Self-administered Remote Neurological Examination Using Mobile Application in Patients With Brain Tumors

The goal of this observational study is to evaluate the feasibility and accuracy of a self-administered remote neurological examination using the "Iskhaa" mobile application in patients with brain tumors aged above 5 years who are able to follow app-based instructions. The main questions it aims to answer are: 1. Development of a mobile application equipped with symptom assessment and recording videos as patients perform specific neurological tasks. 2. Development and validation of the AI model to detect functional changes and predict subsequent neurological deterioration. Participants will: 1. Use the Iskhaa mobile application to perform guided self-neurological examinations following pre-recorded video instructions. 2. Complete EORTC QLQ-C30 and BN20 questionnaires for quality of life assessment. 3. Record and upload videos (e.g., speech, walking, limb movements) using their mobile camera for analysis. 4. In Phase 1 (onsite), 100 participants will use the app under supervision to ensure usability and accuracy. 5. In Phase 2 (offsite), 500 participants will use the app independently at home for monthly self-assessments, with reminders and follow-up support. 6. Continue routine clinic visits every 3-6 months and imaging every 6-12 months as per standard clinical care. The study will compare app-recorded data with physician assessments to determine agreement and validity of remote neurological monitoring using artificial intelligence analysis.

The mobile application "Iskhaa" will be developed in collaboration with the Electronics Division, Bhabha Atomic Research Centre (BARC). The application will be compatible with Android version 8.0 or above. The software will have two components-mobile application (for patients and healthcare providers) and a desktop application (for healthcare providers). The data recorded using the app will be stored in a secured hospital server accessible via the hospital network. The sample of different clinical features to be recorded by the application is included in Annexure A. The mobile application will be initially developed in English and subsequently translated to Hindi, Marathi, and Bengali in upgraded versions over initial 3-6 months. Pre-recorded videos will be used to guide the patients in repeating the actions that will be recorded for neurological examination. Examples of such instructions will include speech (patient asked to speak fixed sentences), gait, e.g., walking 20 steps, tandem walking to check cerebellar function, lifting individual limbs, etc. In addition, the European Organisation for Research and Treatment of Cancer (EORTC) quality of life core questionnaire (QLQ-C30) and brain cancer module (BN20) will be available in the mobile application. The recorded symptoms will be stored in a secured database, and the study investigators will have access to review them using either mobile or desktop applications. Once an assessment (either onsite or onsite) is completed, the data will be updated electronically at the central site housed in Tata Memorial Centre, and the investigator will receive a reminder. Version upgradation for better improvement of the application interface for better user compliance and software updates will be done periodically as required. A sample of the application interface is provided in the diagram below. Patients aged more than 5 years with a diagnosis of brain tumor, having an expected life expectancy of more than 6 months, and able to follow instructions of the mobile application will be screened for the study. They will be accrued after signing a written consent or assent form as appropriate. The initial phase of the study will include an onsite assessment of 100 patients who will be required to use the mobile application to assess symptoms and record videos for neurological functional assessment. In this phase, the application will be installed on specific mobile phones (not belonging to the patient). After performing the evaluation, all the patients will be evaluated by the responsible physician in the clinic, who will follow standard protocols. A detailed neurological examination will be conducted to cover all aspects of the symptoms and functional assessment as available in the mobile application. The primary purpose of the on-site evaluation is to provide any technical support for the application use by the in-house team before installation on mobile devices for patients to allow seamless transition and use. The second phase of assessment will include 500 patients who will perform the virtual examination (off-site). The mobile application will be installed on individual phones when consent is obtained for study participation. Patients will be required to undergo monthly assessments at home with reminders set up for performing the timely evaluations, and they will be reminded by telephone in case of delay in taking the assessment by 7 days. Patients and caregivers will be able to record and upload the video using the mobile camera. Additionally, patients will be instructed to undergo self-assessments (unscheduled) in case of symptomatic worsening. Patients will continue follow-up in the clinic for physical evaluation every 3-6 months as per standard institutional practice and imaging scheduled every 6-12 monthly specific for the histological diagnosis or as mandated clinically in case of neurological worsening as decided by the responsible physician without any influence of the study participation. All the data completed will be available for study investigators to retrieve. The symptomatology assessment will be presented as descriptive statistics and serial changes will be analyzed. The data recorded using the app will be correlated with the assessment by the physician during each clinic visit. Computational analysis will be performed for all the individual tasks undertaken by the patient and recorded in the video format. Neurological functioning (e.g., gait and movement) will be recorded using artificial intelligence algorithms with a hybrid model that combines convolutional neural networks (CNNs) and vision transformers (ViTs). The video data will undergo preprocessing steps to extract relevant features, such as silhouettes, skeletal representations, or pose key points, using advanced pose estimation tools like OpenPose. Each frame will be processed by a CNN backbone, such as ResNet or EfficientNet, to extract low-level spatial features. These features will then be fed into a Vision Transformer module, which will model global temporal dependencies across the sequence of frames, providing a robust representation of the patient's gait. This combined framework will enable the detection of subtle anomalies, abnormalities, or variations in gait, which are critical for clinical monitoring. The model will be trained and validated using video data captured from diverse patient populations, and clinical scenarios. By leveraging the granularity of high-speed footage, the model will aim to deliver high-precision outputs such as gait classification, anomaly detection, or parameter regression tailored to the specific needs of applications. The integration of state-of-the-art techniques, such as 3D Vision Transformers for enhanced spatiotemporal analysis, will ensure the model's robustness and generalizability. Furthermore, Explainable AI (XAI) methodologies will be incorporated to interpret the model's predictions, providing actionable insights and fostering trust in the system.