The present study reported a novel approach to analyze the evolution of osseous structure within the femoral head affected by ONFH. We found ONFH affected not only the necrotic lesion but also the osseous structure outside the lesion. Additionally, the so-called lesion boundary (sclerotic boundary) was not always continuous, which was the main reason that the machine segmentation of the necrotic lesion still needs human assistance. The present study tried to conquer the technological limitation by analyzing the osseous structure without segmenting the necrotic lesion. We found that the volume of sclerotic bone, the volume of soft tissue, and the BMD of the entire femoral head changed with disease duration. The volume of sclerotic bone, when integrated with the BMD, might provide expectable performance on predicting radiographic progression.
Since ONFH primarily affects a limited subchondral part of the femoral head, instead of the entire femoral head, for a long time, researchers had been actively discussing the criteria to segment and classify the necrotic lesion [12]. To date, clinical practitioners get used to categorizing the necrotic lesions into large lesions, medium lesions, and small lesions [25, 26]. In real-world clinical practice, plain radiographs are the most used imaging modality to estimate the lesion size. At the same time, it is also generally acknowledged that plain radiographs are not an accurate modality to measure the lesion size [12]. A low signal line with a fatty center on T1-weighted MRI is a specified feature to confirm the diagnosis of ONFH, and a low signal line is frequently used as the criterion to segment the necrotic lesion. However, the low signal line is not always clear and continuous. The automatic process of segmenting the necrotic lesion, in fact, still relies on human assistance [12, 19]. CT is a better modality to observe the evolution of the osseous structure, and the sclerotic boundary is used as a segmentation criterion of the lesion. According to our study, the sclerotic boundary changed over time; meanwhile, it shared features similar to the low signal line that it was also not always clear and continuous.
Therefore, the present study analyzes the evolution of osseous structures within the femoral head instead of segmenting and analyzing the necrotic lesion. CT images are sensitive to identify the change of osseous structures that are technologically reflected by the HU [27]. HU is used as the tissue classification criteria to segment different tissues and organs based on their thresholds. In the present study, the tissues within the femoral head are classified into sclerotic bone, cancellous bone, and soft tissue, with HU thresholds different from each other. After the automatic identification of HU in the Mimics software, the three different tissues can also be automatically segmented based on their distinct HU thresholds.
The present study reports osseous structure-time variations that the volume of sclerotic bone and the BMD grow with disease duration, and the volume of soft tissue decreases. Most classification systems of ONFH acknowledged that the osseous structure within the lesion changes with time, like in ARCO stage I, the osseous change is absent on plain radiographs, but in ARCO stage II, sclerosis or cyst can be seen. However, the current classification systems do not take into consideration the dynamic evolution of the osseous structure. Previous research supported that the change of the osseous structure is the root cause of the progression of femoral head collapse [13, 18]. Indeed, lesion size and lesion location measurements are based on the significant change of the osseous structure. There are arguments on the effect of changing in the osseous structure. Yu et al. reported that the sclerotic boundary might prevent the collapse of the femoral head [17], but Utsunomiya et al. concluded that the onset of the sclerotic boundary might trigger the progression to collapse [18]. In our opinion, these valuable studies are generally isolated and static studies, since they ignore the dynamic evolution process of the osseous structure.
The present study found that the volume of sclerotic bone and the BMD were closely associated with the radiographic progression, which has potentials for the prediction of radiographic progression when being integrated together. Moreover, as mentioned above, the volume of sclerotic bone and the BMD grow with disease duration. This may explain why different conclusions were made when discussing the effect of sclerotic boundary on the radiographic progression. Previous researchers analyzed the sclerotic rim in a single follow-up time point, but the sclerotic rim would grow with time.
There are several limitations to the study. According to our analysis, disease duration could only explain the changes in the osseous structure to a limited extent (Table 2). The bone repair process in ONFH remains largely unknown. Several different biological processes are involved in this process, including oxidative stress, angiogenesis, bone turnover, and inflammation [2]. Our finding was consistent with the present understanding of the bone repair process. Secondly, only a small sample size of participants without a history of operative treatment was included in the present study, since most patients with large lesions should have undergone operative treatments. Thirdly, we abandoned to analyze the evolution of the same lesions of the same participants, since it is not ethical to perform CT scannings on the same participant in the interval of every 3 months. Fourthly, the radiographic progression is defined by the collapse or progressive collapse in the follow-up duration of 1 year. Thus, the prediction performance of the integrated indicators should be strictly limited to a short term of not more than 1 year. Lastly, this was a retrospective study, and we only included patients without collapse or with mild collapse. Hence, the evolution pattern of the osseous structure reported in the present study should be carefully reconsidered before it was used for the evaluation of patients with severe collapse.
In conclusion, the present study used a classification segmentation method to analyze the osseous structure within the femoral head instead of segmenting and analyzing the necrotic lesion. We found the dynamic evolution process of the osseous structures that the volume of sclerotic bone and the BMD grow with disease duration, but the volume of soft tissue decreases. The volume of sclerotic bone and the BMD are closely associated with the occurrence and progression of collapse. Hence, the collapse risk might change with disease duration due to the dynamic evolution of the osseous structure. Our findings might complement the static collapse risk evaluation method which only includes lesion size and lesion location, and help explain the controversial debates regarding the effect of osseous structure on radiographic progression.