Keole, Sameer

AbstractBackgroundDeep learning auto‐segmentation (DLAS) models have been adopted in the clinic; however, they suffer from performance deterioration owing to the clinical practice variability. Some commercial DLAS software provide an incremental retraining function that enables users to train a custom model using their institutional data to account for clinical practice variability.PurposeThis study was performed to evaluate and implement the commercial DLAS software with the incremental retraining function for definitive treatment of patients with prostate cancer in a multi‐user environment.MethodsCT‐based target organs and organs‐at‐risk (OAR) delineation of 215 prostate cancer patients were utilized. The performance of three commercial DLAS software built‐in models was validated with 20 patients. A retrained custom model was developed using 100 patients and evaluated on the remaining data (n = 115). Dice similarity coefficient (DSC), Hausdorff distance (HD), mean surface distance (MSD), and surface DSC (SDSC) were utilized for quantitative evaluation. A multi‐rater qualitative evaluation was blindly performed with a five‐level scale. Visual inspection was performed in consensus and non‐consensus unacceptable cases to identify the failure modes.ResultsThree commercial DLAS vendor built‐in models achieved sub‐optimal performance in 20 patients. The retrained custom model had a mean DSC of 0.82 for prostate, 0.48 for seminal vesicles (SV), and 0.92 for rectum, respectively. This represents a significant improvement over the built‐in model with DSC of 0.73, 0.37, and 0.81 for the corresponding structures. Compared to the acceptance rate of 96.5% and consensus unacceptable rate (i.e., both reviewers rated as unacceptable) of 3.5% achieved by manual contours, the custom model achieved a 91.3% acceptance rate and 8.7% consensus unacceptable rate. The failure modes of retrained custom model were attributed to the following: cystogram (n = 2), hip prosthesis (n = 2), low dose rate brachytherapy seeds (n = 2), air in endorectal balloon(n = 1), non‐iodinated spacer (n = 2), and giant bladder(n = 1).ConclusionThe commercial DLAS software with the incremental retraining function was validated and clinically adopted for prostate patients in a multi‐user environment. AI‐based auto‐delineation of the prostate and OARs is shown to achieve improved physician acceptance, overall clinical utility, and accuracy.

Abstract Background Craniopharyngioma is a histologically benign tumor of the suprasellar region for which survival is excellent but quality of life is often poor secondary to functional deficits from tumor and treatment. Standard therapy consists of maximal safe resection with or without radiation therapy. Few prospective trials have been performed, and response assessment has not been standardized. Methods The Response Assessment in Pediatric Neuro-Oncology (RAPNO) committee devised consensus guidelines to assess craniopharyngioma response prospectively. Results Magnetic resonance imaging is the recommended radiologic modality for baseline and follow-up assessments. Radiologic response is defined by 2-dimensional measurements of both solid and cystic tumor components. In certain clinical contexts, response to solid and cystic disease may be differentially considered based on their unique natural histories and responses to treatment. Importantly, the committee incorporated functional endpoints related to neuro-endocrine and visual assessments into craniopharyngioma response definitions. In most circumstances, the cystic disease should be considered progressive only if growth is associated with acute, new-onset or progressive functional impairment. Conclusions Craniopharyngioma is a common pediatric central nervous system tumor for which standardized response parameters have not been defined. A RAPNO committee devised guidelines for craniopharyngioma assessment to uniformly define response in future prospective trials.

PURPOSE: Quality payment programs aim to adjust payments on the basis of quality and cost; however, few quality metrics exist in radiation oncology. This study evaluates and predicts the top spenders (TS) after radiation therapy (RT). MATERIALS AND METHODS: Patient characteristics, cancer details, treatments, toxicity, and survival data were collected for patients treated with RT at Mayo Clinic from 2007 to 2016. Standardized costs were obtained and adjusted for inflation. TSs were identified as those with greater than 93rd percentile costs (≥ $120,812). Prediction models were developed to predict TSs using training and validation sets using information available at consultation, after RT, and at last follow-up. RESULTS: A total of 15,131 patients were included and 1,065 TSs identified. Mean cost overall was $55,290 (median, $39,996) for all patients. Prediction models 1, 2, and 3 had concordance statistics of 0.83 to 0.83, 0.85 to 0.85, and 0.87 to 0.88, respectively in training and validation, indicating excellent prediction of TSs. Factors that were most predictive of TSs included stage N/A and stage 4 ( v stage 0; odds ratio [OR], 18.23 and 8.44, respectively; P < .001); hematologic, upper GI, skin and lung cancers ( v breast; OR, 11.45, 7.69, 3.81, and 2.43, respectively; P < .01); immunotherapy, surgery, and chemotherapy use (OR, 4.36, 2.51, and 1.61, respectively; P < .01); hospitalizations within 90 days of RT (OR, 2.26; P < .01); or death during the episode (OR, 1.56; P < .01). CONCLUSION: This is the first study of its kind to predict with high accuracy the highest spenders in radiation oncology. These patients may benefit from pre-emptive management to mitigate costs, or may require exclusion or adjustment from quality payment programs.