Reconstruction of Palatal Defects Using a Composite Chondromucosal Nasoseptal Flap and Comparative Analysis.

OBJECTIVE: The use of composite chondromucosal nasal septal flaps (ccNSF) has been demonstrated to be effective in cadaveric studies for the anterior skull base and the orbit. However, their application in the clinical setting remains unexplored. Our study aims to introduce a new method for treating palatal defects using ccNSF. Additionally, we studied the average NSF area and compared it to the average palate area. METHODS: We collected 108 CT scans from the medical records of patients without head and neck pathologies from a tertiary medical institution. We quantified the quadrangular (septal) cartilage and palate areas. Furthermore, we included a clinical case in which we used the ccNSF for the palatal defect reconstruction. This was to compare the mean area between the palate and the septal cartilage. RESULTS: The ccNSF covered the palatal defect without any significant complications for the first 9 months of follow-up. A total of 102 CT scans met the inclusion criteria and were measured. We found that the mean quadrangular cartilage had a length of 2.50 (±0.52) cm, a width of 2.28 (±0.51) cm, and an area of 5.43 (±1.68) cm2. The mean palate length was 2.73 (±0.44) cm, with a width of 3.13 (±0.34) cm, and area of 7.87 (±1.43) cm2. CONCLUSIONS: The ccNSF proved successful in palatal defect reconstruction, resulting in positive outcomes and no major complications until the 9-month follow-up. The ccNSF is a useful flap that avoids the use of free flap transfer and its associated morbidities. LEVEL OF EVIDENCE: 4 Laryngoscope, 2024.

Predicting Therapeutic Response to Hypoglossal Nerve Stimulation Using Deep Learning.

OBJECTIVES: To develop and validate machine learning (ML) and deep learning (DL) models using drug-induced sleep endoscopy (DISE) images to predict the therapeutic efficacy of hypoglossal nerve stimulator (HGNS) implantation. METHODS: Patients who underwent DISE and subsequent HGNS implantation at a tertiary care referral center were included. Six DL models and five ML algorithms were trained on images from the base of tongue (BOT) and velopharynx (VP) from patients classified as responders or non-responders as defined by Sher's criteria (50% reduction in apnea-hypopnea index (AHI) and AHI < 15 events/h). Precision, recall, F1 score, and overall accuracy were evaluated as measures of performance. RESULTS: In total, 25,040 images from 127 patients were included, of which 16,515 (69.3%) were from responders and 8,262 (30.7%) from non-responders. Models trained on the VP dataset had greater overall accuracy when compared to BOT alone and combined VP and BOT image sets, suggesting that VP images contain discriminative features for identifying therapeutic efficacy. The VCG-16 DL model had the best overall performance on the VP image set with high training accuracy (0.833), F1 score (0.78), and recall (0.883). Among ML models, the logistic regression model had the greatest accuracy (0.685) and F1 score (0.813). CONCLUSION: Deep neural networks have potential to predict HGNS therapeutic efficacy using images from DISE, facilitating better patient selection for implantation. Development of multi-institutional data and image sets will allow for development of generalizable predictive models. LEVEL OF EVIDENCE: N/A Laryngoscope, 2024.

Quantitative measurements of radiation-induced fibrosis for head and neck cancer: A narrative review.

Objectives: To provide a comprehensive summary of the different modalities available to measure soft tissue fibrosis after radiotherapy in head and neck cancer patients. Data Sources: PubMed, Scopus, and Web of Sciences. Review Methods: A search was conducted using a list of medical subject headings and terms related to head and neck oncology, radiation fibrosis, and quantitative measurements, including bioimpedance, MRI, and ultrasound. Original research related to quantitative measurement of neck fibrosis post-radiotherapy was included without time constraints, while reviews, case reports, non-English texts, and inaccessible studies were excluded. Discrepancies during the review were resolved by discussing with the senior author until consensus was reached. Results: A total of 284 articles were identified and underwent title and abstract screening. Seventeen articles had met our criteria for full-text review based on relevance, of which nine had met our inclusion criteria. Young's modulus (YM) and viscoelasticity measures have demonstrated efficacy in quantifying neck fibrosis, with fibrotic tissues displaying significantly higher YM values and altered viscoelastic properties such as increased stiffness rate-sensitivity and prolonged stress-relaxation post-radiation. Intravoxel incoherent motion offers detailed insights into tissue changes by assessing the diffusion of water molecules and blood perfusion, thereby differentiating fibrosed from healthy tissues. Shear wave elastography has proven to be an effective technique for quantifying radiation-induced fibrosis in the head and neck region by measuring shear wave velocity. Conclusion: There are various modalities to measure radiation-induced fibrosis, each with its unique strengths and limitations. Providers should be aware of these implications and decide on methodologies based on their specific clinical workflow. Level of Evidence: Step 5.

Impact of Adjuvant Radiotherapy Setting on Quality-of-Life in Head and Neck Squamous Cell Carcinoma.

OBJECTIVE: To determine differences in post-treatment QoL across treatment settings in patients receiving adjuvant radiation therapy for head and neck squamous cell carcinoma (HNSCC). METHODS: This was a prospective observational cohort study of patients with HNSCC initially evaluated in a head and neck surgical oncologic and reconstructive clinic at an academic medical center (AMC). Participants were enrolled prior to treatment in a prospective registry collecting demographic, social, and clinical data. Physical and social-emotional QoL (phys-QoL and soc-QoL, respectively) was measured using the University of Washington-QoL questionnaire at pre-treatment and post-treatment visits. RESULTS: A cohort of 177 patients, primarily male and White with an average age of 61.2 ± 11.2 years, met inclusion criteria. Most patients presented with oral cavity tumors (n = 132, 74.6%), had non-HPV-mediated disease (n = 97, 61.8%), and were classified as Stage IVa (n = 72, 42.8%). After controlling for covariates, patients treated at community medical centers (CMCs) reported a 7.15-point lower phys-QoL compared with those treated at AMCs (95% CI: -13.96 to -0.35, p = 0.040) up to 12 months post-treatment. Additionally, patients who were treated at CMCs had a 5.77-point (-11.86-0.31, p = 0.063) lower soc-QoL score compared with those treated at an AMC, which was not statistically significant. CONCLUSION: This study revealed that HNSCC patients treated with radiation at AMCs reported significantly greater phys-QoL in their first-year post-treatment compared to those treated at CMCs, but soc-QoL did not differ significantly. Further observational studies are needed to explore potential factors, including treatment planning and cancer resource engagement, behind disparities between AMCs and CMCs. LEVEL OF EVIDENCE: Step 3 Laryngoscope, 2024.

Six-week postoperative opioid use and pain following a randomized controlled trial evaluating multimodal analgesia for head and neck free flap patients.

Introduction: Head and neck malignancy treatment often involves invasive surgeries, necessitating effective postoperative pain control. However, chronic reliance on opioid medications remains a challenge for many patients after surgery. Multimodal analgesia (MMA) within enhanced recovery after surgery protocols has shown success in limiting narcotic pain medications for other cancer types. In a prior study, MMA comprising acetaminophen, ketorolac, gabapentin, and a neurogenic block reduced opioid use in the 7-day postoperative period for major head and neck reconstructive surgery. This study investigates the impact of multimodal analgesia on opioid prescription and pain during the 6-week postoperative period for patients undergoing major head and neck oncologic surgeries, aiming to understand the longer-term effects of narcotic use. Methods: The study retrospectively examined participants in a [hybrid type 1 effectiveness-implementation pragmatic trial to assess multimodal analgesia's long-term effectiveness in head and neck free flap surgery. Arm A received scheduled acetaminophen and as-needed opioids, while Arm B received scheduled gabapentin, ketorolac, a regional nerve block at the donor site, scheduled acetaminophen, and as-needed opioids. Retrospective data collection included opioid prescription use and pain scores up to 6 weeks after surgery, gathered from the Kansas prescription drug monitoring program, K-TRACS. Results: Thirty patients participated, 14 in Arm A and 16 in Arm B. The average morphine milligram equivalents per day of filled prescriptions were not significantly different between Arm A and Arm B (7.23 vs. 7.88, p = .845). Additionally, average pain scores at 6 weeks showed no significant difference between the two groups (1.4 vs. 1.9, p = .612). Conclusion: Patients with head and neck cancer treated with multimodal analgesia during the perioperative period did not exhibit significant differences in opioid use and pain within 6 weeks after discharge. To confirm these findings, a re-examination with strict measures of opioid use and scheduled pain assessments in a prospective manner is warranted. Level of Evidence: 4.

Right Versus Left Cuff Position for Upper Airway Stimulation.

OBJECTIVE: Upper airway stimulation (UAS) is a treatment option for obstructive sleep apnea in which electrical stimulation is applied to the hypoglossal nerve. Nerve branches that control tongue protrusion are located inferiorly. Due to positioning, left-sided implants are typically placed with an inferiorly oriented electrode cuff (L-down) as opposed to superiorly on the right (R-up). In this study, we assess the impact of left- versus right-sided UAS on patient outcomes. STUDY DESIGN: Retrospective cohort study. SETTING: Tertiary Academic Medical Center. METHODS: Patients who underwent UAS implantation between 2016 and 2021 with an L-down or R-up oriented cuff as confirmed by X-ray were included. Data were collected retrospectively. Most recent sleep study variables were used for analysis. RESULTS: A total of 190 patients met the inclusion criteria. The average age was 61.0 ± 11.0 years, with 55 (28.9%) females. L-down orientation was present in 21 (11.1%) patients vs 169 (88.9%) R-up. Indications for L-down included hunting/shooting (n = 15), prior radiation/surgery (n = 4), central port (n = 1), and brachial plexus injury (n = 1). Adherence was higher among L-down patients (47.1 vs 41.0 hours use/week, P = .037) in univariate analysis, with a similar time to adherence data collection (4.4 vs 4.2 months, P = .612), though this finding was not maintained in the multivariate regression analysis. Decrease in apnea-hypopnea index (21.3 vs 22.8, P = .734), treatment success (76.5% vs 84.0%, P = .665), functional threshold (1.5 vs 1.6, P = .550), therapeutic amplitude (2.3 vs 2.4, P = .882), and decrease in Epworth Sleepiness Scale (4.9 vs 2.6, P = .060) were not significantly different between cohorts. CONCLUSION: This study is the first to examine the orientation of the UAS electrode cuff concerning the electrodes' natural position and the potential effect on postoperative outcomes. Our study found no significantly different treatment outcomes between the L-down versus R-up cohort, with the exception of device adherence, which was significantly higher in the L-down group on univariate analysis though not on multivariate analysis. Future studies with larger patient cohorts are needed to further investigate this potential relationship between treatment outcomes and electrode cuff orientation.

Trends in the Utilization of Surgical Modalities for the Treatment of Drug-Resistant Epilepsy: A Comprehensive 10-Year Analysis Using the National Inpatient Sample.

BACKGROUND AND OBJECTIVES: Epilepsy is considered one of the most prevalent and severe chronic neurological disorders worldwide. Our study aims to analyze the national trends in different treatment modalities for individuals with drug-resistant epilepsy and investigate the outcomes associated with these procedural trends in the United States. METHODS: Using the National Inpatient Sample database from 2010 to 2020, patients with drug-resistant focal epilepsy who underwent laser interstitial thermal therapy (LITT), open surgical resection, vagus nerve stimulation (VNS), or responsive neurostimulation (RNS) were identified. Trend analysis was performed using piecewise joinpoint regression. Propensity score matching was used to compare outcomes between 10 years prepandemic before 2020 and the first peak of the COVID-19 pandemic. RESULTS: This study analyzed a total of 33 969 patients with a diagnosis of drug-resistant epilepsy, with 3343 patients receiving surgical resection (78%), VNS (8.21%), RNS (8%), and LITT (6%). Between 2010 and 2020, there was an increase in the use of invasive electroencephalography monitoring for seizure zone localization (P = .003). There was an increase in the use of LITT and RNS (P < .001), while the use of surgical resection and VNS decreased over time (P < .001). Most of these patients (89%) were treated during the pre-COVID pandemic era (2010-2019), while a minority (11%) underwent treatment during the COVID pandemic (2020). After propensity score matching, the rate of pulmonary complications, postprocedural hematoma formation, and mortality were slightly higher during the pandemic compared with the prepandemic period (P = .045, P = .033, and P = .026, respectively). CONCLUSION: This study indicates a relative decrease in the use of surgical resections, as a treatment for drug-resistant focal epilepsy. By contrast, newer, minimally invasive surgical approaches including LITT and RNS showed gradual increases in usage.

Surgical management of malignant melanotic nerve sheath tumors: an institutional experience and systematic review of the literature.

OBJECTIVE: Malignant melanotic nerve sheath tumors are rare tumors characterized by neoplastic melanin-producing Schwann cells. In this study, the authors report their institution's experience in treating spinal and peripheral malignant melanotic nerve sheath tumors and compare their results with the literature. METHODS: Data were collected from 8 patients who underwent surgical treatment for malignant melanotic nerve sheath tumors between 1996 and 2023 at Mayo Clinic and 63 patients from the literature. Time-to-event analyses were performed for the combined group of 71 cases to evaluate the risk of recurrence, metastasis, and death based on tumor location and type of treatment received. Unpaired 2-sample t-tests and Fisher's exact tests were used to determine statistical significance between groups. RESULTS: Between 1996 and 2023, 8 patients with malignant melanotic nerve sheath tumors underwent surgery at the authors' institution, while 63 patients were identified in the literature. The authors' patients and those in the literature had the same mean age at diagnosis (43 years). At the authors' institution, 5 patients (63%) experienced metastasis, 6 patients (75%) experienced long-term recurrence, and 5 patients (62.5%) died. In the literature, most patients (60.3%) were males, with a peak incidence between the 4th and 5th decades of life. Nineteen patients (31.1%) were diagnosed with Carney complex. Nerve root tumors accounted for most presentations (n = 39, 61.9%). Moreover, 24 patients (38.1%) had intradural lesions, with 54.2% (n = 13) being intramedullary and 45.8% (n = 11) extramedullary. Most patients underwent gross-total resection (GTR) (n = 41, 66.1%), followed by subtotal resection (STR) (n = 12, 19.4%), STR with radiation therapy (9.7%), and GTR with radiation therapy (4.8%). Sixteen patients (27.6%) experienced metastasis, 23 (39.7%) experienced recurrence, and 13 (22%) died. Kaplan-Meier analyses showed no significant differences among treatment approaches in terms of recurrence-free, metastasis-free, and overall survival (p > 0.05). Similar results were obtained when looking at the differences with respect to intradural versus nerve root location of the tumor (p > 0.05). CONCLUSIONS: Malignant melanotic nerve sheath tumors are rare tumors with a high potential for malignancy. They carry a dismal prognosis, with a pooled local recurrence rate of 42%, distant metastasis rate of 27%, and mortality rate of 26%. The findings from this study suggest a trend favoring the use of GTR alone or STR with radiation therapy over STR alone. Mortality was similar regardless, which highlights the need for the development of effective treatment options to improve survival in patients with melanotic schwannomas.

General Versus Nongeneral Anesthesia for Spinal Surgery: A Comparative National Analysis of Reimbursement Trends Over 10 Years.

BACKGROUND AND OBJECTIVES: Nongeneral anesthesia (non-GA) spine surgery is growing in popularity and has facilitated earlier postoperative recovery, reduced cost, and fewer complications compared with spine surgery under general anesthesia (GA). Changes in reimbursement policies have been demonstrated to correlate with clinical practice; however, they have yet to be studied for GA vs non-GA spine procedures. We aimed to investigate trends in physician reimbursement for GA vs non-GA spine surgery in the United States. METHODS: We queried the ACS-NSQIP for GA and non-GA (regional, epidural, spinal, and anesthesia care/intravenous sedation) spine surgeries during 2011-2020. Work relative value units per operative hour (wRVUs/h) were retrieved for decompression or stabilization of the cervical, thoracic, and lumbar spine. Propensity score matching (1:1) was performed using all baseline variables. RESULTS: We included 474 706 patients who underwent spine decompression or stabilization procedures. GA was used in 472 248 operations, whereas 2458 operations were non-GA. The proportion of non-GA spine operations significantly increased during the study period. Operative times ( P < .001) and length of stays ( P < .001) were shorter in non-GA when compared with GA procedures. Non-GA lumbar procedures had significantly higher wRVUs/h when compared with the same procedures performed under GA (decompression; P < .001 and stabilization; P = .039). However, the same could not be said about cervicothoracic procedures. Lumbar decompression surgeries using non-GA witnessed significant yearly increase in wRVUs/h ( P < .01) contrary to GA ( P = .72). Physician reimbursement remained stable for procedures of the cervical or thoracic spine regardless of the anesthesia. CONCLUSION: Non-GA lumbar decompressions and stabilizations are associated with higher and increasing reimbursement trends (wRVUs/h) compared with those under GA. Reimbursement for cervical and thoracic surgeries was equal regardless of the type of anesthesia and being relatively stable during the study period. The adoption of a non-GA technique relative to the GA increased significantly during the study period.

Immunohistochemical markers predicting recurrence following resection and radiotherapy in chordoma patients: insights from a multicenter study [RETRACTED].

OBJECTIVE: Chordomas are rare tumors that often recur regardless of surgery with negative margins and postoperative radiotherapy. The predictive accuracy of widely used immunohistochemical (IHC) markers in addressing the recurrence of skull base chordomas (SBCs) is yet to be determined. This study aimed to investigate IHC markers in the prediction of recurrence after SBC resection with adjuvant radiation therapy. METHODS: The authors reviewed the records of patients who had treatment for SBC between January 2017 and June 2021 across the Mayo Clinic in Minnesota, Florida, and Arizona. Exclusion criteria included patients who had no histopathology or recurrence as an outcome. Histopathological markers included cytokeratin A1/A3 only, epithelial membrane antigen (EMA), S100 protein, pan-cytokeratin, IN1, GATA3, CAM5.2, OSCAR, and chondroid. Information from patient records was abstracted, including treatment, clinical and radiological follow-up duration, demographics, and histopathological factors. Decision tree and random forest classifiers were trained and tested to predict the recurrence based on unseen data using an 80/20 split. RESULTS: A total of 38 patients with a diagnosis of SBC who underwent resection (gross-total resection: 42.1%; and subtotal resection: 57.9%) and radiation therapy were extracted from the medical records. The mean patient age was 48.2 (SD 19.6) years; most patients were male (n = 23; 60.5%) and White (n = 36; 94.7%). Pan-cytokeratin was associated with an increased risk of postoperative recurrence (OR 14.67, 95% CI 2.44-88.13; p = 0.00517) after resection and adjuvant radiotherapy. The decision tree analysis found pan-cytokeratin-positive tumors to have a 78% chance of being classified as a recurrence, with an accuracy of 75%. The distribution of minimal depth in the prediction of postoperative recurrence indicates that the most important variables were pan-cytokeratin, followed by cytokeratin A1/A3 and EMA. CONCLUSIONS: The authors' machine learning algorithm identified pan-cytokeratin as the largest contributor to recurrence among other IHC markers after SBC resection. Machine learning may facilitate the prediction of outcomes in rare tumors, such as chordomas.

Association between immunohistochemical markers and tumor progression following resection of spinal chordomas: a multicenter study.

OBJECTIVE: Chordomas are slow-growing tumors derived from notochord remnants. Despite margin-negative excision and postoperative radiation therapy, spinal chordomas (SCs) often progress. The potential of immunohistochemical (IHC) markers, such as epithelial membrane antigen (EMA), combined with machine learning algorithms to predict long-term (≥ 12 months) postoperative tumor progression, has been understudied. The authors aimed to identify IHC markers using trained tree-based algorithms to predict long-term (≥ 12 months) postoperative tumor progression. METHODS: The authors reviewed the records of patients who underwent resection of SCs between January 2017 and June 2021 across the Mayo Clinic enterprise. Demographics, type of treatment, histopathology, and other relevant clinical factors were abstracted from each patient's record. Low tumor progression was defined as more than a 94.3-mm3 decrease in the tumor size at the latest radiographic follow-up. Decision trees and random forest classifiers were trained and tested to predict the long-term volumetric progression after an 80/20 data split. RESULTS: Sixty-two patients diagnosed with and surgically treated for SC were identified, of whom 31 were found to have a more advanced tumor progression based on the tumor volume change cutoff of 94.3 mm3. The mean age was 54.3 ± 13.8 years, and most patients were male (62.9%) and White (98.4%). The most common treatment modality was subtotal resection with radiation therapy (35.5%), with proton beam therapy being the most common (71%). Most SCs were sacrococcygeal (41.9%), followed by cervical (32.3%). EMA-positive SCs had a postoperative progression risk of 67%. Pancytokeratin-positive SCs had a progression rate of 67%; however, patients with S100 protein-positive SCs had a 54% risk of progression. The accuracy of this model in predicting the progression of unseen test data was 66%. Pancytokeratin (mean minimal depth = 1.57), EMA (mean minimal depth = 1.58), cytokeratin A1/A3 (mean minimal depth = 1.59), and S100 protein (mean minimal depth = 1.6) predicted the long-term volumetric progression. Multiway variable importance plots show the relative importance of the top 10 variables based on three measures of varying significance and their predictive role. CONCLUSIONS: These IHC variables with tree-based machine learning tools successfully demonstrate a high capacity to identify a patient's tumor progression pattern with an accuracy of 66%. Pancytokeratin, EMA, cytokeratin A1/A3, and S100 protein were the IHC drivers of a low tumor progression. This shows the power of machine learning algorithms in analyzing and predicting outcomes of rare conditions in a small sample size.

Using machine learning to predict 30-day readmission and reoperation following resection of supratentorial high-grade gliomas: an ACS NSQIP study involving 9418 patients.

OBJECTIVE: High-grade gliomas (HGGs) are among the rarest yet most aggressive tumor types in neurosurgical practice. In the current literature, few studies have assessed the drivers of early outcomes following resection of these tumors and investigated their association with quality of care. The authors aimed to identify the clinical predictors for 30-day readmission and reoperation following HGG surgery using the American College of Surgeons (ACS) National Surgical Quality Improvement Project (NSQIP) database and sought to create web-based applications predicting each outcome. METHODS: Using the ACS NSQIP database, the authors conducted a retrospective, multicenter cohort analysis of patients who underwent resection of supratentorial HGGs between January 1, 2016, and December 31, 2020. Demographics and comorbidities were extracted. The primary outcomes were 30-day unplanned readmission and reoperation. A stratified 80:20 split of the available data was carried out. Supervised machine learning algorithms were trained to predict 30-day outcomes. RESULTS: A total of 9418 patients were included in our cohort. The observed rate of unplanned readmission within 30 days of surgery was 13.0% (n = 1221). In terms of predictors, weight, chronic steroid use, preoperative blood urea nitrogen level, and white blood cell count were associated with a higher risk of readmission. The observed rate of unplanned reoperation within 30 days of surgery was 5.2% (n = 489). In terms of predictors, increased weight, longer operative time, and more days between hospital admission and operation were associated with an increased risk of early reoperation. The random forest algorithm showed the highest predictive performance for early readmission (area under the curve [AUC] = 0.967), while the XGBoost algorithm showed the highest predictive performance for early reoperation (AUC = 0.985). Web-based tools for both outcomes were deployed (https://glioma-readmission.herokuapp.com/ and https://glioma-reoperation.herokuapp.com/). CONCLUSIONS: In this study, the authors provide the first nationwide analysis for short-term outcomes in patients undergoing resection of supratentorial HGGs. Multiple patient, hospital, and admission factors were associated with readmission and reoperation, confirmed by machine learning predicting patients' prognosis, leading to better planning preoperatively and subsequently improved personalized patient care.

Immunohistochemical markers predicting long-term recurrence following clival and spinal chordoma resection: a multicenter study.

OBJECTIVE: Chordomas are rare tumors from notochordal remnants and account for 1%-4% of all primary bone malignancies, often arising from the clivus and sacrum. Despite margin-negative resection and postoperative radiotherapy, chordomas often recur. Further, immunohistochemical (IHC) markers have not been assessed as predictive of chordoma recurrence. The authors aimed to identify the IHC markers that are predictive of postoperative long-term (≥ 1 year) chordoma recurrence by using trained multiple tree-based machine learning (ML) algorithms. METHODS: The authors reviewed the records of patients who had undergone treatment for clival and spinal chordomas between January 2017 and June 2021 across the Mayo Clinic enterprise (Minnesota, Florida, and Arizona). Demographics, type of treatment, histopathology, and other relevant clinical factors were abstracted from each patient record. Decision tree and random forest classifiers were trained and tested to predict long-term recurrence based on unseen data using an 80/20 split. RESULTS: One hundred fifty-one patients diagnosed and treated for chordomas were identified: 58 chordomas of the clivus, 48 chordomas of the mobile spine, and 45 chordomas sacrococcygeal in origin. Patients diagnosed with cervical chordomas were the oldest among all groups (58 ± 14 years, p = 0.009). Most patients were male (n = 91, 60.3%) and White (n = 139, 92.1%). Most patients underwent resection with or without radiation therapy (n = 129, 85.4%). Subtotal resection followed by radiation therapy (n = 51, 33.8%) was the most common treatment modality, followed by gross-total resection then radiation therapy (n = 43, 28.5%). Multivariate analysis showed that S100 and pan-cytokeratin are more likely to predict the increase in the risk of postoperative recurrence (OR 3.67, 95% CI 1.09-12.42, p= 0.03; and OR 3.74, 95% CI 0.05-2.21, p = 0.02, respectively). In the decision tree analysis, a clinical follow-up > 1897 days was found in 37% of encounters and a 90% chance of being classified for recurrence (accuracy = 77%). Random forest analysis (n = 500 trees) showed that patient age, type of surgical treatment, location of tumor, S100, pan-cytokeratin, and EMA are the factors predicting long-term recurrence. CONCLUSIONS: The IHC and clinicopathological variables combined with tree-based ML tools successfully demonstrated a high capacity to identify recurrence patterns with an accuracy of 77%. S100, pan-cytokeratin, and EMA were the IHC drivers of recurrence. This shows the power of ML algorithms in analyzing and predicting outcomes of rare conditions of a small sample size.

Deep Learning Approaches for Glioblastoma Prognosis in Resource-Limited Settings: A Study Using Basic Patient Demographic, Clinical, and Surgical Inputs.

BACKGROUND: Glioblastoma (GBM) is the most common brain tumor in the United States, with an annual incidence rate of 3.21 per 100,000. It is the most aggressive type of diffuse glioma and has a median survival of months after treatment. This study aims to assess the accuracy of different novel deep learning models trained on a set of simple clinical, demographic, and surgical variables to assist in clinical practice, even in areas with constrained health care infrastructure. METHODS: Our study included 37,095 patients with GBM from the SEER (Surveillance Epidemiology and End Results) database. All predictors were based on demographic, clinicopathologic, and treatment information of the cases. Our outcomes of interest were months of survival and vital status. Concordance index (C-index) and integrated Brier scores (IBS) were used to evaluate the performance of the models. RESULTS: The patient characteristics and the statistical analyses were consistent with the epidemiologic literature. The models C-index and IBS ranged from 0.6743 to 0.6918 and from 0.0934 to 0.1034, respectively. Probabilistic matrix factorization (0.6918), multitask logistic regression (0.6916), and logistic hazard (0.6916) had the highest C-index scores. The models with the lowest IBS were the probabilistic matrix factorization (0.0934), multitask logistic regression (0.0935), and logistic hazard (0.0936). These models had an accuracy (1-IBS) of 90.66%; 90.65%, and 90.64%, respectively. The deep learning algorithms were deployed on an interactive Web-based tool for practical use available via https://glioblastoma-survanalysis.herokuapp.com/. CONCLUSIONS: Novel deep learning algorithms can better predict GBM prognosis than do baseline methods and can lead to more personalized patient care regardless of extensive electronic health record availability.