Smoking Detection and Cessation: An Updated Scoping Review of Digital and Mobile Health Technologies

IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, VOL. 29, NO. 7, JULY 2025 5191 Smoking Detection and Cessation: An Updated Scoping Review of Digital and Mobile Health Technologies Mirko Casu , Francesco Guarnera , Giusy Rita Maria La Rosa , Sebastiano Battiato , Senior Member, IEEE, Pasquale Caponnetto , Riccardo Polosa , and Rosalia Emma Abstract—Digital and mobile health technologies offer promising solutions for smoking detection and cessation. This scoping review examines the current state of research and development in this field, encompassing smartphone applications, wearable devices, and sensor-based systems. We analyzed 49 studies published between 2019 and 2023 from PubMed and ACM Digital Library, focusing on technology features, outcomes, and evaluation methods. Wearable sensors and smartphone apps show potential in combating smoking addiction and improving quit rates. Motion sensors for hand-to-mouth gesture detection achieve high accuracy in controlled settings but face challenges in realworld applications. Machine learning models and wireless signal detection techniques yield encouraging results but Received 17 September 2024; revised 23 December 2024 and 27 January 2025; accepted 3 March 2025. Date of publication 10 March 2025; date of current version 4 July 2025. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. The work of Riccardo Polosa (as task leader), Mirko Casu, Sebastiano Battiato and Rosalia Emma has been supported by Ministry of University and Research (MUR) in the framework of the National Recovery and Resilience Plan (PNRR)E63C22000900006, SPOKE 1, Work Page Health under project SAMOTHRACE (Sicilian Micro and Nano Technology Research and Innovation Center). The work of Francesco Guarnera has been supported by MUR in the framework of PNRR PE0000013, under project “Future Artificial Intelligence Research –FAIR”.(Corresponding author: Mirko Casu.) Mirko Casu is with the Department of Mathematics and Computer Science, University of Catania, 95123 Catania, Italy, and also with the Department of Educational Sciences, Section of Psychology, University of Catania, 95123 Catania, Italy (e-mail: mirko[email protected]). Francesco Guarnera is with the Department of Mathematics and Computer Science, University of Catania, 95123 Catania, Italy (e-mail: francesco.guar[email protected]). Giusy Rita Maria La Rosa is with the Department of Clinical & Experimental Medicine, University of Catania, 95123 Catania, Italy (e-mail: giusy[email protected]). Sebastiano Battiato is with the Department of Mathematics and Computer Science, University of Catania, 95123 Catania, Italy, and also with the Center of Excellence for the Acceleration of Harm Reduction, University of Catania, 95123 Catania, Italy (e-mail: [email protected]). Pasquale Caponnetto is with the Department of Educational Sciences, Section of Psychology, University of Catania, 95123 Catania, Italy, and also with the Center of Excellence for the Acceleration of Harm Reduction, University of Catania, 95123 Catania, Italy (e-mail: p[email protected]). Riccardo Polosa and Rosalia Emma are with the Department of Clinical & Experimental Medicine, University of Catania, 95123 Catania, Italy, and also with the Center of Excellence for the Acceleration of Harm Reduction, University of Catania, 95123 Catania, Italy (e-mail: [email protected]; [email protected]). Digital Object Identifier 10.1109/JBHI.2025.3549255 require further refinement. Smartphone apps provide personalized plans and progress tracking, though most rely on manual logging and lack rigorous scientific evaluation. Our findings suggest that digital health technologies could significantly enhance smoking cessation efforts. However, more robust evaluation methods and integration of sensor data with machine learning are needed to improve usability and effectiveness. Continued research and innovation in this field are crucial for developing reliable, practical solutions and integrating these technologies into clinical programs. Index Terms—Smoking detection, health technologies, smoking cessation, medical mobile apps, technology review, wearable devices. I. INTRODUCTION TOBACCO smoking remains a leading cause of preventable illness and premature death worldwide, despite declining prevalence rates [1],[2]. In the U.K., smoking-related deaths accounted for 16% of all deaths in 2016 [1]. The economic impact of smoking is substantial, with global annual costs exceeding US$500 billion [3]. Smoking behavior is maintained by nicotine’s reinforcing properties and the distant nature of health consequences [2]. Effective interventions to reduce smoking prevalence include tax increases, social marketing, and brief advicefromhealthprofessionals[2].Workplacesmokingcessation programshaveshowncost-effectiveness,withbenefit-cost ratios up to 8.75 and significant employer cost savings [3]. While various cessation measures have proven effective and cost-effective, challenges remain in addressing persistent inequalities in smoking rates among certain groups, such as manual workers and individuals with serious mental illness [1]. Over the past decade, we have witnessed a rapid proliferation of portable devices that have become central to our daily lives [4],[5]. Notably, smartphone technology, coupled with ever-expanding bandwidth connectivity and the growth of social networks, has fundamentally transformed the way we conduct nearly all our daily activities, ushering in an era of pervasive digital technology [4],[6],[7],[8]. In addition to smartphones, there has been a significant uptick in the adoption of various wearable devices and home/office installations [9],[10],[11], [12], all interconnected and controllable through simple smartphone applications. This interconnected device ecosystem is © 2025 The Authors. This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ 5192 IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, VOL. 29, NO. 7, JULY 2025 geared towards enhancing the intelligence of our devices and environments, leading to the emergence of concepts like smart homes and smart offices [13],[14],[15],[16]. Crucially, these wearable and remote devices are equipped with specific sensors that can capture data related to individuals or their surroundings, which can then be shared and processed collaboratively among different devices [17],[18],[19]. The goal is to derive insights and create added value for the user experience, offering opportunities for both data capture and user support. A novel and innovative application of smart device technology lies in its potential to assist with smoking cessation treatments [20],[21], [22]. In the realm of smoking detection and cessation technologies, there is a recognized issue that these technologies are not fully optimized for real-life scenarios. While existing technologies have demonstrated potential, their performance in real-life scenarios continues to pose a challenge. The present work is a scoping review updating a previously published work [17] with the aim of conducting a comparative examination of various smartphoneapplications(apps),wearabletechnologiesdesigned for automatic smoking detection, and other instances where technology can play a role in supporting smoking cessation interventions. In this paper, we aim to provide an overview and analysis of the current state-of-the-art technology focusing on automated smoking detection and smoking cessation technologies. An automatic smoking detection technology is a solution designed to ascertain the number of cigarettes smoked by an individual within a specified observation period [16],[23]. This encompasses approaches that necessitate minimal user intervention (i.e., automatic), encompassing all stages involved in detecting smoking events, from collecting sensory data to making the final inference, as opposed to solutions reliant on self-reporting by participants (e.g., diary apps). In the following sections, the most relevant apps and technologies designed to help people stop smoking are shown and compared. A summary of the revised solutions to help users quit smoking is presented in the Discussion. II. METHODS A. Research Question This scoping review was conducted following the Preferred Reporting Items for Systematic Reviews and MetaAnalyses (PRISMA) Extension for Scoping Reviews (Suppl. Mater.1) [24]. The aim was to synthesize and explore the current applications of digital devices for automatically detecting the use of cigarettes. Additionally, this review will also report on the use of smoking cessation smartphone applications, which represent a significant stride in leveraging technology to aid in smoking cessation. Our study, structured following the PICO format [25], focused on individuals who smoke (P), examining the application of digital health technologies, such as smartphone apps, wearable devices, sensors, and machine learning techniques, to detect smoking events and support smoking cessation (I). These innovative approaches were compared to traditional methods, including manual self-monitoring, standardbehavioraltherapies,ortheabsenceofintervention(C).The outcomes of interest included improvements in the automatic detection of smoking events, such as recall rates and accuracy, increased smoking cessation rates, usability and acceptance of these technologies, and their successful integration into clinical practice (O). B. Systematic Search of Patents A systematic search was conducted on Google Patents to identify patents related to smoking cessation systems and technologies. Google Patents was chosen as the search engine due to its comprehensive coverage of patent databases from multiple jurisdictions, including the United States Patent and Trademark Office (USPTO), European Patent Office (EPO), and World Intellectual Property Organization (WIPO). The search strategy was designed to be broad to capture as many relevant patents as possible. The search terms used were combinations of the following keywords: (“smoking cessation system” OR “automated smoking detection”) AND (“patent” OR “application” OR “method”). The search, unrestricted by date or jurisdiction, screened all results for relevance based on title and abstract. Patents detailing smoking cessation systems or technologies were further analyzed. Additional relevant patents were identified through screening the reference lists of these patents.Data extractedfromeachpatent, includingtitle,number, filing and publication dates, inventors, assignees, abstract, and claims, provided an overview of the latest technology in automated smoking detection and cessation systems. The systematic search results were incorporated into the PRISMA flow diagram (Fig. 1), visually representing the search and selection process for transparency and reproducibility of the study. C. Literature Search An updated search for smoking detection technologies and smoking cessation applications was conducted in December 2023 using PubMed and ACM Digital Library databases. The following search strategy was used: (“smoking” AND “detection system”) OR (“smoking” AND “sensor”) OR ((“smoking” AND “detection system”) OR (“smoking” AND “sensor”)) OR ((“smoking cessation”) AND (“application” OR “app” OR “smartphone app”)). The full search strategy is provided in Suppl. Mater. 2. All the studies published since 2019, year of publication of the previous review, were included. There were no limitations based on language. The reference lists of the included studies underwent additional scrutiny to identify additional potential studies. We manually searched key peerreviewed scientific journals in the field of tobacco research (specifically, Nicotine & Tobacco Research, Tobacco Control, Carcinogenesis, Health Education Research, and Contributions to Tobacco and Nicotine Research). Two authors of the review independently examined and chose studies from the conducted searches. Any disagreements were resolved through discussion or, if necessary, with the involvement of a third reviewer. 1) Web-Based Search for Smoking Cessation Applications: For smoking cessation applications, an additional web-based search was carried out. The selection process was conducted as follows: we performed multiple searches using Bing, Google, CASU et al.: SMOKING DETECTION AND CESSATION: AN UPDATED SCOPING REVIEW OF DIGITAL AND MOBILE HEALTH TECHNOLOGIES 5193 Fig. 1. PRISMA-ScR (Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews) flow diagram representing the article selection process in accordance with the guidelines for updates of systematic reviews [24]. and DuckDuckGo search engines using the query “quit smoking apps.”Fromthe initialsearch results,weexcludedentries related to sponsored links promoting specific apps. Instead, we focused on links associated with blogs dedicated to health topics (e.g., healthline.com). From this resulting list of apps, we chose those with high average user ratings in the Android and iOS app stores (i.e., ratings of 4/5 or higher). Then, a further literature search was conducted on Google Scholar combining the names of app identified via web as follows: (“name of the application” AND “application”). This additional step, which was performed for each application, enabled verification of which apps were clinically assessed. D. Eligibility Criteria The eligibility criteria for the inclusion of studies in this scoping review were as follows: rStudies that reported on development, evaluation, or applicationof adigital ormobilehealthtechnologyfor smoking detection or cessation. rTechnology that involved smartphone, smartwatch, wearable device, or other sensor-based system. rStudies in the form of original research articles (includingrandomizedcontrolled trials),cross-sectional, cohorts, brief reports, case reports, case series communications, methodologies, and methods. rStudies published in peer-reviewed journals or conference proceedings. rStudies published between 2019 and 2023. E. Exclusion Criteria The exclusion criteria were the following: rStudies that did not focus on smoking detection or cessation as a primary or secondary outcome. rTechnology that did not involve sensor or motion data collection or analysis. rStudies not written in English. rApps that were not available in English. rStudies in the forms of abstract, preprint, editorial, commentary, letter, or review. rStudies published before 2019 (except those included in the previous version of this review). F. Data Extraction Two reviewers independently performed data extraction. Any inconsistencies were resolved through discussion or with the assistance of a third reviewer. In our analysis, we categorized the reviewed technologies into two main groups: smoking detection technologies and smoking cessation applications. 1) Smoking Detection Technologies: For each study, the following elements were systematically extracted and compiled 5194 IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, VOL. 29, NO. 7, JULY 2025 in study tables: product, technology, operating system (OS), goal, participants, hours, recall, F-score and Area Under Curve (AUC). 2) Smoking Cessation Applications: For each study, the following items were extracted and adapted into appropriate tables: product, technology, mobile operating systems, scientific evaluation (Sc. Eval.), public availability, and list price. Within each category, we also highlighted whether smoking cessation applications have been supported by scientific articles and whether these have undergone assessment, for example, on a clinical level. Smartphone applications were divided into two subsections based on whether they have been subjected to clinical assessment or not, and therefore at least one phase of testing on real samples. III. RESULTS A. Study Characteristics The search yielded a total of 37 distinct documents, to be added to 12 studies retrieved from the previous version of the review (6 related to smoking detection technologies, and 6 related to smoking cessation applications). Given that the utilization of wearable devices for smoking detection is relatively recent, most of the located works pertain to products currently undergoing evaluation or still in the experimental prototype phase. The article selection process is reported in the PRISMA-ScR, compiled in accordance with the guidelines for updates of systematic reviews (Fig. 1). The full list of included studies is reported in Table I. A total of 26 studies were identified for smoke detection technologies and 20 for smoking cessation applications, 2 of which were subsequently added to suggestions obtained through the review process. Furthermore, 3 studies related to generic smoking detection were included. The main characteristics of the included studies are explained below. The remaining 60 references cited throughout this paper provide contextual background, theoretical framing, or supplementary discussion but were excluded from formal analysis to maintain focus on the core research questions. This approach aligns with scoping review methodologies, which prioritize depth on key themes while acknowledging broader scholarly discourse. B. Technologies for Smoking Events Detection The technologies discussed in this paragraph aim to detect smoking events in real-time, eliminating manual tracking. Some are market-ready, others are under evaluation. They typically use a wearable device and smartphone app to identify smokingrelated movements like hand-to-mouth actions. The development of smoking detection systems has undergone significant advancements, particularly in leveraging wearable technologies and machine learning. Lopez-Meyer et al. [26],[27] laid early foundations using respiratory inductive plethysmography (RIP) sensors and wrist-worn devices to detect smoking gestures (see Fig. 2). Their approach utilized Support Vector Machines (SVM) and threshold-based algorithms, achieving recall rates of 80–90%. However, their system was limited to controlled settings, requiring offline processing Fig. 2. Sensors of the system depicted in Lopez-Meyer et al.’s work [26]. and providing minimal adaptability to free-living environments. Moving forward, systems like SmokeBeat [28] enhanced detection by incorporating accelerometers and gyroscopes into commercial smartwatches. SmokeBeat combined probabilistic models with gesture segmentation, yielding precision and recall ratesexceeding85%.Similarly,RisQ[29]leveragedConditional Random Fields to sequence smoking gestures in free-living conditions,whileStopWatch[30]adoptedRandomForestclassifiers to distinguish smoking from other activities. These systems demonstrated the potential for low-cost, user-friendly platforms, achieving operational accuracies between 70–90%. Post-2020 studies demonstrated remarkable advances in smoking behavior detection through increasingly sophisticated methodological approaches. Senyurek and colleagues [31] developed a wearable system integrating respiratory inductive plethysmography (RIP) and inertial measurement unit (IMU) sensors, employing a hybrid deep learning framework combining convolutional neural networks (CNN) and long short-term memory (LSTM) networks. Their research utilized a comprehensive dataset evaluated through leave-one-subject-out crossvalidation, achieving an F1-score of 78%. Similarly, Kirmizis et al. [32] employed an artificial neural network with convolutional and recurrent layers, utilizing the Smoking Event Detection (SED) and Smoking Event Detection Free-Living (SEDFL) datasets. Their two-step methodology leveraged smartwatch data to detect individual puffs and localize smoking sessions, achieving impressive weighted accuracies of 0.968 and F1-scores of 0.878. Agac et al. [33] advanced sensor fusion methodologies, utilizing accelerometers and gyroscopes from smartwatches (LG Watch R, LG Watch Urbane or Sony Watch 3) and smartphones (Samsung Galaxy S2 or S3). Their framework incorporated user-specific features, such as body dimensions, into a Random Forest classifier to achieve 83% recall in distinguishing smoking gestures from other hand-to-mouth activities. They validated the model using a comprehensive dataset collected under free-living conditions, demonstrating the importance of personalization in wearable systems. More recent advancements, such as Hnoohom et al.’s [34], advanced CASU et al.: SMOKING DETECTION AND CESSATION: AN UPDATED SCOPING REVIEW OF DIGITAL AND MOBILE HEALTH TECHNOLOGIES 5195 TABLE I SUMMARY OF THE STUDIES INCLUDED IN THIS SCOPING REVIEW,LISTEDINTHEORDER THEY APPEAR IN THE TEXT,REFLECTING THEIR RELEVANCE TO DIFFERENT TOPICS AND SECTIONS OF THE ARTICLE smoking gesture detection through a sophisticated ResNetSE framework, integrating deep residual networks with attention mechanisms. By analyzing the UT-Smoke dataset collected from 11 volunteers over three months, the researchers compared their approach against five baseline models (CNN, LSTM, BiLSTM, GRU, and BiGRU; see Fig. 3). The ResNetSE model demonstrated exceptional performance, consistently achieving top accuracy and F1-scores of 98.65%, 98.39%, and 98.63% across multiple scenarios, highlighting its superior capabilities in real-time gesture recognition. Thakur and colleagues [35] developed a robust activity recognitionframeworkusinga6-axisinertialmeasurementunit(IMU) sensor, exploring multi-class classification models including Logistic Regression, k-Nearest Neighbor, Adaptive Boosting, Random Forest, Support Vector Machine, and Decision Tree. Maguire et al. [21] introduced a particularly innovative multimodal system combining a smartwatch (with accelerometers and gyroscopes) and a wearable finger sensor, and an Android app (Fig. 4), using a TensorFlow Lite model for activity classification. Their smartwatch-only system achieved accuracy improvements from 75.8% to 85.5% by integrating the finger sensor. Furthermore, Sharma et al. [22] advanced the field with a microcontroller-based system employing a convolution-based network and Neural Architecture Search (NAS) to develop custom Deep Neural Network (DNN) models. Mukhopadhyay’s research [36] utilized reinforcement learning to optimize CNN 5196 IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, VOL. 29, NO. 7, JULY 2025 Fig. 3. ResNetSE model included in Hnoohom and colleagues’ work [34]. TABLE II SUMMARY OF THE SMOKING DETECTION TECHNOLOGIES DESCRIBED IN THIS ARTICLE Fig. 4. Diagram of smoking cessation technology described by Maguire and colleagues [21]. architectures, achieving a puff detection F1-score of 0.81, while Alharbi et al. [37] introduced SmokeMon, a chest-worn thermalsensing system that demonstrated high-precision smoking event detection across laboratory and real-world environments. A summary of these studies and their recall rates can be seen in Table II, while a summary of the smartwatches employed in them can be found in Table III. TABLE III SUMMARY OF THE COMMERCIALLY AVAILABLE-TO-PUBLIC SMARTWATCH EMPLOYED IN THE STUDIES INCLUDED IN THIS SCOPING REVIEW C. Smoking Cessation Applications The applications delineated in this section are founded upon the annotation of smoking behaviors, accomplishments, and instances of craving. These applications have attained exceedingly elevatedmean feedbackratings fromusersand have experienced substantial rates of downloads within the application markets. 1) Clinically Assessed Applications: Smoking cessation applications highlights a robust and innovative landscape of digital interventions, each employing unique strategies to support users in their quitting journeys. Through scientific third-party evaluation or endorsement, these applications demonstrate a commitmenttointegratingevidence-basedmethodologies,personalized CASU et al.: SMOKING DETECTION AND CESSATION: AN UPDATED SCOPING REVIEW OF DIGITAL AND MOBILE HEALTH TECHNOLOGIES 5197 features,andadvancedtechnologiestoenhanceuserengagement and efficacy. Chen et al. [38] showcased an Android-based system combining wearable sensors and tailored quitting plans, emphasizing personalization through demographic and behavioral data. This approach aligns closely with mindfulness-based strategies, such as the RAIN method [39], to aid in managing cravings, while integrating supportive messaging for both users and their social networks. Pivot, a widely evaluated program, stands out for its incorporation of an FDA-cleared breath sensor that offers realtime physiological feedback, alongside a comprehensive app providing customized lessons, progress tracking, and coaching. Its compatibility with iOS and Android platforms, coupled with multiple clinical validations [40],[41],[42],[43], underscores its scalability and effectiveness. Similarly, CureApp Smoking Cessation (CASC) integrates behavioral and pharmacological therapies with a mobile exhaled CO checker, demonstrating significant improvements in abstinence rates and reductions in nicotine dependence compared to control groups [44]. This app exemplifies the potential of hybrid digital and pharmacological approaches. Apps like SmartStop and Craving To Quit! focus on combining technology with behavioral science. SmartStop leverages a programmable nicotine patch synchronized with a smartphone app to address peak craving periods, while Craving To Quit! integrates cognitive behavioral therapy (CBT) and mindfulness practices to disrupt smoking patterns [45],[46]. These interventions highlight the interconnected physiological and psychological dimensions of smoking cessation. By examining how stress compromises prefrontal cortex function and increases smoking vulnerability, researchers illuminate the neurological underpinnings of addiction [47]. Mindfulness therapy offers a promising approach to modulating desire and cigarette use, revealing the complex mechanisms that sustain tobacco dependency [48]. Apps employing gamification and interactive features, such as Clickotine, Smoke Free, Kwit, and Quit Genius, have demonstrated effectiveness in enhancing self-efficacy and motivation through rewards systems, progress tracking, and engaging challenges [49],[50],[51],[52],[53],[54],[55],[56]. In this regard, QuitSTART exemplifies another facet of smoking cessation support, combining progress tracking with strategies to manage cravings and negative moods. The app employs user data to offer personalized challenges, advice, and motivation, ensuring an interactive and engaging cessation journey [57]. Gamified elementsappearparticularly influentialin fostering userengagement and addressing cognitive factors critical to quitting. Notably, Acceptance and Commitment Therapy (ACT) has emerged as a recurrent theme, underpinning the design of 2MorrowQuit, SmartQuit, and iCanQuit [12],[58],[59]. These apps leverage ACT principles [60] to build psychological flexibility, mitigate cravings, and promote mindfulness, with promising outcomes in abstinence rates and behavior modification. Finally, Quit Genius and other CBT-based apps demonstrate a holistic approach, addressing not only smoking cessation but also broader addiction challenges. Their integration of personalized plans, extensive CBT exercises, and supportive communities reflects a comprehensive strategy aimed at sustaining long-term change [61],[62]. 2) Applications Not Yet Clinically Assessed: Several publicly available apps aid in smoking cessation and do not, yet, provide clinical assessment.The LIVESTRONG MyQuit Coach and Quit Smoking: Cessation Nation offer goal-setting and community support. Quit Now! provides motivational messages and supports multiple languages. The Quit Smoking with Andrew Johnson app uses self-hypnosis, while Butt Out provides insights for a smoke-free lifestyle. Get Rich or Die Smoking motivates through monetary incentives, and SmokeFree—Quit Smoking Slowly offers options to quit abruptly or gradually. The Quit Smoking NOW—Max Kirsten app uses hypnosis and NLP techniques, and the Quit Tracker: Stop Smoking app displays financial savings and health benefits. Quit It Lite helps users set personalized goals, and Quit Smoking Hypnosis offers daily hypnosis sessions. Quitter’s Circle supports smoking cessation with resources and a quit fund tool. EasyQuit provides a personalized quit plan and a distraction game. Flamy offers personalized plans and rewards, and Smoking Log helps reduce cigarette consumption. All these apps are available on iOS and Android, with some offering premium features. This subsection provided a brief summary of some smartphone apps designed for smoking cessation that lack published peer-review. A more complete list can be found in Table IV. 3) Smartphone Apps Limitations: Engagement with smartphone apps, particularly those designed for smoking cessation, faces several limitations. One key issue is the lack of personalization and adaptability to the user’s changing needs and contexts, which can lead to decreased engagement over time [63],[64]. Moreover, many apps do not adequately assess the user’s readiness to quit smoking or arrange follow-up, which are crucial for maintaining engagement [65]. Improving engagement could involve incorporating more user-centered design principles, such as real-time messaging with support networks and reducing barriers to access [63]. Furthermore, the use of assessment tools like the Mobile Application Rating Scale (MARS) can provide valuable insights into app quality, includingengagement,functionality,aesthetics,andinformation quality [66],[67],[68]. However, it’s important to note that commercialization of apps does not necessarily imply widespread availability to the general public. For instance, the Pivot App operates on a B2B model [69],[70],[71], which may limit its accessibility to only certain organizations or groups. Therefore, while commercial apps may be widely marketed, their actual accessibility may be more limited [72],[73]. 4) Apps Usability: The usability and convenience of the described applications play a crucial role. While a generic application may achieve high performance in terms of smoking detection, it could prove inconvenient to use. A fundamental distinction exists between apps that provide information in a standalone manner, without the need for additional devices, and those thatoperate withmultimodal informationfrommultiple sources. Table IV illustrates that most of the described applications do not require additional devices, making them user-friendly tools. However, it is evident that applications utilizing supplementary information, such as SmartStop, Pivot, and CureApp Smoking 5198 IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, VOL. 29, NO. 7, JULY 2025 TABLE IV SUMMARY OF THE QUIT SMOKING APPLICATIONS DESCRIBED IN THIS STUDY,INCLUDING THOSE WITH “FREEMIUM”MODELS,WHERE THE APP IS FREE TO DOWNLOAD AND USE,BUT OFFERS ADDITIONAL PAID FEATURES OR CONTENT Cessation, achieve higher performance, following the principle that more data equates to greater knowledge and, consequently, better performance. On the other hand, however, the use of additional devices is in some cases an unfeasible and in others an inconvenience that discourages their long-term use. Therefore, although it will always be easier to use a stand-alone application and simpler to use a multimodal system, the correct trade-off depends on the scenarios where it is to be used. D. Nicotine Detection in Smoke Detection Systems Tai et al. [74] introduced the “s-band,” a wearable nicotine sensor employing a gold nanodendrite-modified working electrode and a self-assembled monolayer, enabling high sensitivity and stability in detecting nicotine from human sweat. Validated in both buffer solutions and real-world samples from smokers, the sensor reliably identified nicotine levels consistent with cigarette nicotine content, highlighting its potential for public health and personalized medicine applications. Rani et al. [75] advanced the field with a metal-organic nanotube (MONT) sensor capable of selectively detecting nicotine in cigarette smoke at concentrations below 23.3 µM. The MONT’s porous structure, combined with rapid response times (20 seconds) and sunlight stability at room temperature, allows for efficient nicotine detection through visible light-driven binding to metal ions. Its reusability after heating at 110◦C under vacuum enhances cost-effectiveness and practicality across gaseous and solution-phase applications. Meanwhile, Rahman et al. [76] developed awireless, battery-free,skin-mountednicotinesensor using vanadium dioxide (VO2) technology to detect nicotine vapor from e-cigarettes. By leveraging electron transfer between nicotine molecules and the VO2 surface, this sensor achieves precise vaporized nicotine detection, supported by density functional theory (DFT) calculations and compositional analysis. Its lightweight design facilitates continuous monitoring for both personal and environmental use. E. Non-Wearable Smoking Detection: Deep Learning, Wireless Signals, Trials, Dataset, and Gesture Detection 1) Smoking Detection Through Vision: Recent advancements in non-wearable smoking detection have made significant strides, leveraging deep learning techniques and novel system designs to enhance accuracy and efficiency. Macalisang et al. [77] developed a smoking detection system using a dataset of 300 images and the YOLOv3 model, achieving high training and validation accuracies of 98.10% and 98.22%, though challenges with detection angles and video quality remained, with accuracies varying from 63% to 98% in real-world testing. Weietal.[78] expanded this work by building a larger dataset of 9,424 smoking images and employing data augmentation techniquessuchasMosaicenhancement,whichimprovedgeneralization. Their model, optimized with the DIoU_Loss function and adjusted learning rates, showed enhanced performance, particularly in Average Precision (AP) and Intersection over Union (IoU), underscoring the model’s robustness. Zhang et al. [79] introduced CBAM-Tiny, a lightweight attention mechanism designed to improve small target detection by refining spatial features with global pooling and utilizing a custom DenseBlock module for better gradient flow. Their model achieved an mAP of 86.32% and a frame rate of 55 frames per second, demonstrating both precision and speed, which is crucial for real-time applications. Finally, Chong [80] developed a real-time system utilizing the Real-Time Streaming Protocol (RTSP) to capture video frames and process them through a custom model trained on the Tsinghua-Tencent 100 K dataset. This system employed CASU et al.: SMOKING DETECTION AND CESSATION: AN UPDATED SCOPING REVIEW OF DIGITAL AND MOBILE HEALTH TECHNOLOGIES 5199 Non-Maximum Suppression (NMS) and a context information correlation algorithm to improve detection accuracy and processing speed, outperforming models like YOLOv3, SSD, and RetinaNet. 2) Detection of Non-Cigarette Smoke: Gaur et al.’s review [81] explores smoking detection, discussing challenges with smoke obscuring data and the features used in algorithms. They highlight the need for dataset testing and advanced methods like quaternionic wavelet features, Kalman filtering, and transmission-based detection. Xu and Xu [82] combined static anddynamicfeaturesforAI-baseddetection.Saponaraetal.[83] used deep learning for real-time fire and smoking detection, leveraging the NVIDIA Jetson Nano’s CPU (Central Processing Unit) and GPU to parallelize neural networks. They focused on the YOLOv2 detector, achieving a detection rate of 21 FPS. Gu et al.’s study [84] evaluates a Deep Dual-Channel Convolutional Neural Network (DCNN) for smoking detection, which outperforms other models in terms of stability and efficiency. The DCNN surpasses processing times of other models and excels at extracting detailed and basic features. These studies collectively usher in a new era in algorithm-driven fire and smoking detection. 3) Smoking Detection Using Deep Learning: Jeong and Ha [85] explored a deep learning-based system for smoking detection using CCTV footage, integrating OpenPose-based skeletonanalysiswithspecializedhardwareforenhancedrecognition. Their system preprocesses image data to recognize smoking behavior, coupling it with sensor-equipped devices to detect smoke components, triggering warnings for non-smoking areas. A neural network built with TensorFlow and Keras, optimized with MobileNetV2, achieves 75% accuracy for smoking images and 70% for non-smoking images, offering a promising real-time smoking detection framework. On a different front, Lai et al. [86] focused on smoking cessation by leveraging data from a program in northern Taiwan spanning from 2010 to 2018. Using machine learning models like artificial neural networks (ANN), support vector machines (SVM), random forests (RF), andothers,theyaimedto predictsmokingcessationprobabilities based on factors such as patient characteristics, smoking habits, and nicotine dependence scores. The ANN model outperformed others with an accuracy of 0.640 and an ROC value of 0.660, offering a valuable predictive tool for smoking cessation programs. Both studies contribute to the understanding of smoking behavior and cessation, with Jeong and Ha’s work enhancing real-time detection through image processing and hardware integration, while Lai et al.’s research provides insights into machine learning’s potential in predicting successful smoking cessation. 4) Human Behavior Detection With Wireless Signals: Song et al. [10] developed a contactless AI technology using Channel State Information (CSI) from wireless signals to detect human motion,focusingon distinguishingbetweensittingand standing. Theyused USRPdevices tocollect CSI datafrom volunteers and analyzed it using MATLAB and scikit-learn. Machine learning models were built and tested, with Random Forest (RF) performing well and K-Nearest Neighbors (KNN) being less effective. An ensemble classifier improved performance, and the CSI dataset outperformed a benchmark dataset. The model was effective inpractical applications,with local testsprovidingGUI predictions and real-time tests offering CSI amplitude graphs and web interface predictions. 5) Smoking Detection Trials: The smoking detection trials conducted across various studies demonstrate the potential for integrating real-time, personalized interventions in smoking cessation. Battalio et al. [87] utilized a Just-in-Time Adaptive Intervention (JITAI) model to help smokers manage stress, a key trigger for relapse. The study incorporated multiple sensors, including chestbands and wristbands, to gather physiological and behavioral data for real-time analysis. By using stress-detection algorithms, the system provided individualized treatment options, such as stress management prompts, to prevent smoking episodes during high-stress moments. In a similar vein, Hernandez et al. [88] focused on the feasibility and effectiveness of mindfulness-based interventions delivered via wearable sensors that tracked physiological indicators associated with negative affect,self-regulation,andsmokingbehaviors.Usingdeeplearning techniques, the study personalized interventions based on real-time data, offering a more dynamic and tailored approach to smoking cessation. Horvath et al. [89] explored the effectiveness of a smartband-based system that provided automatic smoking detection and mindfulness interventions, including the RAIN technique, which was tailored to help participants recognize and manage cravings. In this trial, data collected from the wearable devices were used to assess treatment fidelity, adherence, and user satisfaction, with smoking behavior and abstinence rates also being tracked. 6) Gesture Detection: Gesture detection has evolved through various approaches, each contributing to the accuracy and efficiency of activity recognition systems. Hnoohom et al. [90] developed an innovative Human Activity Recognition (HAR) workflow incorporating data collection fromwearables,pre-processing,modeltraining,andassessment. They introduced the Att-BiLSTM model, which integrated a BiLSTM layer, an attention layer, and a fully connected layer, demonstrating superior performance on the WISDM-HARB Dataset. This model achieved higher accuracy and F1-scores when combining wrist-worn accelerometer and gyroscope data with a 20-second window size, evaluated using metrics such as F-Score, Recall, Precision, and confusion matrices. In contrast, Agac et al. [91] focused on a dynamically adaptable parameter selection method with the Conawact algorithm for activity recognition, which tailored sensor parameters based on activity complexity. This dynamic approach significantly improved the F1-score by 7% for complex activities and by 6% overall, while also reducing energy consumption by 38%, maintaining memory size, and lowering CPU usage by 15%. Their method proved to be particularly effective for activities like “smoking in a group” and “drinking while sitting down,” showing improvements of over 20%. Meanwhile, Patel et al. [92] explored 3D gesture recognition through wearables, emphasizing the integration of sensor data from smartwatches and armbands with image/video data. Their work aimed at improving human-machine collaboration, with a focus on gesture and pattern recognition to enhance