Predicting neurosurgical referral outcomes in patients with chronic subdural hematomas using machine learning algorithms - A multi-center feasibility study

Background

Chronic subdural hematoma (CSDH) incidence and referral rates to neurosurgery are increasing. Accurate and automated evidence-based referral decision-support tools that can triage referrals are required. Our objective was to explore the feasibility of machine learning (ML) algorithms in predicting the outcome of a CSDH referral made to neurosurgery and to examine their reliability on external validation.

Methods

Multicenter retrospective case series conducted from 2015 to 2020, analyzing all CSDH patient referrals at two neurosurgical centers in the United Kingdom. 10 independent predictor variables were analyzed to predict the binary outcome of either accepting (for surgical treatment) or rejecting the CSDH referral with the aim of conservative management. 5 ML algorithms were developed and externally tested to determine the most reliable model for deployment.

Results

1500 referrals in the internal cohort were analyzed, with 70% being rejected referrals. On a holdout set of 450 patients, the artificial neural network demonstrated an accuracy of 96.222% (94.444-97.778), an area under the receiver operating curve (AUC) of 0.951 (0.927-0.973) and a brier score loss of 0.037 (0.022-0.056). On a 1713 external validation patient cohort, the model demonstrated an AUC of 0.896 (0.878-0.912) and an accuracy of 92.294% (90.952-93.520). This model is publicly deployed: https://medmlanalytics.com/neural-analysis-model/.

Conclusion

ML models can accurately predict referral outcomes and can potentially be used in clinical practice as CSDH referral decision making support tools. The growing demand in healthcare, combined with increasing digitization of health records raises the opportunity for ML algorithms to be used for decision making in complex clinical scenarios.

Development and Validation of the Chronic Subdural HematOma Referral oUtcome Prediction Using Statistics (CHORUS) Score: A Retrospective Study at a National Tertiary Center

BACKGROUND

Chronic subdural hematoma (CSDH) is a common neurosurgical condition with an increasing rate of patient referrals. CSDH referral decision-making is a subjective clinical process, and our aim was to develop a simple scoring system capable of acting as a decision support tool aiding referral triage.

METHODS

A single tertiary center retrospective case series analysis of all CSDH patient referrals from 2015 to 2020 was conducted. Ten independent variables used in the referral process were analyzed to predict the binary outcome of either accepting or rejecting the CSDH referral. Following feature selection analysis, a multivariable scoring system was developed and evaluated.

RESULTS

1500 patient referrals were included. Stepwise multivariable logistic and least absolute shrinkage and selection operator regression identified age <85 years, the presence of headaches, dementia, motor weakness, radiological midline shift, a reasonable premorbid quality of life, and a large sized hematoma to be statistically significant predictors of CSDH referral acceptance (P <0.04). These variables derived a scoring system ranging from -9 to 6 with an optimal cut-off for referral acceptance at any score >1 (P <0.0001). This scoring system demonstrated optimal calibration (brier score loss = 0.0552), with a score >1 predicting referral acceptance with an area under the curve of 0.899 (0.876-0.922), a sensitivity of 83.838% (76.587-91.089), and a specificity of 96.000% (94.080-97.920).

CONCLUSIONS

Certain patient specific clinical and radiological characteristics can predict the acceptance or rejection of a CSDH referral. Considering the precision of this scoring system, it has the potential for effectively triaging CSDH referrals.

Is a wound drain needed routinely after anterior cervical discectomy and fusion?

Background

Postoperative subfascial wound drains are commonly used in anterior cervical discectomy and fusion (ACDF) surgeries to help lower the risk of complications from postoperative haematomas (POHs) and surgical site infections (SSIs). However, controversy surrounding their efficacy remains, due to limited guidelines and a lack of conclusive evidence. The primary aim of this study is to determine whether placing a drain routinely offers any therapeutic benefit for postsurgical outcomes in ACDF intervention, hence re-evaluating their routine use in clinical practice.

Methods

This retrospective study analyzed on patients who underwent ACDF procedures between January 2013 and December 2023 at a tertiary neurosurgical centre in the UK. Patients were categorized into drain and nondrain cohorts and stratified according to several predictor variables including baseline demographics, and clinico-social characteristics. The primary outcome measure investigated in this study was the incidence of postoperative complications and length of stay (LOS) between those who received a drain and those who did not.

Results

A total of 1,938 patients met the inclusion criteria and were included in our analysis. Of these patients, 1,614 (83.3%) had subfascial drains placed during surgery. Baseline demographics differed slightly between both cohorts, with patients in the drain group having a higher median age (p=.02), and a significantly higher proportion were male (p=.02), hypertensive (p<.01), drank alcohol (p<.01) and smoked (p=.04). Between the 2 groups, we observed no significant difference in rates of reoperation for POHs (p=.43), SSIs (p=.85) or LOS (p=.18).

Conclusions

Our study found no significant differences in the incidence of POHs or SSIs between post-ACDF patients with drains and those without, nor was there any difference in postoperative length of stay between the groups. Therefore, our results do not support the routine use of subfascial drains in clinical practice.

What Degree of Radiological Compression Predicts Cauda Equina Syndrome: A Retrospective Study at a National Tertiary Center

OBJECTIVE

Cauda equina syndrome (CES) is diagnosed both clinically and radiologically. However, it's unclear if a specific degree of cauda equina compression on imaging can confirm the diagnosis. This study aimed to identify an optimal percentage of prolapse to canal ratio (PCR) on magnetic resonance imaging that correlates with cauda equina compression, facilitating reliable decision-making for CES symptoms.

METHODS

A single-center retrospective case series analysis was conducted from 2020 to 2021. Sixty-one patients who underwent emergency lumbar discectomy within 48 hours of presentation were included, divided into a CES group and a disc prolapse group. CES cases were identified using International Classification of Diseases, 10th Revision code G83.4. PCR was calculated by dividing the width of the disc herniation by the total width of the spinal canal at the level of the herniation on single mid-sagittal T2 magnetic resonance imaging scans, using the IC Measure software on Windows.

RESULTS

CES most frequently occurred at the L4/5 level (37/61). The median PCR in CES cases was 76.19% (66.67-85.71), significantly higher than in disc prolapse cases, 48.08% (31.33-55.56) (P < 0.001). A PCR threshold of 40% maximized sensitivity at 100% but had a specificity of 45%. Conversely, a threshold of 75% maximized specificity at 100% with a sensitivity of 50%. The optimal PCR, determined by the Youden index, was 66%, yielding a sensitivity of 75% and specificity of 97%, with an area under the curve of 0.923.

CONCLUSIONS

Simple radiological measurements of PCR can potentially triage CES patients and guide their management. Future studies should correlate PCR with clinical signs and symptoms for a comprehensive assessment.

First record of Wroughton’s Small Spiny Mouse Mus phillipsi Wroughton, 1912 (Rodentia: Muridae) from Odisha, India with notes on diversity and distribution of other rodents

We report the occurrence of Wroughton’s Small Spiny Mouse Mus phillipsi Wroughton, 1912 based on a specimen collected from Gajapati District, Odisha.  With this species, the diversity of order Rodentia in Odisha increases to 17 species under three families and 12 genera.  An updated checklist of the rodents with distribution localities and threats to various species in Odisha is also presented. 

Is Robust Optimization Essential When Planning Pencil Beam Scanned Proton Therapy for Intracranial Lesions?

PURPOSE/OBJECTIVE(S)

Creating intensity modulated proton therapy (IMPT) plans usually involves a robust optimization step to account for uncertainties in proton range and positioning instead of using the PTV margins typically seen in photon IMRT planning. Because robust optimization adds significant planning time per iteration, and proton planning typically involves many iterations to obtain an optimal plan, this project evaluates whether a PTV approach can be used to more efficiently create plans for intracranial lesions by comparing plan quality from both approaches to determine if they produce equivalent plans.

MATERIALS/METHODS

Five patients with intracranial lesions treated with IMPT at our center were randomly chosen for this study. 3-4 beams were used to treat CTVs ranging between 46 and 246 cc. Patients were treated on a Mevion S250i single-vault proton machine with Hyperscan. Static blocks (7mm conformed to the CTV) were used for all cases and plans were created in a treatment planning system using robustness criteria of 3mm (position uncertainty) and 3.5% (range uncertainty). For each patient, the CTV was uniformly grown by 3 mm to create a PTV. The optimization criteria used in the clinical plan were used as baseline to create two plans - one using robust optimization for CTV coverage and one without robust optimization for PTV coverage. A script was used to time two otherwise identical optimization runs. All plans were normalized so that the prescription was delivered to 95% of the CTV. The plan was robustly evaluated under conditions of 3.5% range uncertainty and 2mm positional uncertainty. For each nominal plan, the CTV dose heterogeneity and conformity as well as the following dose metrics were collected: CTV D99% and D98%, Normal Brain V100%, V90%, V80%, V50% as well as overall plan Dmax (to 0.03cc). For each plan, the minimum CTV D99 and maximum Brain Dmax were also collected for the robust evaluation scenarios. All collected metrics were compared between the robust CTV-based and non-robust PTV-based plans using paired t-tests with 5% significance.

RESULTS

For the five cases investigated here, all dosimetric metrics investigated were not significantly different between the CTV-based and PTV-based plans except for the plan maximum dose (CTV-based: 104.6% Rx - 110.0% Rx, PTV-based: 105.6% Rx-110.6% Rx, p = 0.029). The optimization times were also significantly different, averaging 1532 s for CTV-based plans versus 252 s for the PTV-based plans (p = 0.004).

CONCLUSION

For the plans investigated here, non-robust PTV planning approach creates plans of very similar quality to a robust CTV-based plan, while having significantly shorter planning times.

CT-on-Rails Utilization for Image Guidance and Plan Adaptation at a Single-Room Proton Therapy Center

PURPOSE/OBJECTIVE(S)

In-room CT-on-rails is a relatively rare imaging modality in proton therapy. This work evaluated the utilization of CT-on-rails (CTOR) for image guidance and plan adaptation at a single-room proton therapy center.

MATERIALS/METHODS

All patients treated from the proton center beginning treatment to the current date were evaluated (May 2021 - January 2023). Patients were treated using a Mevion S250i scanning beam proton therapy system with an in-room Definition Edge CTOR. The image guidance technique utilized for each patient was evaluated (CTOR, orthogonal planar kV imaging, or both). For patients who required a new CT for plan adaptation, the use of CT images from CTOR or a new simulation CT was assessed.

RESULTS

Of the 132 patients treated over the evaluated time period, 56.0% utilized CTOR for image guidance at some point during their treatment. 21.2% utilized CTOR as the only image guidance technique, while 34.8% utilized CTOR in combination with kV planar imaging. For 43.9% of patients, only orthogonal planar kV imaging was utilized. Head and neck treatment sites most often utilized CTOR alone as an image guidance technique, while central nervous system sites most often utilized kV planar imaging alone. Of the patients who required new CT imaging for plan adaptation, 75.0% utilized CTOR images. Some reasons for patients not being able to utilize CTOR imaging for re-planning were the need for 4DCT imaging (which was not available on the CTOR) or the need to re-make immobilization devices (such as thermoplastic masks due to becoming loose).

CONCLUSION

CTOR was shown to be utilized for image guidance in the majority of patients receiving proton therapy at a single-room proton therapy center. CTOR was able to be used for plan adaptation in a large majority of patients, eliminating the need to schedule additional CT simulation appointments for these patients.