Combination of a Big Data Analytics Resource System With an Artificial Intelligence Algorithm to Identify Clinically Actionable Radiation Dose Thresholds for Dysphagia in Head and Neck Patients.

PURPOSE: We combined clinical practice changes, standardizations, and technology to automate aggregation, integration, and harmonization of comprehensive patient data from the multiple source systems used in clinical practice into a big data analytics resource system (BDARS). We then developed novel artificial intelligence algorithms, coupled with the BDARS, to identify structure dose volume histograms (DVH) metrics associated with dysphagia. METHODS AND MATERIALS: From the BDARS harmonized data of ≥22,000 patients, we identified 132 patients recently treated for head and neck cancer who also demonstrated dysphagia scores that worsened from base line to a maximum grade ≥2. We developed a method that used both physical and biologically corrected (α/ß = 2.5) DVH curves to test both absolute and percentage volume based DVH metrics. Combining a statistical categorization algorithm with machine learning (SCA-ML) provided more extensive detailing of response threshold evidence than either approach alone. A sensitivity guided, minimum input, machine learning (ML) model was iteratively constructed to identify the key structure DVH metric thresholds. RESULTS: Seven swallowing structures producing 738 candidate DVH metrics were ranked for association with dysphagia using SCA-ML scoring. Structures included superior pharyngeal constrictor (SPC), inferior pharyngeal constrictor (IPC), larynx, and esophagus. Bilateral parotid and submandibular gland (SG) structures were categorized by relative mean dose (eg, SG_high, SG_low) as a dose versus tumor centric analog to contra and ipsilateral designations. Structure DVH metrics with high SCA-ML scores included the following: SPC: D20% (equivalent dose [EQD2] Gy) ≥47.7; SPC: D25% (Gy) ≥50.4; IPC: D35% (Gy) ≥61.7; parotid_low: D60% (Gy) ≥13.2; and SG_high: D35% (Gy) ≥61.7. Larynx: D25% (Gy) ≥21.2 and SG_low: D45% ≥28.2 had high SCA-ML scores but were segmented on less than 90% of plans. A model based on SPC: D20% (EQD2 Gy) alone had sensitivity and area under the curve of 0.88 ± 0.13 and 0.74 ± 0.17, respectively. CONCLUSIONS: This study provides practical demonstration of combining big data with artificial intelligence to increase volume of evidence in clinical learning paradigms.

Predictive Models to Determine Clinically Relevant Deviations in Delivered Dose for Head and Neck Cancer.

PURPOSE: This study aimed to improve the understanding of deviations between planned and accumulated doses and to establish metrics to predict clinically significant dosimetric deviations midway through treatment to evaluate the potential need to re-plan during fractionated radiation therapy (RT). METHODS AND MATERIALS: A total of 100 patients with head and neck cancer were retrospectively evaluated. Contours were mapped from the planning computed tomography (CT) scan to each fraction cone beam CT via deformable image registration. The dose was calculated on each cone beam CT and evaluated based on the mapped contours. The mean dose at each fraction was averaged to approximate the accumulated dose for structures with mean dose constraints, and the daily maximum dose was summed to approximate the accumulated dose for structures with maximum dose constraints. A threshold deviation value was calculated to predict for patients needing midtreatment re-planning. This predictive model was applied to 52 patients treated at a separate institution. RESULTS: Dose was accumulated on 10 organs over 100 patients. To generate a threshold deviation that predicted the need to re-plan with 100% sensitivity, the submandibular glands required re-planning if the delivered dose was at least 3.5 Gy higher than planned by fraction 15. This model predicts the need to re-plan the submandibular glands with 98.7% specificity. In the independent evaluation cohort, this model predicts the need to re-plan the submandibular glands with 100% sensitivity and 98.0% specificity. The oral cavity, intermediate clinical target volume, left parotid, and inferior constrictor patient groups each had 1 patient who exceeded the threshold deviation by the end of RT. By fraction 15 of 30 to 35 total fractions, the left parotid gland, inferior constrictor, and intermediate clinical target volume had a dose deviation of 3.1 Gy, 5.9 Gy, and 4.8 Gy, respectively. When a deformable image registration failure was observed, the dose deviation exceeded the threshold for at least 1 organ, demonstrating that an automated deformable image registration-based dose assessment process could be developed with user evaluation for cases that result in dose deviations. CONCLUSIONS: A midtreatment threshold deviation was determined to predict the need to replan for the submandibular glands by fraction 15 of 30 to 35 total fractions of RT.

Incorporating big data into treatment plan evaluation: Development of statistical DVH metrics and visualization dashboards.

PURPOSE: To develop statistical dose-volume histogram (DVH)-based metrics and a visualization method to quantify the comparison of treatment plans with historical experience and among different institutions. METHODS AND MATERIALS: The descriptive statistical summary (ie, median, first and third quartiles, and 95% confidence intervals) of volume-normalized DVH curve sets of past experiences was visualized through the creation of statistical DVH plots. Detailed distribution parameters were calculated and stored in JavaScript Object Notation files to facilitate management, including transfer and potential multi-institutional comparisons. In the treatment plan evaluation, structure DVH curves were scored against computed statistical DVHs and weighted experience scores (WESs). Individual, clinically used, DVH-based metrics were integrated into a generalized evaluation metric (GEM) as a priority-weighted sum of normalized incomplete gamma functions. Historical treatment plans for 351 patients with head and neck cancer, 104 with prostate cancer who were treated with conventional fractionation, and 94 with liver cancer who were treated with stereotactic body radiation therapy were analyzed to demonstrate the usage of statistical DVH, WES, and GEM in a plan evaluation. A shareable dashboard plugin was created to display statistical DVHs and integrate GEM and WES scores into a clinical plan evaluation within the treatment planning system. Benchmarking with normal tissue complication probability scores was carried out to compare the behavior of GEM and WES scores. RESULTS: DVH curves from historical treatment plans were characterized and presented, with difficult-to-spare structures (ie, frequently compromised organs at risk) identified. Quantitative evaluations by GEM and/or WES compared favorably with the normal tissue complication probability Lyman-Kutcher-Burman model, transforming a set of discrete threshold-priority limits into a continuous model reflecting physician objectives and historical experience. CONCLUSIONS: Statistical DVH offers an easy-to-read, detailed, and comprehensive way to visualize the quantitative comparison with historical experiences and among institutions. WES and GEM metrics offer a flexible means of incorporating discrete threshold-prioritizations and historic context into a set of standardized scoring metrics. Together, they provide a practical approach for incorporating big data into clinical practice for treatment plan evaluations.

The big data effort in radiation oncology: Data mining or data farming?

Although large volumes of information are entered into our electronic health care records, radiation oncology information systems and treatment planning systems on a daily basis, the goal of extracting and using this big data has been slow to emerge. Development of strategies to meet this goal is aided by examining issues with a data farming instead of a data mining conceptualization. Using this model, a vision of key data elements, clinical process changes, technology issues and solutions, and role for professional societies is presented. With a better view of technology, process and standardization factors, definition and prioritization of efforts can be more effectively directed.

Whisker spot patterns: a noninvasive method of individual identification of Australian sea lions (Neophoca cinerea).

Reliable methods for identification of individual animals are advantageous for ecological studies of population demographics and movement patterns. Photographic identification, based on distinguishable patterns, unique shapes, or scars, is an effective technique already used for many species. We tested whether photographs of whisker spot patterns could be used to discriminate among individual Australian sea lion (Neophoca cinerea). Based on images of 53 sea lions, we simulated 5,000 patterns before calculating the probability of duplication in a study population. A total of 99% (± 1.5 SD) of patterns were considered reliable for a population of 50, 98% (± 1.7 SD) for 100, 92% (± 4.7 SD) for 500, and 88% (± 5.7 SD) for 1,000. We tested a semiautomatic approach by matching 16 known individuals at 3 different angles (70°, 90°, and 110°), 2 distances (1 and 2 m), and 6 separate times over a 1-year period. A point-pattern matching algorithm for pairwise comparisons produced 90% correct matches of photographs taken on the same day at 90°. Images of individuals at 1 and 2 m resulted in 89% correct matches, those photographed at different angles and different times (at 90°) resulted in 48% and 73% correct matches, respectively. Our results show that the Chamfer distance transform can effectively be used for individual identification, but only if there is very little variation in photograph angle. This point-pattern recognition application may also work for other otariid species.

Ecological and mutation-order speciation in digital organisms.

Reproductive isolation between populations often evolves as a by-product of independent adaptation to new environments, but the selective pressures of these environments may be divergent ("ecological speciation" or uniform ("mutation-order speciation." In this study, we use an artificial life platform to directly compare the strength of reproductive isolation (specifically, postzygotic) generated by ecological and mutation-order processes. We also tested the effect of gene flow as well as the dimensionality (i.e., number of selective pressures) of the environments on the strength of postzygotic isolation. We found that ecological speciation generally formed stronger isolation than mutation-order speciation, mutation-order speciation was more sensitive to gene flow than ecological speciation, and environments with high dimensionality formed stronger reproductive isolation than those with low dimensionality. How various factors affect the strength of reproductive isolation has been difficult to test in biological organisms, but the use of artificial life, which provides its own genetic system that evolves, allowed us to computationally test the effect of these factors more easily.