Background: Psychological test reports are essential in assessing intellectual functioning, aiding in diagnosing and treating intellectual disability (ID) and attention-deficit/ hyperactivity disorder (ADHD). However, these reports can have several problems because they are diverse, unstructured, subjective, and involve human errors. Additionally, physicians often do not read the entire report, and the number of reports is lower than that of diagnoses. Methods: We developed explainable predictive models for classifying IDs and ADHDs based on written reports to address these issues. The reports of 1,475 patients with IDs and ADHDs who underwent intelligence tests were used for the models. These models were developed by analyzing reports using natural language processing (NLP) and incorporating the physician’s diagnosis for each report. We selected n-gram features from the models’ results by extracting important features using SHapley Additive exPlanations and permutation importance to make the models explainable. Developing the n-gram feature-based original text search system compensated for the lack of human readability caused by NLP and enabled the reconstruction of human-readable texts from the selected n-gram features. Results: The maximum model accuracy was 0.92, and the 80 human-readable texts were restored from four models. Conclusion The results showed that the models could accurately classify IDs and ADHDs, even with a few reports. The models were also able to explain their predictions. The explainability-enhanced model can help physicians understand the classification process of IDs and ADHDs and provide evidence-based insights.

Background: Psychological test reports are essential in assessing intellectual functioning, aiding in diagnosing and treating intellectual disability (ID) and attention-deficit/ hyperactivity disorder (ADHD). However, these reports can have several problems because they are diverse, unstructured, subjective, and involve human errors. Additionally, physicians often do not read the entire report, and the number of reports is lower than that of diagnoses. Methods: We developed explainable predictive models for classifying IDs and ADHDs based on written reports to address these issues. The reports of 1,475 patients with IDs and ADHDs who underwent intelligence tests were used for the models. These models were developed by analyzing reports using natural language processing (NLP) and incorporating the physician’s diagnosis for each report. We selected n-gram features from the models’ results by extracting important features using SHapley Additive exPlanations and permutation importance to make the models explainable. Developing the n-gram feature-based original text search system compensated for the lack of human readability caused by NLP and enabled the reconstruction of human-readable texts from the selected n-gram features. Results: The maximum model accuracy was 0.92, and the 80 human-readable texts were restored from four models. Conclusion The results showed that the models could accurately classify IDs and ADHDs, even with a few reports. The models were also able to explain their predictions. The explainability-enhanced model can help physicians understand the classification process of IDs and ADHDs and provide evidence-based insights.

Background: Psychological test reports are essential in assessing intellectual functioning, aiding in diagnosing and treating intellectual disability (ID) and attention-deficit/ hyperactivity disorder (ADHD). However, these reports can have several problems because they are diverse, unstructured, subjective, and involve human errors. Additionally, physicians often do not read the entire report, and the number of reports is lower than that of diagnoses. Methods: We developed explainable predictive models for classifying IDs and ADHDs based on written reports to address these issues. The reports of 1,475 patients with IDs and ADHDs who underwent intelligence tests were used for the models. These models were developed by analyzing reports using natural language processing (NLP) and incorporating the physician’s diagnosis for each report. We selected n-gram features from the models’ results by extracting important features using SHapley Additive exPlanations and permutation importance to make the models explainable. Developing the n-gram feature-based original text search system compensated for the lack of human readability caused by NLP and enabled the reconstruction of human-readable texts from the selected n-gram features. Results: The maximum model accuracy was 0.92, and the 80 human-readable texts were restored from four models. Conclusion The results showed that the models could accurately classify IDs and ADHDs, even with a few reports. The models were also able to explain their predictions. The explainability-enhanced model can help physicians understand the classification process of IDs and ADHDs and provide evidence-based insights.

Background: Psychological test reports are essential in assessing intellectual functioning, aiding in diagnosing and treating intellectual disability (ID) and attention-deficit/ hyperactivity disorder (ADHD). However, these reports can have several problems because they are diverse, unstructured, subjective, and involve human errors. Additionally, physicians often do not read the entire report, and the number of reports is lower than that of diagnoses. Methods: We developed explainable predictive models for classifying IDs and ADHDs based on written reports to address these issues. The reports of 1,475 patients with IDs and ADHDs who underwent intelligence tests were used for the models. These models were developed by analyzing reports using natural language processing (NLP) and incorporating the physician’s diagnosis for each report. We selected n-gram features from the models’ results by extracting important features using SHapley Additive exPlanations and permutation importance to make the models explainable. Developing the n-gram feature-based original text search system compensated for the lack of human readability caused by NLP and enabled the reconstruction of human-readable texts from the selected n-gram features. Results: The maximum model accuracy was 0.92, and the 80 human-readable texts were restored from four models. Conclusion The results showed that the models could accurately classify IDs and ADHDs, even with a few reports. The models were also able to explain their predictions. The explainability-enhanced model can help physicians understand the classification process of IDs and ADHDs and provide evidence-based insights.

Background: The diagnosis of major depressive disorder (MDD) relies primarily on clinical interviews, which can be subjective and time consuming. Thus, there is a need for more objective diagnostic tools. The aim of this study was to develop an artificial intelligence (AI) application that predicts the antidepressant drug response of individual patients with MDD based on longitudinal data. Methods: Longitudinal data from patient records, including sex, age, outpatient or inpatient status, medication type and dosage, and the Hamilton Depression Rating Scale (HAMD) scores, were used to train the Transformer model and the 1-dimensional convolutional neural network model. Individual patient records were allocated to training (80%), validation (10%), and testing (10%) datasets. Results: The AI model demonstrated 88% sensitivity and 92% specificity for predicting the treatment response. Significant factors independently associated with the antidepressant response included age, sex, history of depression, and baseline HAMD scores. Conclusion This AI-driven software application provides a clinically valuable tool for predicting treatment response.While promising, further research is needed to incorporate voice data into the AI model using the voice recording feature to further improve diagnostic accuracy.

Background: The diagnosis of major depressive disorder (MDD) relies primarily on clinical interviews, which can be subjective and time consuming. Thus, there is a need for more objective diagnostic tools. The aim of this study was to develop an artificial intelligence (AI) application that predicts the antidepressant drug response of individual patients with MDD based on longitudinal data. Methods: Longitudinal data from patient records, including sex, age, outpatient or inpatient status, medication type and dosage, and the Hamilton Depression Rating Scale (HAMD) scores, were used to train the Transformer model and the 1-dimensional convolutional neural network model. Individual patient records were allocated to training (80%), validation (10%), and testing (10%) datasets. Results: The AI model demonstrated 88% sensitivity and 92% specificity for predicting the treatment response. Significant factors independently associated with the antidepressant response included age, sex, history of depression, and baseline HAMD scores. Conclusion This AI-driven software application provides a clinically valuable tool for predicting treatment response.While promising, further research is needed to incorporate voice data into the AI model using the voice recording feature to further improve diagnostic accuracy.

Background: The diagnosis of major depressive disorder (MDD) relies primarily on clinical interviews, which can be subjective and time consuming. Thus, there is a need for more objective diagnostic tools. The aim of this study was to develop an artificial intelligence (AI) application that predicts the antidepressant drug response of individual patients with MDD based on longitudinal data. Methods: Longitudinal data from patient records, including sex, age, outpatient or inpatient status, medication type and dosage, and the Hamilton Depression Rating Scale (HAMD) scores, were used to train the Transformer model and the 1-dimensional convolutional neural network model. Individual patient records were allocated to training (80%), validation (10%), and testing (10%) datasets. Results: The AI model demonstrated 88% sensitivity and 92% specificity for predicting the treatment response. Significant factors independently associated with the antidepressant response included age, sex, history of depression, and baseline HAMD scores. Conclusion This AI-driven software application provides a clinically valuable tool for predicting treatment response.While promising, further research is needed to incorporate voice data into the AI model using the voice recording feature to further improve diagnostic accuracy.

Background: The diagnosis of major depressive disorder (MDD) relies primarily on clinical interviews, which can be subjective and time consuming. Thus, there is a need for more objective diagnostic tools. The aim of this study was to develop an artificial intelligence (AI) application that predicts the antidepressant drug response of individual patients with MDD based on longitudinal data. Methods: Longitudinal data from patient records, including sex, age, outpatient or inpatient status, medication type and dosage, and the Hamilton Depression Rating Scale (HAMD) scores, were used to train the Transformer model and the 1-dimensional convolutional neural network model. Individual patient records were allocated to training (80%), validation (10%), and testing (10%) datasets. Results: The AI model demonstrated 88% sensitivity and 92% specificity for predicting the treatment response. Significant factors independently associated with the antidepressant response included age, sex, history of depression, and baseline HAMD scores. Conclusion This AI-driven software application provides a clinically valuable tool for predicting treatment response.While promising, further research is needed to incorporate voice data into the AI model using the voice recording feature to further improve diagnostic accuracy.

Background: The diagnosis of major depressive disorder (MDD) relies primarily on clinical interviews, which can be subjective and time consuming. Thus, there is a need for more objective diagnostic tools. The aim of this study was to develop an artificial intelligence (AI) application that predicts the antidepressant drug response of individual patients with MDD based on longitudinal data. Methods: Longitudinal data from patient records, including sex, age, outpatient or inpatient status, medication type and dosage, and the Hamilton Depression Rating Scale (HAMD) scores, were used to train the Transformer model and the 1-dimensional convolutional neural network model. Individual patient records were allocated to training (80%), validation (10%), and testing (10%) datasets. Results: The AI model demonstrated 88% sensitivity and 92% specificity for predicting the treatment response. Significant factors independently associated with the antidepressant response included age, sex, history of depression, and baseline HAMD scores. Conclusion This AI-driven software application provides a clinically valuable tool for predicting treatment response.While promising, further research is needed to incorporate voice data into the AI model using the voice recording feature to further improve diagnostic accuracy.

Objectives: Despite the growing concern regarding the adverse effects related to problematic smartphone use (PSU), little is known about underlying morphologic changes in the brain. The brainstem is a deep brain structure that consists of several important nuclei associated with emotions, sensations, and motor functions. In this study, we sought to examine the difference in the volume of brainstem substructures among adolescents with and without PSU. Methods: A total of 87 Korean adolescents participated in this study. The PSU group (n=20, age=16.2±1.1, female:male=12:8) was designated if participants reported a total Smartphone Addiction Proneness Scale (SAPS) score of ≥42, whereas the remaining participants were assigned to the control group (n=67, age=15.3±1.7, female:male=19:48). High-resolution T1 magnetic resonance imaging was performed, and the volume of each of the four brainstem substructures [midbrain, pons, medulla, and superior cerebellar peduncle (SCP)] was measured. Analysis of covariance was conducted to reveal group differences after adjusting for effects of age, gender, whole brainstem volume, depressive symptoms, and impulsivity. Results: The PSU group showed a significantly smaller volume of the SCP than the control group (F=8.273, p=0.005). The volume of the SCP and the SAPS score were negatively correlated (Pearson’s r=-0.218, p=0.047). Conclusion The present study is the first to reveal an altered volume of the brainstem substructure among adolescents with PSU. This finding suggests that the altered white matter structure in the brainstem could be one of the neurobiological mechanisms underlying behavioral changes in PSU.