Explainable deep learning-based survival prediction for non-small cell lung cancer patients undergoing radical radiotherapy.

BACKGROUND AND PURPOSE: Survival is frequently assessed using Cox proportional hazards (CPH) regression; however, CPH may be too simplistic as it assumes a linear relationship between covariables and the outcome. Alternative, non-linear machine learning (ML)-based approaches, such as random survival forests (RSFs) and, more recently, deep learning (DL) have been proposed; however, these techniques are largely black-box in nature, limiting explainability. We compared CPH, RSF and DL to predict overall survival (OS) of non-small cell lung cancer (NSCLC) patients receiving radiotherapy using pre-treatment covariables. We employed explainable techniques to provide insights into the contribution of each covariable on OS prediction. MATERIALS AND METHODS: The dataset contained 471 stage I-IV NSCLC patients treated with radiotherapy. We built CPH, RSF and DL OS prediction models using several baseline covariable combinations. 10-fold Monte-Carlo cross-validation was employed with a split of 70%:10%:20% for training, validation and testing, respectively. We primarily evaluated performance using the concordance index (C-index) and integrated Brier score (IBS). Local interpretable model-agnostic explanation (LIME) values, adapted for use in survival analysis, were computed for each model. RESULTS: The DL method exhibited a significantly improved C-index of 0.670 compared to the CPH and a significantly improved IBS of 0.121 compared to the CPH and RSF approaches. LIME values suggested that, for the DL method, the three most important covariables in OS prediction were stage, administration of chemotherapy and oesophageal mean radiation dose. CONCLUSION: We show that, using pre-treatment covariables, a DL approach demonstrates superior performance over CPH and RSF for OS prediction and use explainable techniques to provide transparency and interpretability.

AUDIO-CASI: Hardware and Software Considerations in Adding Sound to a Computer-Assisted Interviewing System.

This article reviews a multimedia application in the area of survey measurement research: adding audio capabilities to a computer-assisted interviewing system. Hardware and software issues are discussed, and potential hardware devices that operate from DOS platforms are reviewed. Three types of hardware devices are considered: PCMCIA devices, parallel port attachments, and laptops with built-in sound.

AUDIO AND VIDEO COMPUTER-ASSISTED SELF INTERVIEWING: Preliminary Tests of New Technologies for Data Collection.

Results are reported from a preliminary study testing a new technology for survey data collection: audio computer-assisted self interviewing. This technology has the theoretical potential of providing privacy (or anonymity) of response equivalent to that of paper self-administered questionnaires (SAQs). In addition, it could offer the advantages common to all computer-assisted methods such as the ability to implement complex questionnaire logic, consistency checking, etc.. In contrast to Video-CASI, Audio-CASI proffers these potential advantages without limiting data collection to the literate segment of the population. In this preliminary study, results obtained using RTI's Audio-CASI system were compared to those for paper SAQs and for Video-CASI. Survey questionnaires asking about drug use, sexual behavior, income, and demographic characteristics were administered to a small sample (N = 40) of subjects of average and below-average reading abilities using each method of data collection. While the small sample size renders many results suggestive rather than definitive, the study did demonstrate that both Audio- and Video-CASI systems work well even with subjects who do not have extensive familiarity with computers. Indeed, respondents preferred the Audio- and Video-CASI to paper SAQs. The computerized systems also eliminated errors in execution of "skip" instructions that occurred when subjects completed paper SAQs. In a number of instances, the computerized systems also appeared to encourage more complete reporting of sensitive behaviors such as use of illicit drugs. Among the two CASI systems, respondents rated Audio-CASI more favorably than Video-CASI in terms of interest, ease of use, and overall preference.