Chronic subdural hematoma (CSDH) is a common disease in neurosurgery, which is more common in elderly patients. Although there are some reports on the treatment of CSDH with drugs, such as glucocorticoids and atorvastatin, surgical intervention is still the main method of treating CSDH, especially in the case of obvious compression symptoms. In the operation, one or multiple burr hole was used, followed by intraoperative irrigation and drainage. Most of the patients have good surgical results, but there is still a high recurrent rate which is 20-26% reported in different reports [1,4,9,14]. In recent years, with the rapid development of medical image digitization, using texture analysis to study medical images has become a hot spot. Texture analysis has the advantage of not relying on doctors’ subjective judgment, avoiding different doctors’ different interpretations of the same imaging data. Using texture analysis to determine whether CSDH relapse after surgery, achieving early prevention and reduction of recurrence can not only reduce the burden of future medical system, but also reduce the incidence rate and mortality related to recurrence.

This study was approved by the Ethics Committee of The Second Affiliated Hospital of Jiaxing University (Institutional review board number : 2023-ZFYJ-037).

CSDH is an enveloped hematoma located between the dura mater and the arachnoid membrane. It usually shows clinical symptoms and signs 3 weeks after the trauma. Burr-hole craniotomy is a common clinical treatment, which has good effect but high recurrence rate. Moreover, researches at home and abroad have different definitions of postoperative recurrence of CSDH. For example, a large multicenter cohort study in the UK defines CSDH recurrence as repeated surgical drainage within 60 days after surgery [15]. You et al. [24] thought if the patient has a hematoma accumulation again on the ipsilateral subdural space with neurological deficits, surgery is needed again, which was defined as CSDH recurrence. Therefore, in this study, recurrence was defined as recurrence of ipsilateral CSDH requiring reoperation within 2 months after the initial surgery.

Many non-surgical risk factors are thought to be related to the recurrence of CSDH, including age, patient status at admission, drinking, systemic diseases such as liver and kidney dysfunction, and bilateral CSDH [23]. Another multicenter retrospective study of 719 cases showed that age, gender, diabetes, hypertension, chronic renal failure, drinking, consciousness disorder on admission, and preoperative CT density were not related to CSDH recurrence, and antiplatelet therapy significantly affected the recurrence of CSDH [18]. A meta-analysis reported by Wang et al. [21] also showed that antithrombotic drugs are a risk factor for CSDH recurrence. Our study found that age (≥60 years), gender, history of trauma, hypertension, diabetes, and history of anticoagulant use were not risk factors for CSDH recurrence. Thus, this may be related to our small amount of data, and further research is needed to determine whether non-surgical risk factors are related to the recurrence of CSDH. But it also shows that it is difficult to judge the recurrence of CSDH only by general data.

In recent years, many literatures have found that the density, drainage, thickness, and typing of CSDHs are related to CSDH recurrence. The articles published by You et al. [24] show that homogeneous high-density hematomas can be used to predict whether recurrence after CSDH. Articles published by Liu et al. [13] showed that mixed density hematomas are independent predictors of CSDH recurrence. Ridwan et al. [15]’s study found that the separated stratified, separated, and trabecular Nakaguchi subtypes were independent predictors of CSDH relapse. However, there are certain limitations in predicting the recurrence of CSDH by imaging. There are differences in different doctors’ interpretation of the same imaging data, and even there will be greater differences. Therefore, the standardized interpretation of the imaging data of patients with CSDHs will eliminate the differences between different observers, which will help predict the recurrence of patients with CSDH in advance, formulate personalized treatments, reduce the recurrence rate and improve the quality of life.

The difference in the number of patients between the recurrence group and the non-recurrence group in this study is related to the recurrence rate of CSDH. Recurrence rates of CSDH have been reported to range from 9.2% to 26.5%, and recurrence rates also range from 5% to 33% [8,11]. The recurrence rate calculated from the number of patients in the two groups in our manuscript was 17.9%, which is consistent with the recurrence rate reported in the literature. We performed statistical analyses using previous literature, in which there were also differences in the number of patients between the two groups, to ensure the robustness of the analyses.

Texture analysis is a process of extracting texture feature parameters through image processing technology to obtain quantitative or qualitative description of texture [5]. At present, texture analysis is widely used in all walks of life, and also in medicine. For example, the application of texture analysis on cardiac MRI T1 and T2 maps provides quantitative imaging parameters for the diagnosis of acute and chronic heart failure-like myocarditis [2]; the texture analysis of susceptibility weighted imaging (SWI) can quantify Identifying patients with amyotrophic lateral sclerosis [10]; placental MRI texture analysis is helpful for prenatal diagnosis of implanted placental spectrum [4]; in the field of tumors, MRI-based texture analysis can be used for the differential diagnosis of pancreatic nonfunctional neuroendocrine tumors and solid pseudopapillary tumors [12]; predict the efficacy of neoadjuvant chemotherapy for cervical cancer [7]; predict bone metastases before prostate cancer treatment [20]; there is a review that texture analysis plays an important role in the evaluation of breast cancer diagnosis, prognosis and treatment response [6]. Moreover, texture analysis also has a good application prospect in neurosurgery, and many literatures have reported it. Texture analysis can identify different types of posterior cranial fossa tumors such as medulloblastoma, brain metastases, and hemangioblastoma [25]. Furthermore, texture analysis can identify glioblastoma and primary central nervous system lymphoma before surgery [22]. Meta-analysis shows that MRI texture analysis can accurately distinguish low-grade and high-grade gliomas [19].

According to literature reports [17], FLAIR sequence can more effectively distinguish subdural hematoma from cerebrospinal fluid than T2 sequence, and is an effective sequence for the detection of subdural hematoma. In this study, it was necessary to accurately delineate the boundary of CSDH and then perform texture analysis. Therefore, FLAIR sequence can effectively distinguish CSDH from cerebrospinal fluid, and is more suitable for this study than T2, SWI and other sequences. T2 FLAIR sequence was used for texture analysis in this study.

Therefore, this study is the first to report the texture analysis of CSDH by Firevoxel software, which avoids the deviation of analysis by different doctors. We found that the heterogeneity of the relapsed group (0.313±0.056) was greater than that of the non-relapsed group (0.224±0.087), and the difference was statistically significant (p=0.017). And when the cut-off point of heterogeneity is 0.820, the sensitivity, specificity, and accuracy of judging the recurrence of CSDH through heterogeneity were 83.3%, 80.4%, and 80.7%, respectively.

Entropy is obtained by computing the probability distribution of the gray values of pixels in the image. Specifically, entropy measures the randomness and uncertainty of the gray value in the image texture. Quantifying texture complexity: entropy can quantify the complexity of image texture. High entropy values indicate complex image texture and high information content, while low entropy values indicate simple image texture and low information content. This quantification facilitates a deeper understanding and analysis of the image. Texture classification and recognition : in texture classification and recognition tasks, entropy can be used as one of the features to help distinguish different textures. By calculating the entropy of different texture images, a classifier can be constructed to achieve automatic texture classification and recognition. Image quality evaluation : entropy can also be used to evaluate the quality of images. In the process of image compression and transmission, the change of entropy value can reflect the loss of image information. Therefore, entropy can be used as an important index for image quality evaluation.

This study confirms that texture analysis can be used to predict postoperative recurrence of CSDH. In addition, texture analysis is a non-invasive, economical, and easy-to-use technique. It is of great significance to predict the recurrence of CSDH after surgery, that is, it can not only provide a reference for the diagnosis and treatment program, ensure the safety of patients, but also appropriately reduce the medical expenses, and reduce the burden for families and society. However, the shortcoming of this study is that the amount of data included is small, and more institutions are needed to join the study, which further confirms the value of texture analysis in the postoperative recurrence of CSDH.

The analysis of the texture of CSDH provides a new method for judging the recurrence of CSDH, which is helpful to achieve early prevention and reduction of recurrence. This will not only reduce the burden on the healthcare system in the future, but also reduce the morbidity and mortality associated with recurrence.