The research group comprises 2 occupational therapists (authors APP and VMCE) and 2 computer scientists (authors OJMN and MDGCP). To design the initial software architecture, the group received input from other health and computing professionals [56].

The study was conducted in a specialized rehabilitation center that requires both ethical committee approval and verification of ethical and safety aspects for studies conducted on its premises. Therapists participating had at least 1 year of experience in stroke rehabilitation and received training on device use before working with patients. The smartphones used in the feasibility study were either from the research team, having been tested and shown to function properly, or were participants' own devices. To minimize malfunction risks, no smartphone was allowed to be charged during use by participants. The Informed Consent Form informed participants about potential discomfort due to heat and the vest and assured them that, in the unlikely event of a smartphone malfunction causing an explosion, immediate assistance and emergency services would be provided. Data acquisition and storage were handled securely, with all data kept anonymized and stored in accounts requiring login via a secure network. Photographs were authorized with face identification removed. Participants consented to the use of data for the study and were informed of their right to withdraw at any time without coercion or obligation. Participation was voluntary and did not include any financial benefit, in compliance with Brazilian regulations.

The study comprised design and development iterations involving brainstorming meetings, interviews with therapists and patients, prototype development, feasibility studies, and experimental sessions (see workflow in Textbox 1).

Throughout the study, our literature review was ongoing and dynamically aligned with emerging themes. Initially, it concentrated on the specific needs of patients with stroke [4] and the effectiveness of trunk training [15]. Next, we reviewed work on novel wearable [21] and affordable smartphone-based solutions [31], including reports in literature reviews [29,57], and noticed the involvement of healthy adults in feasibility studies [26,58], as well as control participants [27]. These works are representative of the broader ongoing review process, which continued to cover these and additional relevant areas throughout the study.

We used the Google Sheets software for statistical analysis relative to the within-subject study. We used thematic analysis [59] for categorizing interview responses. For the kinematic assessment, we compared angular momentum from values calculated by Vicon and the smartphone’s sensors.

Our study involved 3 groups: poststroke participants, therapists, and healthy participants. We used interviews and questionnaires to collect requirements, demographic data, and their impressions of Postural SmartVest.

For the interviews with therapists, the inclusion criterion was having at least 1 year of experience in stroke rehabilitation. We invited professionals from a rehabilitation center using printed flyers and email invitations. In the feasibility study, we recruited participants via university email: the inclusion criteria were being 18 years or older and owning a smartphone. The exclusion criterion was self-reported physical or cognitive impairments. For the pilot and within-subjects studies, the inclusion criteria were having chronic hemiparesis, being 18 years or older, walking independently, having cognitive ability for communication, and attending rehabilitation sessions at least twice a week for a minimum of 1 month.

For the brainstorming meetings, the research team held three 2-hour sessions to gather functional and nonfunctional requirements for the solution to develop the initial prototype comprising the vest and the app, Prototype #1 (Figure 1).

For the interviews with therapists, one team member (APP) conducted semi-structured interviews with therapists. The interview adhered to a structured protocol comprising 8 questions categorized into 4 categories: resources for posture improvement, general technology acceptance, specific technology requirements, and smartphone use (Multimedia Appendix 1: App A).

For the feasibility study, we used the first version of the app. We created 8 vests and conducted sessions with 8 participants at a time. We provided smartphones to participants who could not use their own. Each session lasted 2 hours and followed a 3-phase protocol: preparation, deployment, and follow-up. In the preparation phase, therapists adjusted the vests and calibrated the seating posture for each participant.

In the deployment phase, participants set their devices to airplane mode and wore headphones to ensure privacy for their feedback. They then engaged in smartphone-free activities like reading or using their laptops (Figure 1). The deployment phase consisted of two 30-minute segments: feedback from the app was turned off during the first part and turned on during the second. The app continuously recorded the participants’ postural data throughout both segments in this phase.

In the follow-up phase, participants responded to 2 questionnaires, one with scaled questions and the other with open-ended inquiries. These questionnaires aimed to assess usability and satisfaction. One questionnaire evaluated Postural SmartVest as an assistive technology for posture correction (Multimedia Appendix 1: App B), while the second assessed Postural SmartVest as an assistive technology in general (Multimedia Appendix 1: App C). For the latter questionnaire, we adapted the Quebec User Evaluation of Satisfaction with Assistive Technology (QUEST; version 2.0), a 12-item assessment designed to measure user satisfaction with devices and services [60]. We used a translated and validated version [61] and focused on the device-related section of the questionnaire. This section included 8 scale questions that assessed size, weight, adjustability, safety, durability, ease of use, comfort, and overall effectiveness of the technology.

The feedback from the previous iteration guided the design of Prototype #2. We then used this prototype in a pilot study involving 1 patient with hemiparesis to observe the solution and detail the within-subjects study design. In the pilot study and the following studies, the smartphone used was a Motorola G with Android (version 5.1).

During the pilot study, the poststroke patient followed a protocol that included 5 ambulant-based daily activities conducted in the therapy room (Figure 2). The protocol consisted of 4 distinct phases. In the first phase, the participant walked 8 m, incorporating forward, sideways, and backward walking. In the second phase, they proceeded with a 12-m free walk to a refrigerator. In the third phase, the participant engaged in an upper limb activity that entailed opening the refrigerator, taking a glass, and returning it to the fridge. Finally, in the fourth phase, they completed a 12-m free walk back to the starting point.

The feedback from the pilot study informed the development of a new prototype (Prototype #3). At this stage, we checked the accelerometer readings obtained by the app running on Motorola G with Android (version 5.1).

On the one hand, sensor type and placement, activity characteristics, and population-specific conditions can influence the accuracy of sensor outcome estimates, as reported in a study that used research-grade wearable sensors on the upper arms, waist, and ankles and included poststroke patients [62]. On the other hand, the literature reports positive results regarding using state-of-the-art smartphone accelerometer technology. For instance, researchers investigated the comparative performance of 3 different commercially available smartphone accelerometers among themselves and to a gold-standard Vicon MX motion capture system, indicating that the devices are valid and reliable measuring instruments for estimating linear accelerations [63].

We assessed the angle calculations of the third prototype by comparing them with the Vicon Gait Plug-in system, a gold-standard in full-body kinematic and kinetic modeling. The Vicon system includes software and fixed cameras that capture signals from a moving target, providing visual feedback through animated vectors and plans, and it also calculates quantitative data, including distances and angles between selected planes.

Under the guidance of one occupational therapist (APP), author OJMN used the prototype to perform 6 movements: trunk flexion, trunk extension, left and right-side bending, and left and right trunk rotation (Figure 3).

The therapist calibrated the subject’s optimal posture position, and for all movements, the subject began in an upright position, executed the movement, and returned to the initial position. The Vicon system and the app prototype collected data simultaneously, following an initial manual synchronization process, with the Vicon recording frames at 250 frames per second and the app prototype recording angular moments at each second along with coordinate values. To reconstruct the trunk segment, we placed markers on the clavicle, sternum, cervical vertebra C7, thoracic vertebra T10, and a point located on the medial border of the right scapula. We conducted 6 sessions using 39 markers on a full-body model (Figure 4).

Following the kinematic assessment, Prototype #3 was referred to as SmartVest and was used for the remainder of the study (Figure 5).

The within-subjects study included individual sessions with poststroke patients, each accompanied by their therapists. Before that, therapists used the Postural SmartVest for about 30 minutes while the researchers provided guidelines to ensure the professionals understood how the solution worked. Only afterward did the therapists apply Postural SmartVest to their patients during their rehabilitation sessions. All sessions used the same smartphone (Motorola G with Android; version 5.1).

In the presession interview with each patient, we gathered information about their posture control during activities, awareness of posture issues, and any potential biases toward one side of the body. We also inquired about their strategies to correct posture problems when they noticed them (Multimedia Appendix 1: App D).

At the beginning of each session, the therapist adjusted the vest on the patient, calibrated the app to recognize the patient’s best-at-the-time standing posture, and initiated the monitoring process (Textbox 2). Subsequently, the patient performed the activities defined in our pilot study twice, in consecutive sessions, each with an estimated duration of 10 minutes (Figures 6 and 7). The app did not provide feedback in the first session but registered all the movements. In the subsequent session, the app provided feedback through screen color changes (green and red), vibrations, and audio guidance.

Afterward, we asked the poststroke patients to fill out 2 questionnaires, one focusing on Postural SmartVest as an assistive technology for posture correction (Multimedia Appendix 1: App B) and the other assessing Postural SmartVest as a general assistive technology (Multimedia Appendix 1: App C).

Additionally, the therapists shared their feedback after working with the patients using a tailored version of the Postural SmartVest QUEST (version 2.0) questionnaire (Multimedia Appendix 1: App E). This study helped us understand the effectiveness of Postural SmartVest during stroke rehabilitation, both from the patient’s and therapist’s perspectives.

The study received prior approval from the Hospital of the Clinics of the Ribeirão Preto School of Medicine at the University of São Paulo, Brazil, identifier 57234816.3.0000.5440. As detailed in the text approved by the ethics committee, the study adhered to ethical principles by ensuring beneficence through rigorous monitoring and support for participants, justice by offering equitable access and fair treatment, and respect by obtaining informed consent and addressing potential discomfort and safety concerns proactively.

For the interviews, we enlisted 20 therapists: 10 occupational therapists and 10 physiotherapists. All participants were female, aged 26-40, with varying years since graduation (Table 1).

For the feasibility study, we recruited 28 healthy participants, aged 20-62, with an average age of 30.9. The group included 17 males and 11 females (Table 1). Of the participants, 25 used their own smartphones, while 3 used devices provided by us. The study covered 22 different smartphone models and 7 versions of the Android operating system.

For the pilot and within-subjects studies, we recruited poststroke participants from patients attending rehabilitation sessions at least twice a week for a month at the exact center. Inclusion criteria included chronic hemiparesis from stroke, aged 18 years or older, independent walking ability, cognitive ability for communication, and participation in rehabilitation sessions for at least a month.

We enrolled 1 patient for the pilot study and 40 patients for the within-subjects study (27 males, 67.5%), with a mean age of 57.1 (SD 11.42) years, and functional independence categorized as independent (n=39) of moderate assistance (n=1) on the functional independence measure [64,65]. Among them, 24 had right hemispheric lesions, 16 had left hemispheric lesions, and 38 had ischemic strokes. The duration of their condition ranged from 6 months to 19 years, with a majority between 1 and 4 years (Table 1).

The Postural SmartVest solution comprises a customizable smartphone app for Android devices securely placed in a transparent pocket on an athletic, lightweight compression tank top (Figure 5).

We designed Postural SmartVest, a solution consisting of a chest garment holding a smartphone running an Android application. When the user wears the Postural SmartVest in a rehabilitation session, the therapist first calibrates the app to the best posture the patient can achieve at the time. The app uses the smartphone accelerometer to identify variations in the user’s body movement in the sagittal and frontal planes relative to the calibrated position. Based on these variations, the app provides multisensorial feedback that guides the patient to perform the movements required to return to the calibrated position.

The iterative development of Postural SmartVest, informed by therapist interviews and feasibility study feedback, tailored the final product to user needs. The app reliably identifies posture changes and provides corresponding multisensorial feedback, as substantiated by kinematic assessment.

Postural SmartVest significantly improved patients’ ability to maintain their best-at-the-time postures, with patients showing a higher capacity for correct positions when using feedback. Although the analysis found no significant difference in the number of movements between stages, the increased movement during the feedback stage suggests an active engagement with the system for postural adjustments. This engagement aligns with rehabilitation goals, emphasizing motor skill enhancement and proprioception. These findings highlight Postural SmartVest’s potential as an assistive technology for poststroke patients, emphasizing the importance of feedback in rehabilitation tools.

Postural SmartVest’s flexibility meets each patient’s unique needs and progress, with a highly customizable app enabling therapists to personalize feedback and adjust thresholds. This multisensory approach enhances engagement and may result in more effective posture correction. The app offers a comprehensive feedback system, including visual cues, tactile vibrations, and audio instructions.

The technology received positive feedback regarding its design and usability, with patients and therapists giving favorable ratings for various aspects of Postural SmartVest, including weight, comfort, dimensions, and effectiveness. Patients believed in its utility for walking assistance and motor perception improvement, and their suggestions aligned with their approval. Many patients were willing to recommend Postural SmartVest to other patients with poststroke hemiparesis, highlighting its perceived value. Healthy users also expressed satisfaction with Postural SmartVest’s posture support.

Our study used a specific sample of poststroke patients, which may not fully represent the diverse stroke survivor population. This highlights the need for including a broader and more diverse sample in future research. Although the findings indicate high satisfaction among both patients and therapists, individual preferences and responses varied. Patient feedback identified several limitations that suggest potential improvements to the app's personalization features, such as the ability to adjust the audio volume while using the vest and the inclusion of customizable vibration patterns and prompt frequencies. Additionally, enabling therapists to record guidance prompts in their own voices could enhance the patient experience, as one patient noted: “Listening to the app's directions reminds me of you (the therapist) talking to me during the rehabilitation sessions.”

The study’s duration might have limited the assessment of long-term effects, as improvements in posture maintenance could potentially change over time. A longer follow-up period could provide insights into whether the benefits observed are sustained in the long term. Additionally, the self-controlled approach used in the study—where participants served as their own controls by using the app with feedback turned off—may introduce bias related to user expectations and behavior. Future research should consider alternative control group designs to minimize potential biases and enhance the validity of the results. Another limitation of this study is the relatively unstructured approach to coding and scoring responses from interviews and feedback sessions. The multidimensional nature of the questions, ranging from levels of agreement to broader attitudes and preferences, presented challenges in achieving uniform analysis. To address this, we plan to use more structured, software-assisted procedures in future research to standardize data collection and analysis [68,69], thereby improving the consistency and depth of understanding gained from participant feedback.

Future research should also explore categorizing participants based on injury time, dominance, or body laterality to better understand the varying impacts of the Postural SmartVest. Including studies that involve patients using the device at home could provide insights into its real-world effectiveness and long-term impact in poststroke rehabilitation outside of clinical settings. Understanding how patients integrate the Postural SmartVest into their daily routines and the effects of extended, unsupervised use could reveal valuable information about its benefits in home-based rehabilitation. Furthermore, while Postural SmartVest records a range of activities, including posture changes and feedback delivery, it currently lacks features for therapists to access and use this data effectively. Future research should focus on developing these functionalities to enhance patient monitoring and rehabilitation planning. Additionally, addressing the requirements identified during initial therapist interviews, such as including a photo and video gallery and a dashboard for progress monitoring, could improve Postural SmartVest’s functionality and user satisfaction.

Postural SmartVest is a modified athletic compression tank top with a transparent pocket for a smartphone running a customizable Android app. This app offers tactile vibrations, dynamic visual feedback, and clear audio instructions for posture monitoring and guidance. It is designed to be minimally intrusive, replicable, and useful for both therapists and poststroke patients, using readily available Android smartphones and avoiding the need for expensive specialized equipment.

The development of Postural SmartVest reflects the evolution of cognitive rehabilitation from classical methods to personalized systems with kinematic analysis and user-centered design [70]. It incorporates feedback from patients with stroke and therapists, aligning with recommendations for wearable technology design [71]. Unlike wearable rehabilitation systems that prioritize sensor accuracy over comfort [24], Postural SmartVest is distinguished by its comfort and effective smartphone placement, addressing both functional and user experience aspects.

Research on user perspectives for wearable systems in upper extremity stroke rehabilitation highlights the importance of monitoring both upper extremity and trunk movements and evaluating their quality and quantity [39]. In our experimental sessions, we included activities such as position maintenance, lateral trunk movements, trunk rotation, forward-backward trunk motions, and object retrieval, aligning with tasks assessed by various trunk ability evaluation tools [72]. While some studies have investigated smartphone-based training to improve balance and trunk performance among seated patients with stroke—using modified balance boards with visual and audio feedback in an external monitor [48,49] or smartphones attached to harnesses for screen feedback [25]—Postural SmartVest supports dynamic ambulatory movements and provides visual feedback directly on the smartphone screen, using a single color (red or green) to indicate posture status. This approach is particularly effective in therapy rooms with mirror walls and accommodates individuals with visual deficiencies, which are common after a stroke [73].

Postural SmartVest uses the smartphone’s accelerometer on the patient’s trunk to assist with trunk posture, while many studies use sensors on different body parts for various purposes [74-77]. While postural SmartVest uses smartphone-generated sensor data to identify trunk postures and facilitate posture correction, existing literature often combines data from multiple sensors placed on the user’s body for posture detection [78].

A wealth of research envisions using their solutions to foster collaboration among health professionals [79] or in the patient’s home environment, including studies that use smartphones as a communication component in a remote rehabilitation platform [18,25,33,35,70,76,80,81]. Although we designed Postural SmartVest primarily to assist patients during rehabilitation sessions, we are actively extending our research to the home environment by integrating Postural SmartVest with our experience sampling and programmed intervention platform [68].

Our goal was to create an affordable wearable, Postural SmartVest, aiding poststroke patients in posture maintenance during rehabilitation. Its simplicity and cost-effectiveness, using widely available devices with efficient sensors, enable multisensory feedback, enhancing accessibility. Patients and therapists expressed utility and satisfaction, emphasizing its clinical potential.

A comparison of patient session data using Postural SmartVest, both with and without feedback, informed essential insights. The significant difference in maintaining correct positions underscored the value of feedback. Although there was no disparity in the number of movements, the increased adjustments during sessions with feedback suggested active patient engagement. These findings emphasize the potential of Postural SmartVest as a valuable assistive technology for poststroke patients and highlight the importance of integrating feedback into rehabilitation tools.

Our study indicates several future research directions, including studying diverse stroke survivor populations, assessing long-term effects, exploring user engagement factors, and evaluating therapist training impact. Additionally, we should prioritize functional outcomes, technology adoption, cost-effectiveness, and addressing unmet needs like enhancing posture visualization and aiding long-term posture monitoring.