The first step in the development process involves the translation of sections of the system that participants interact with into Malaysia’s 3 major languages: Bahasa Melayu (the national language) [21], Mandarin (the most widely spoken Chinese dialect in Malaysia) [22], and Tamil (commonly used within the Tamil community) [23]. Machine translation tools, such as Google Translate and Bing Translate, as well as AI-based assistance (eg, ChatGPT), will be used to provide initial drafts that bilingual translators will further refine. Subsequently, a team of bilingual reviewers will collaborate with the translators to further improve readability, consistency, and contextual relevance.

The development of the Intake24-MY food list process will commence with a review of the Australian Intake24 food list, which comprises 2226 items. Globally familiar items, such as ice cream, bread, and eggs, will be retained, whereas nonrelevant items will be removed. The retained food and drink items will be updated with relevant food composition data to reflect local values. Subsequently, 680 prepared or cooked foods and drinks from the Malaysian Food Composition Database (MyFCD) [24] will be incorporated. To address remaining gaps, additional food items will be identified through expert knowledge, popular restaurant menus, food blogs, food delivery apps, and major grocery outlets.

Intake24 offers 4 portion-size estimation options: as-served images, guide images, a sliding scale for drinks, and standard portion descriptions [25]. As-served images show a series of 7 photographs depicting incremental food portions, as shown in Figure 2. In contrast, guide images illustrate size differences for fixed items, such as cream buns, Malaysian kuih, and hot-drink mugs, as shown in Figure 3.

A total of 527 images and 160 standard portions from the Australian Intake24 database will be retained for use in Malaysia, particularly for common items such as packaged snacks and beverages, as well as for items that use similar drinkware and tableware. Where applicable, one-to-one visual matches, such as rice on a plate, ice cream in a bowl, and peanut butter spread on bread, will also be reused. To represent local foods not found in the Australian database, 210 new image sets, comprising as-served and guide images, will be created using an iPhone 15 Pro, a light box, and standard-size crockery, following standard guidelines [26].

A subset of newly developed food photographs will be selected for validation. The selected foods will represent a range of appearances, textures, consistencies, and portion characteristics, including beverages, liquid foods, mixed dishes, solid foods with undefined portion sizes, foods served as distinct pieces, single-item foods, preportioned foods, and foods commonly eaten in combination or as part of a meal [27].

A small-scale validation of the food photographs will be conducted using the “food-in-front” technique. Participants will be asked to fill out a form to estimate the portion sizes of foods placed in front of them using the corresponding portion-size images (Multimedia Appendix 1). Their estimated weights will then be compared with the actual measured weights of the foods [28]. Based on the results, a decision will be made on whether to include the image series in Intake24-MY or to use a standard portion option instead.

The food lists will be linked to the appropriate food composition tables. As MyFCD mainly includes generic foods and a limited number of prepared dishes [24], additional data will be sourced from databases in Singapore, Indonesia, and other countries, prioritized by relevance [29].

MyFCD provides macronutrient data for most foods, but micronutrient information is limited. Where no suitable match exists, recipe calculations will be used to estimate nutrient content. Food matching will follow the Food and Agriculture Organization guidelines [30], with quality codes and source references recorded. For unmatched dishes, detailed recipes with ingredient weights and cooking methods will be used, applying nutrient retention and yield factors [30,31]. These data will be compiled into a new food composition database.

The feedback component will be adapted to reflect the Malaysian Dietary Guidelines [32] and will be provided to users upon completion of their dietary recalls. The feedback will be age-specific and sex-specific and will provide culturally relevant guidance on potential dietary adjustments to promote healthier eating patterns.

A screen-capture walkthrough will be developed to guide users through the Intake24-MY system. It will demonstrate various tasks within the system, such as searching for food items, matching foods within the database, reporting portion sizes, using the “same as before” option, prompting for associated foods, and submitting dietary recalls. A voice-over explaining the process will be recorded in each of the 3 languages: Bahasa Melayu, Mandarin, and Tamil. Additionally, frequently asked questions (FAQ) documents in all 3 languages will also be prepared, featuring screenshots to support user navigation.

A small-scale system test will be conducted first to identify any missing food items, technical issues, and user challenges. This will be followed by pilot testing, where participants will complete 2 nonconsecutive weekday dietary recalls using Intake24-MY and an interviewer-led 24hDR. The participants will be asked to provide feedback on language, missing food items, and system usability. System refinements will then be implemented by developers, prioritizing critical issues based on their impact and severity.

Intake24-MY will adopt a user-centered, iterative approach to system localization, consistent with previous adaptations [33]. Each development stage will be followed by expert review, user testing, and system refinement, as shown in Figure 4.

Different methods will be applied at each stage of the iterative process. Translation will begin with machine-generated drafts, which will be refined by bilingual experts and adjusted based on user feedback. The food list will first be compiled from existing data sources, and then expanded to include any missing or culturally specific items. Portion-size images will undergo photograph validation to assess their appropriateness. The iterative cycle will continue through small-scale testing, pilot testing, and relative validation phases.

Participants will be asked to estimate their meal intake using Intake24-MY and compare it with direct observation. On the first day, they will be served a single meal and asked to consume it while leaving any leftovers at the study site. The next day, participants will receive an email with a link to log into Intake24-MY to complete a recall of the previous day’s lunch. They will be asked to identify the food and drink items and their corresponding portion sizes. Trained researchers will observe the participants, weigh the foods served, and measure any leftovers.

Data collected using Intake24-MY will be downloaded and compared with the food and drink items and portion sizes recorded by the observer for each individual. Foods will be coded as matches, omissions, or intrusions. When a match is found, the accuracy of the portion-size estimates will be assessed by calculating the ratio of the food weight reported in Intake24-MY to the weight measured by the observer.

Participants will first provide demographic information and then complete a dietary recall using Intake24-MY and an interviewer-led 24hDR. The interviewer-led 24hDR will be conducted using the externally validated multiple-pass method [34] and will be used as the reference method for comparison with Intake24-MY.

For the pilot study and relative validation, the dietary data will be collected over 2 nonconsecutive days and 4 nonconsecutive days, including 1 weekday and 1 weekend day, across 2 and 3 weeks, respectively. On each assessment day, dietary intake during the previous 24 hours will be assessed using both methods with the same reference day. Participants will receive a unique link by email or text message to access the Intake24-MY system. They will follow the system instructions to complete the dietary recalls. For the interviewer-led dietary recall, the first session will be conducted face-to-face, with participants completing dietary recall under the interviewer’s guidance. Subsequent dietary recall interviews will be conducted by phone. To minimize potential order effects [33], participants will be randomly assigned as a counterbalance measure: 75% will begin with Intake24-MY, followed by the interviewer-led recall to evaluate sequential performance, while 25% will start with the interviewer-led recall, followed by Intake24-MY. Once they finish their final recall, participants will be asked to give feedback (Multimedia Appendix 2) and assess usability through a structured questionnaire (Multimedia Appendix 3).

Dietary data collected from the interviewer-led 24hDRs will be analyzed using Nutritionist Pro software (Axxya Systems). Reported foods and beverages will be coded and matched to food items available in the software database, supplemented with MyFCD values, regional food composition databases, and recipe calculations where required. Portion sizes reported during the interviews will be converted to grams before nutrient analysis. For mixed dishes, ingredients and serving sizes will be estimated based on participant-reported recipes, standard recipes, or commonly used preparation methods.

A multiple-pass protocol, developed based on the USDA Automated Multiple-Pass Method [34] and adapted to the Malaysian context, will be used together with a 24hDR form (Multimedia Appendix 4). The interviewer-led 24hDR will follow a standardized 5-pass approach.

Participants will first provide a quick list of all foods and beverages consumed on the previous day, followed by prompts to identify any forgotten items. Contextual information, including the time of intake, eating location, and source of food, will then be collected. A detailed pass will gather information on food descriptions, cooking methods, brand names, ingredients, and serving sizes for mixed dishes, as well as portion-size estimates, using food photographs, flipcharts, and standard household measures. Finally, the interviewer will review the recall with the participant and probe for any additional items before closing the session.

These structured procedures, including forgotten-food probes, portion-size estimation aids, interviewer training, and a final review, are intended to minimize recall bias, misreporting, and interviewer-related variation. Data collectors will receive comprehensive training to ensure the consistent and accurate administration of interviewer-led dietary recalls. The Malaysia Food Album [35] will be used to aid portion-size estimation. A complete interview guide is shown in Figure 5, and the training procedure is shown in Figure 6.