An Explainable Intelligence Driven Query Prioritization Using Balanced Decision Tree Approach for Multi-Level Psychological Disorders Assessment

Explainable artificial intelligence for mental health through transparency and interpretability for understandability

The literature on artificial intelligence (AI) or machine learning (ML) in mental health and psychiatry lacks consensus on what "explainability" means. In the more general XAI (eXplainable AI) literature, there has been some convergence on explainability meaning model-agnostic techniques that augment a complex model (with internal mechanics intractable for human understanding) with a simpler model argued to deliver results that humans can comprehend. Given the differing usage and intended meaning of the term "explainability" in AI and ML, we propose instead to approximate model/algorithm explainability by understandability defined as a function of transparency and interpretability. These concepts are easier to articulate, to "ground" in our understanding of how algorithms and models operate and are used more consistently in the literature. We describe the TIFU (Transparency and Interpretability For Understandability) framework and examine how this applies to the landscape of AI/ML in mental health research. We argue that the need for understandablity is heightened in psychiatry because data describing the syndromes, outcomes, disorders and signs/symptoms possess probabilistic relationships to each other-as do the tentative aetiologies and multifactorial social- and psychological-determinants of disorders. If we develop and deploy AI/ML models, ensuring human understandability of the inputs, processes and outputs of these models is essential to develop trustworthy systems fit for deployment.

A Nomogram for Predicting Cancer-Specific Survival in Children With Wilms Tumor: A Study Based on SEER Database and External Validation in China

Background

Wilms tumor (WT) is the most common tumor in children. We aim to construct a nomogram to predict the cancer-specific survival (CSS) of WT in children and externally validate in China.

Methods

We downloaded the clinicopathological data of children with WT from 2004 to 2018 in the SEER database. At the same time, we used the clinicopathological data collected previously for all children with WT between 2013 and 2018 at Children's Hospital of Chongqing Medical University (Chongqing, China). We analyzed the difference in survival between the patients in the SEER database and our hospital. Cox regression analysis was used to screen for significant risk factors. Based on these factors, a nomogram was constructed to predict the CSS of children with WT. Calibration curve, concordance index (C-index), the area under the receiver operating curve (AUC) and decision curve analysis (DCA) was used to evaluate the accuracy and reliability of the model.

Results

We included 1,045 children with WT in the SEER database. At the same time, we collected 112 children with WT in our hospital. The Kaplan-Meier curve suggested that children in China with WT had a higher mortality rate than those in the United States. Cox regression analysis revealed that age, lymph node density (LND), and tumor stage were significant prognostic factors for the patients in the SEER database. However, the patients in our hospital only confirmed that the tumor stage and the number of positive regional lymph nodes were significant factors. The prediction model established by the SEER database had been validated internally and externally to prove that it had good accuracy and reliability.

Conclusion

We have constructed a survival prognosis prediction model for children with WT, which has been validated internally and externally to prove accuracy and reliability.

A Web-Based Prediction Model for Cancer-Specific Survival of Middle-Aged Patients With Non-metastatic Renal Cell Carcinoma: A Population-Based Study

Background

Renal cell carcinoma (RCC) is one of the most common cancers in middle-aged patients. We aimed to establish a new nomogram for predicting cancer-specific survival (CSS) in middle-aged patients with non-metastatic renal cell carcinoma (nmRCC).

Methods

The clinicopathological information of all patients from 2010 to 2018 was downloaded from the SEER database. These patients were randomly assigned to the training set (70%) and validation set (30%). Univariate and multivariate COX regression analyses were used to identify independent risk factors for CSS in middle-aged patients with nmRCC in the training set. Based on these independent risk factors, a new nomogram was constructed to predict 1-, 3-, and 5-year CSS in middle-aged patients with nmRCC. Then, we used the consistency index (C-index), calibration curve, and area under receiver operating curve (AUC) to validate the accuracy and discrimination of the model. Decision curve analysis (DCA) was used to validate the clinical application value of the model.

Results

A total of 27,073 patients were included in the study. These patients were randomly divided into a training set (N = 18,990) and a validation set (N = 8,083). In the training set, univariate and multivariate Cox regression analysis indicated that age, sex, histological tumor grade, T stage, tumor size, and surgical method are independent risk factors for CSS of patients. A new nomogram was constructed to predict patients' 1-, 3-, and 5-year CSS. The C-index of the training set and validation set were 0.818 (95% CI: 0.802-0.834) and 0.802 (95% CI: 0.777-0.827), respectively. The 1 -, 3 -, and 5-year AUC for the training and validation set ranged from 77.7 to 80.0. The calibration curves of the training set and the validation set indicated that the predicted value is highly consistent with the actual observation value, indicating that the model has good accuracy. DCA also suggested that the model has potential clinical application value.

Conclusion

We found that independent risk factors for CSS in middle-aged patients with nmRCC were age, sex, histological tumor grade, T stage, tumor size, and surgery. We have constructed a new nomogram to predict the CSS of middle-aged patients with nmRCC. This model has good accuracy and reliability and can assist doctors and patients in clinical decision making.