities between -150 HU and 100 HU were used on the color-coded bone marrow maps. BME is around 0 HU and appears as green on the color-coded map. On the color-coded map, higher values appear yellow turning eventually red and lower values appear blue gradually turning purple.

All scans of the feet were independently assessed by two medically trained researchers (C.M.B. and M.A.M.). The researchers were instructed and supervised to conduct the measurements by a musculoskeletal radiologist (M.M.) with over 30 years of experience in diabetic foot disease. The researchers were blinded for the radiology report and clinical information.

The anatomical locations where CN is most prevalent can be described according to the classification of Brodsky-Trepman [23, 24]. This is a reliable classification and important to the clinical communication about CN [13]. The Brodsky classification describes five types of CN according to their anatomical location. Type 1 (tarsometatarsal and naviculocuneiform joints) and type 2 (subtalar, talonavicular or calcaneocuboid joints) account for around 90% of all CN cases [25]. We decided to limit the measurement locations to five bones in the midfoot, in accordance with Brodsky type 1 and 2. This makes the measurements representative for a clinical setting. CN of the hindfoot, Brodsky-Trepman type 3, accounts approximately for 10% of the cases, mostly in the ankle (type 3A). CN of the calcaneus (type 3B) can lead to an avulsion fracture of the tendon-Achilles [23, 25]. The calcaneal body can be used as a reference standard due to its size and the rarity of CN presenting in the calcaneus or affecting the measurements in the calcaneus body.

A circular region of interest (ROI) of 0.5 cm2 was manually placed in the five locations in the midfoot: medial, intermediate and lateral cuneiform bones, cuboid bone and navicular bone (Fig. 1). The exact placement of the ROI was not fixed, but based on the color-coded map. The ROI was placed in a region where the color-coded map showed a lot of green (indicative of elevated CT values). When the color-coded map showed no green areas in one of the bones, the region that was estimated to contain the highest CT value was used. This was the region that contained the least amount of purple color-coded areas (indicative of very low CT values). ROIs were placed in the sagittal reconstruction and with a minimum of 2 mm from the cortex. The contralateral calcaneus (or, if unavailable, the ipsilateral calcaneus) was used as a reference location. A large ROI of 5.0 cm2 was placed in the middle of the calcaneus (Fig. 1), thereby providing a better representation of the average CT values. If a bone was fragmented and a circular ROI could not be placed, a freehand ROI was drawn. The mean CT value inside the ROI was noted. Afterwards, the measurements of both observers were pooled and the averaged CT value of the five measurements was used for further calculations. To evaluate if there are differences between the measurements in the five midfoot bones and to assess if one bone is more suitable for these kinds of measurements, an analysis between the two study groups with the separate measurements in the different bones was performed as well.

ROI placement in the medial cuneiform bone (left) and in the calcaneus (right) on the VNCa map 

The VNCa images were assessed qualitatively by a physician with three years of research experience in musculoskeletal radiology (M.A.M.) and an experienced musculoskeletal radiologist (M.M.) for the presence of BME indicative of active CN. The readers were blinded for the radiology report and clinical information and used the color-coded map for the assessment (Fig. 2). The presence of BME was considered positive if one or both reviewers considered BME to be present. Inter observer agreement was assessed.

Sagittal VNCa images of a foot with active CN (left) and without active CN (right). The white arrows point to the cuboid bone. In the foot with active CN the average CT value measured in the cuboid bone higher, shown as green or yellow indicating the presence of BME

SPSS (version 26.0, IBM, Armonk, New York, USA) was used for the statistical tests. The Kolmogorov–Smirnov test was used to test for normality. A student’s t-test or Mann–Whitney U test was used to test for significance between the groups and between the average CT values in the midfoot and reference location. The midfoot bones were also analyzed separately between the groups, as well as the highest measured CT value regardless of location. The Chi-squared test was conducted to test the significance for binary data. A p-value < 0.05 was considered as statistically significant. The intraclass correlation coefficient (ICC) or Cohen’s kappa were used to determine inter-observer agreement for respectively the quantitative and qualitative measurements. An ICC < 0.5 was considered “poor”, 0.5–0.74 “moderate”, 0.75–0.9 “good” and > 0.9 “excellent” [26]. A kappa of < 0 was “poor”, 0–0.20 “slight”, 0.21–0.40 “fair”, 0.41–0.60 “moderate”, 0.61–0.80 “substantial” and > 0.81 “almost perfect” [27]. In addition, a Bland–Altman analysis was conducted. A Receiver Operating Characteristic (ROC) analysis was performed. A cutoff CT value for diagnosing BME was determined using the Youden’s index based on the ROC analysis. Additionally, we calculated sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV).