Jeremy Thomas Gaskins
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Jeremy Thomas Gaskins has a total of 135 co-authors. Here are the top co-authors by number of shared publications:
- Scott R. Silva
Affiliation Department of Radiation Oncology, Brown Cancer Center, University of Louisville School of Medicine, Louisville, Kentucky, United States of America
5 shared publications Adrianna Masters
Affiliation Department of Radiation Oncology, Brown Cancer Center, University of Louisville School of Medicine, Louisville, Kentucky, United States of America
3 shared publications- Andres Portocarrero-Bonifaz
Affiliations Department of Radiation Oncology, Brown Cancer Center, University of Louisville School of Medicine, Louisville, Kentucky, United States of America, Department of Radiation Oncology, Mayo Clinic, Jacksonville, Florida, United States of America
3 shared publications Bryce M. Forry
Affiliation Department of Radiation Oncology, Brown Cancer Center, University of Louisville School of Medicine, Louisville, Kentucky, United States of America
3 shared publications- Dan O. Popa
Louisville Automation Research and Robotics Institute (LARRI), University of Louisville, Louisville, KY 40208, USA
3 shared publications - Grant McKenzie
University of Louisville School of Medicine
3 shared publications
Latest Publications
- Open Access
- 4 November 2026
- Published by MDPI AG in Preprints.org
Acute Symptoms of Tear Gas Exposure and Adverse Reproductive Health Outcomes in Males
Exposure to tear gas, a chemical agent used primarily to disperse crowds, results in acute adverse effects on health which are well-documented but may have longer-term impacts on health. This study investigated the association between tear gas exposure during protests in 2020 and male reproductive health outcomes. Data were collected through an anonymous online questionnaire in 2020–2021 from adults exposed to tear gas. Exposure was assessed using reported acute symptoms affecting the eyes, lungs, skin, and heart. A symptom-based exposure score ranging from 0–14 was categorized as low, medium, or high. Male participants reported whether they experienced erectile dysfunction, ejaculatory dysfunction, or difficulty conceiving following exposure. Logistic regression was used to estimate adjusted odds ratios (OR) and 95% confidence intervals (CI). Among 92 male respondents, approximately 46% reported at least one reproductive issue, including 28 with erectile dysfunction and 26 with ejaculatory dysfunction. High tear gas exposure was significantly associated with erectile dysfunction compared with low exposure (adjusted OR=6.57; 95% CI: 1.28–40.08). No significant association was observed for ejaculatory dysfunction (adjusted OR=1.96; 95% CI: 0.35–11.36). These findings suggest potential reproductive health effects of tear gas exposure and may help to guide public policy and education.
- 4 November 2026
- Published by MDPI AG in Preprints.org
Open AccessAcute Symptoms of Tear Gas Exposure and Adverse Reproductive Health Outcomes in Males
Exposure to tear gas, a chemical agent used primarily to disperse crowds, results in acute adverse effects on health which are well-documented but may have longer-term impacts on health. This study investigated the association between tear gas exposure during protests in 2020 and male reproductive health outcomes. Data were collected through an anonymous online questionnaire in 2020–2021 from adults exposed to tear gas. Exposure was assessed using reported acute symptoms affecting the eyes, lungs, skin, and heart. A symptom-based exposure score ranging from 0–14 was categorized as low, medium, or high. Male participants reported whether they experienced erectile dysfunction, ejaculatory dysfunction, or difficulty conceiving following exposure. Logistic regression was used to estimate adjusted odds ratios (OR) and 95% confidence intervals (CI). Among 92 male respondents, approximately 46% reported at least one reproductive issue, including 28 with erectile dysfunction and 26 with ejaculatory dysfunction. High tear gas exposure was significantly associated with erectile dysfunction compared with low exposure (adjusted OR=6.57; 95% CI: 1.28–40.08). No significant association was observed for ejaculatory dysfunction (adjusted OR=1.96; 95% CI: 0.35–11.36). These findings suggest potential reproductive health effects of tear gas exposure and may help to guide public policy and education.
- Open Access
- 3 October 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Human-Robot Interaction
A Pilot Study on Facial Expression Recognition by Autistic Children of a Social Robot
This study investigates the accuracy of recognition of four basic facial expressions (Angry, Happy, Sad, and Surprised) on a social assistive robot's face by children with and without autism spectrum disorder (ASD). Human-robot interaction experiments (HRI) were conducted with Milo, an expressive social robot. A total of 60 children aged 8–12 took part in the study: 41 non-autistic children and 19 children who met DSM-5 diagnostic criteria for ASD. Milo displayed four basic facial expressions in a randomized order across two consecutive attempts. Participants identified the expressions, allowing for a quantitative analysis of potential differences between groups.The proportion of correct answers was used for validity and reliability ratings. A logistic mixed-effect regression model was employed to analyze the relationship between accuracy, demographic variables, emotion type, and ASD status. Our results suggest that although overall differences in facial expression recognition between autistic and non-autistic participants may be subtle, more detailed emotion-specific analyses are necessary to identify and characterize these differences. Results indicated that overall group differences in labeling accuracy were modest, whereas emotion-specific analyses identified a significant group difference for Milo's Angry facial expression. The number of repeated attempts showed a marginally significant effect in overall accuracy in both groups, with the recognition accuracy decreasing on the second attempt, particularly among autistic children. However, this decline was sensitive to the treatment of “None of the above” responses. Therefore, the attempt effect should be interpreted cautiously. These findings highlight the importance of considering specific emotions, task structure, and response options when evaluating children's interpretations of robot-generated facial expressions.
- 3 October 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Human-Robot Interaction
Open AccessA Pilot Study on Facial Expression Recognition by Autistic Children of a Social Robot
This study investigates the accuracy of recognition of four basic facial expressions (Angry, Happy, Sad, and Surprised) on a social assistive robot's face by children with and without autism spectrum disorder (ASD). Human-robot interaction experiments (HRI) were conducted with Milo, an expressive social robot. A total of 60 children aged 8–12 took part in the study: 41 non-autistic children and 19 children who met DSM-5 diagnostic criteria for ASD. Milo displayed four basic facial expressions in a randomized order across two consecutive attempts. Participants identified the expressions, allowing for a quantitative analysis of potential differences between groups.The proportion of correct answers was used for validity and reliability ratings. A logistic mixed-effect regression model was employed to analyze the relationship between accuracy, demographic variables, emotion type, and ASD status. Our results suggest that although overall differences in facial expression recognition between autistic and non-autistic participants may be subtle, more detailed emotion-specific analyses are necessary to identify and characterize these differences. Results indicated that overall group differences in labeling accuracy were modest, whereas emotion-specific analyses identified a significant group difference for Milo's Angry facial expression. The number of repeated attempts showed a marginally significant effect in overall accuracy in both groups, with the recognition accuracy decreasing on the second attempt, particularly among autistic children. However, this decline was sensitive to the treatment of “None of the above” responses. Therefore, the attempt effect should be interpreted cautiously. These findings highlight the importance of considering specific emotions, task structure, and response options when evaluating children's interpretations of robot-generated facial expressions.
- Open Access
- 22 May 2026
- Published by MDPI AG in Preprints.org
Protest Attendance and Tear Gas Exposure: Associations with Menstrual Function
Background: During the racial justice protests of 2020 and 2021, crowd control chemical irritants (referred to as “tear gas”) were deployed against protesters, after which there were anecdotal reports of altered menstrual cycles. Methods: Data from 82 women (ages 18-45) who attended protests/demonstrations in 2020-2021 were collected through an online questionnaire. The associations between proxy measurements of tear gas exposure (number of protests attended, number of acute symptoms of exposure, and number of tear gas exposures) and menstrual cycle outcomes were determined through linear regression, adjusted for potential confounders. Menstrual cycle outcomes were light bleeding, heavy bleeding, period pains, irregular cycles, long bleeding, short bleeding, short cycles, or long cycles. Results: High protest attendance (> 9 times) (β: 2.60, 95% CI: 1.41, 3.80), and exposure to tear gas at least three times (β: 2.18, 95% CI: 0.63, 3.72) were significantly associated with a greater number of menstrual cycle outcomes. The strongest association was observed for irregular cycles. The number of acute symptoms was not significantly associated with menstrual cycle outcomes. Conclusions: Our findings support the hypothesis that tear gas exposure causes menstrual cycle disturbances. Additional research is needed to determine whether there are longer-term menstrual cycle disturbances or other adverse effects on reproductive health.
- 22 May 2026
- Published by MDPI AG in Preprints.org
Open AccessProtest Attendance and Tear Gas Exposure: Associations with Menstrual Function
Background: During the racial justice protests of 2020 and 2021, crowd control chemical irritants (referred to as “tear gas”) were deployed against protesters, after which there were anecdotal reports of altered menstrual cycles. Methods: Data from 82 women (ages 18-45) who attended protests/demonstrations in 2020-2021 were collected through an online questionnaire. The associations between proxy measurements of tear gas exposure (number of protests attended, number of acute symptoms of exposure, and number of tear gas exposures) and menstrual cycle outcomes were determined through linear regression, adjusted for potential confounders. Menstrual cycle outcomes were light bleeding, heavy bleeding, period pains, irregular cycles, long bleeding, short bleeding, short cycles, or long cycles. Results: High protest attendance (> 9 times) (β: 2.60, 95% CI: 1.41, 3.80), and exposure to tear gas at least three times (β: 2.18, 95% CI: 0.63, 3.72) were significantly associated with a greater number of menstrual cycle outcomes. The strongest association was observed for irregular cycles. The number of acute symptoms was not significantly associated with menstrual cycle outcomes. Conclusions: Our findings support the hypothesis that tear gas exposure causes menstrual cycle disturbances. Additional research is needed to determine whether there are longer-term menstrual cycle disturbances or other adverse effects on reproductive health.
- Open Access
- 11 May 2026
- Published by Wiley in International Journal of Gynecology & Obstetrics
- Vol. 175 (2) , 741-748
Rate of cesarean delivery among patients with class III obesity at term based on body mass index: A retrospective cohort study
Objectives: This study investigates the risk of cesarean delivery based on body mass index (BMI) in patients with class III obesity who undergo induction of labor. Method: This was a retrospective cohort study set in a university hospital. Study participants included patients presenting for delivery at term with class III obesity at the time of admission. A retrospective chart review was conducted of patients with a pre‐delivery BMI ≥40 who delivered between May 2016 and December 2020 at a single institution. Patients were divided into the following BMI groups: 40–44.9, 45–49.99, 50–59.99, ≥60. Demographic and other clinical features were analyzed using analysis of variance and Fisher tests across BMI groups. Rates of labor induction and cesarean delivery (CD) were further analyzed by logistic regression. Results: We identified 1010 patients with pre‐pregnancy BMI ≥40 who delivered in the study period, of whom 497 met the study criteria. The median BMI in this population was 43.9 (interquartile range: 41.8–48.5, range 40.0–92.9). Twenty‐seven percent of patients who underwent an induction of labor (IOL) delivered via CD compared to 12% for patients who presented in spontaneous labor. There were higher rates of induction seen in patients as BMI increased, but this effect was not seen in patients with BMI 60 or higher. The rate of CD was increased in patients who underwent IOL ( P = 0.0332). This finding was only seen for those undergoing an elective IOL ( P = 0.0088) but not a medically indicated IOL ( P = 0.3588). Conclusion: Rates of induction of labor are higher in patients with higher BMI. In our cohort, those who underwent elective IOL were more likely to deliver via CD. Further study is needed to evaluate outcomes for patients with a BMI of 60 and greater, and more clarity is needed around recommendations for timing of elective induction of labor in patients with class III obesity.
- 11 May 2026
- Published by Wiley in International Journal of Gynecology & Obstetrics
- Vol. 175 (2) , 741-748
Open AccessRate of cesarean delivery among patients with class III obesity at term based on body mass index: A retrospective cohort study
Objectives: This study investigates the risk of cesarean delivery based on body mass index (BMI) in patients with class III obesity who undergo induction of labor. Method: This was a retrospective cohort study set in a university hospital. Study participants included patients presenting for delivery at term with class III obesity at the time of admission. A retrospective chart review was conducted of patients with a pre‐delivery BMI ≥40 who delivered between May 2016 and December 2020 at a single institution. Patients were divided into the following BMI groups: 40–44.9, 45–49.99, 50–59.99, ≥60. Demographic and other clinical features were analyzed using analysis of variance and Fisher tests across BMI groups. Rates of labor induction and cesarean delivery (CD) were further analyzed by logistic regression. Results: We identified 1010 patients with pre‐pregnancy BMI ≥40 who delivered in the study period, of whom 497 met the study criteria. The median BMI in this population was 43.9 (interquartile range: 41.8–48.5, range 40.0–92.9). Twenty‐seven percent of patients who underwent an induction of labor (IOL) delivered via CD compared to 12% for patients who presented in spontaneous labor. There were higher rates of induction seen in patients as BMI increased, but this effect was not seen in patients with BMI 60 or higher. The rate of CD was increased in patients who underwent IOL ( P = 0.0332). This finding was only seen for those undergoing an elective IOL ( P = 0.0088) but not a medically indicated IOL ( P = 0.3588). Conclusion: Rates of induction of labor are higher in patients with higher BMI. In our cohort, those who underwent elective IOL were more likely to deliver via CD. Further study is needed to evaluate outcomes for patients with a BMI of 60 and greater, and more clarity is needed around recommendations for timing of elective induction of labor in patients with class III obesity.
- 31 January 2026
- Published by Elsevier in Journal of Minimally Invasive Gynecology
- Vol. 33 (6) , 744-751.e1
Perioperative Outcomes of Major Laparoscopic Surgery for Benign Gynecologic Conditions and the Role of Body Mass Index
- 31 January 2026
- Published by Elsevier in Journal of Minimally Invasive Gynecology
- Vol. 33 (6) , 744-751.e1
Perioperative Outcomes of Major Laparoscopic Surgery for Benign Gynecologic Conditions and the Role of Body Mass Index
- Open Access
Predictor-Informed Bayesian Nonparametric Clustering
In this project we are interested in performing clustering of observations such that the cluster membership is influenced by a set of predictors. To that end, we employ the Bayesian nonparameteric Common Atoms Model, which is a nested clustering algorithm that utilizes a (fixed) group membership for each observation to encourage more similar clustering of members of the same group. CAM operates by assuming each group has its own vector of cluster probabilities, which are themselves clustered to allow similar clustering for some groups. We extend this approach by treating the group membership as an unknown latent variable determined as a flexible nonparametric form of the covariate vector. Consequently, observations with similar predictor values will be in the same latent group and are more likely to be clustered together than observations with disparate predictors. We propose a pyramid group model that flexibly partitions the predictor space into these latent group memberships. This pyramid model operates similarly to a Bayesian regression tree process except that it uses the same splitting rule for at all nodes at the same tree depth which facilitates improved mixing. We outline a block Gibbs sampler to perform posterior inference from our model. Our methodology is demonstrated in simulation and real data examples. In the real data application, we utilize the RAND Health and Retirement Study to cluster and predict patient outcomes in terms of the number of overnight hospital stays.
Open AccessPredictor-Informed Bayesian Nonparametric Clustering
In this project we are interested in performing clustering of observations such that the cluster membership is influenced by a set of predictors. To that end, we employ the Bayesian nonparameteric Common Atoms Model, which is a nested clustering algorithm that utilizes a (fixed) group membership for each observation to encourage more similar clustering of members of the same group. CAM operates by assuming each group has its own vector of cluster probabilities, which are themselves clustered to allow similar clustering for some groups. We extend this approach by treating the group membership as an unknown latent variable determined as a flexible nonparametric form of the covariate vector. Consequently, observations with similar predictor values will be in the same latent group and are more likely to be clustered together than observations with disparate predictors. We propose a pyramid group model that flexibly partitions the predictor space into these latent group memberships. This pyramid model operates similarly to a Bayesian regression tree process except that it uses the same splitting rule for at all nodes at the same tree depth which facilitates improved mixing. We outline a block Gibbs sampler to perform posterior inference from our model. Our methodology is demonstrated in simulation and real data examples. In the real data application, we utilize the RAND Health and Retirement Study to cluster and predict patient outcomes in terms of the number of overnight hospital stays.
- Open Access
Acceptance of an Adaptive Robotic Nursing Assistant for Ambulation Tasks
The effective use of nursing assistant robots requires an understanding of key acceptance factors. The study examined the differences in attitudes among 58 nursing students while performing ambulation tasks with and without an Adaptive Robotic Nursing Assistant (ARNA) robot. An ARNA is driven by tactile cues from the patient through a force–torque-measuring handlebar, whose signals are fed into a neuro-adaptive controller to achieve a specific admittance behavior regardless of patient strength, weight, or floor incline. Ambulation tasks used two fall-prevention devices: a gait belt and a full-body harness. The attitude toward the robot included perceived satisfaction, usefulness, and assistance, replacing the perceived ease-of-use construct found in the standard technology acceptance model. The effects of external demographic variables on those constructs were also analyzed. The modified technology acceptance model was validated with the simultaneous estimation of the effects of perceived usefulness and assistance on satisfaction. Our analysis employed an integrated hierarchical linear mixed-effects regression model to analyze the complex relationships between model variables. Our results suggest that nursing students rated the ARNA’s performance higher across all model constructs compared to a human assistant. Furthermore, male subjects rated the perceived usefulness of the robot higher than female subjects.
Open AccessAcceptance of an Adaptive Robotic Nursing Assistant for Ambulation Tasks
The effective use of nursing assistant robots requires an understanding of key acceptance factors. The study examined the differences in attitudes among 58 nursing students while performing ambulation tasks with and without an Adaptive Robotic Nursing Assistant (ARNA) robot. An ARNA is driven by tactile cues from the patient through a force–torque-measuring handlebar, whose signals are fed into a neuro-adaptive controller to achieve a specific admittance behavior regardless of patient strength, weight, or floor incline. Ambulation tasks used two fall-prevention devices: a gait belt and a full-body harness. The attitude toward the robot included perceived satisfaction, usefulness, and assistance, replacing the perceived ease-of-use construct found in the standard technology acceptance model. The effects of external demographic variables on those constructs were also analyzed. The modified technology acceptance model was validated with the simultaneous estimation of the effects of perceived usefulness and assistance on satisfaction. Our analysis employed an integrated hierarchical linear mixed-effects regression model to analyze the complex relationships between model variables. Our results suggest that nursing students rated the ARNA’s performance higher across all model constructs compared to a human assistant. Furthermore, male subjects rated the perceived usefulness of the robot higher than female subjects.
- Open Access
- 7 August 2025
- Published by MDPI AG in Preprints.org
Acceptance of an Adaptive Nursing Assistant Robot for Ambulation Tasks
Effective use of nursing assistant robots requires understanding of key acceptance factors. The study examined the differences in attitudes among 58 nursing students while performing the ambulation tasks with and without an Adaptive Robot Nursing Assistant (ARNA) robot. ARNA is driven by tactile cues from the patient through a force-torque measuring handlebar, whose signals are fed into a neuro-adaptive controller to achieve a specific admittance behavior regardless of patient strength, weight, or floor incline. Ambulation tasks used two fall prevention devices: a gait belt and a full-body harness. The attitude toward the robot included perceived satisfaction, usefulness, and assistance, replacing the perceived ease of use construct found in the standard Technology Acceptance Model. The effects of external demographic variables on those constructs were also analyzed. The technology acceptance model was validated with the simultaneous estimation of the effects of perceived usefulness and assistance on the satisfaction employing an integrated hierarchical linear mixed effect regression model to analyze complex relationships between the model variables. The results suggest that students rated ARNA’s performance higher across all constructs compared to that of a human assistant. Male subjects rated perceived usefulness of the robot higher than females in a reduced model.
- 7 August 2025
- Published by MDPI AG in Preprints.org
Open AccessAcceptance of an Adaptive Nursing Assistant Robot for Ambulation Tasks
Effective use of nursing assistant robots requires understanding of key acceptance factors. The study examined the differences in attitudes among 58 nursing students while performing the ambulation tasks with and without an Adaptive Robot Nursing Assistant (ARNA) robot. ARNA is driven by tactile cues from the patient through a force-torque measuring handlebar, whose signals are fed into a neuro-adaptive controller to achieve a specific admittance behavior regardless of patient strength, weight, or floor incline. Ambulation tasks used two fall prevention devices: a gait belt and a full-body harness. The attitude toward the robot included perceived satisfaction, usefulness, and assistance, replacing the perceived ease of use construct found in the standard Technology Acceptance Model. The effects of external demographic variables on those constructs were also analyzed. The technology acceptance model was validated with the simultaneous estimation of the effects of perceived usefulness and assistance on the satisfaction employing an integrated hierarchical linear mixed effect regression model to analyze complex relationships between the model variables. The results suggest that students rated ARNA’s performance higher across all constructs compared to that of a human assistant. Male subjects rated perceived usefulness of the robot higher than females in a reduced model.
- Open Access
- 14 May 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (5) , e0312208
Advancing patient care: Machine learning models for predicting grade 3+ toxicities in gynecologic cancer patients treated with HDR brachytherapy
Background: Gynecological cancers are among the most prevalent cancers in women worldwide. Brachytherapy, often used as a boost to external beam radiotherapy, is integral to treatment. Advances in computation, algorithms, and data availability have popularized the use of machine learning to predict patient outcomes. Recent studies have applied models such as logistic regression, support vector machines, and deep learning networks to predict specific toxicities in patients who have undergone brachytherapy. Objective: To develop and compare machine learning models for predicting grade 3 or higher toxicities in gynecological cancer patients treated with high dose rate (HDR) brachytherapy, aiming to contribute to personalized radiation treatments. Methods: A retrospective analysis was performed on gynecological cancer patients who underwent HDR brachytherapy with Syed-Neblett or Tandem and Ovoid applicators from 2009 to 2023. After applying exclusion criteria, 233 patients were included in the analysis. Dosimetric variables for the high-risk clinical target volume (HR-CTV) and organs at risk, along with tumor, patient, and toxicity data, were collected and compared between groups with and without grade 3 or higher toxicities using statistical tests. Seven supervised classification machine learning models (Logistic Regression, Random Forest, K-Nearest Neighbors, Support Vector Machines, Gaussian Naive Bayes, Multi-Layer Perceptron Neural Networks, and XGBoost) were constructed and evaluated. The training process involved sequential feature selection (SFS) when appropriate, followed by hyperparameter tuning. Final model performance was characterized using a 25% withheld test dataset. Results: The top three ranking models were Support Vector Machines, Random Forest, and Logistic Regression, with F1 testing scores of 0.63, 0.57, and 0.52; normMCC testing scores of 0.75, 0.77, and 0.71; and accuracy testing scores of 0.80, 0.85, and 0.81, respectively. The SFS algorithm selected 10 features for the highest-ranking model. In traditional statistical analysis, HR-CTV volume, Charlson Comorbidity Index, Length of Follow-Up, and D2cc - Rectum differed significantly between groups with and without grade 3 or higher toxicities. Conclusions: Machine learning models were developed to predict grade 3 or higher toxicities, achieving satisfactory performance. Machine learning presents a novel solution to creating multivariable models for personalized radiation therapy.
- 14 May 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (5) , e0312208
Open AccessAdvancing patient care: Machine learning models for predicting grade 3+ toxicities in gynecologic cancer patients treated with HDR brachytherapy
Background: Gynecological cancers are among the most prevalent cancers in women worldwide. Brachytherapy, often used as a boost to external beam radiotherapy, is integral to treatment. Advances in computation, algorithms, and data availability have popularized the use of machine learning to predict patient outcomes. Recent studies have applied models such as logistic regression, support vector machines, and deep learning networks to predict specific toxicities in patients who have undergone brachytherapy. Objective: To develop and compare machine learning models for predicting grade 3 or higher toxicities in gynecological cancer patients treated with high dose rate (HDR) brachytherapy, aiming to contribute to personalized radiation treatments. Methods: A retrospective analysis was performed on gynecological cancer patients who underwent HDR brachytherapy with Syed-Neblett or Tandem and Ovoid applicators from 2009 to 2023. After applying exclusion criteria, 233 patients were included in the analysis. Dosimetric variables for the high-risk clinical target volume (HR-CTV) and organs at risk, along with tumor, patient, and toxicity data, were collected and compared between groups with and without grade 3 or higher toxicities using statistical tests. Seven supervised classification machine learning models (Logistic Regression, Random Forest, K-Nearest Neighbors, Support Vector Machines, Gaussian Naive Bayes, Multi-Layer Perceptron Neural Networks, and XGBoost) were constructed and evaluated. The training process involved sequential feature selection (SFS) when appropriate, followed by hyperparameter tuning. Final model performance was characterized using a 25% withheld test dataset. Results: The top three ranking models were Support Vector Machines, Random Forest, and Logistic Regression, with F1 testing scores of 0.63, 0.57, and 0.52; normMCC testing scores of 0.75, 0.77, and 0.71; and accuracy testing scores of 0.80, 0.85, and 0.81, respectively. The SFS algorithm selected 10 features for the highest-ranking model. In traditional statistical analysis, HR-CTV volume, Charlson Comorbidity Index, Length of Follow-Up, and D2cc - Rectum differed significantly between groups with and without grade 3 or higher toxicities. Conclusions: Machine learning models were developed to predict grade 3 or higher toxicities, achieving satisfactory performance. Machine learning presents a novel solution to creating multivariable models for personalized radiation therapy.