Neurodegenerative disorders including Alzheimer’s disease and other forms of dementia are becoming increasingly prevalent worldwide. Memory decline that compromises a person’s quality of life is one of the earliest and most prominent clinical symptoms (Jahn, 2013). Consequently, with the aim of providing effective treatment, substantial research efforts have focused on developing therapeutic strategies that can ameliorate cognitive deficits. In preclinical studies, such interventions are typically assessed for their impact on memory function in rodent models using behavioural paradigms. Distinct aspects of memory can be evaluated through well-validated tests such as Y-maze, T-maze, novel object recognition test and Morris water maze (Ghafarimoghadam et al., 2022 and Khan et al., 2025).
       
The passive avoidance (PA) test is also widely used to evaluate both short-term and long-term memory based on fear-motivated learning (d’Isa et al., 2014). The PA is a two-chamber apparatus comprising an illuminated compartment and an adjacent dark compartment. The dark chamber is equipped with an electrifiable grid floor that can be activated to deliver a mild foot shock. In this task, rodents learn to suppress their innate preference for dark, confined spaces by avoiding the dark compartment where this aversive stimulus was previously delivered during the acquisition session. However, the relatively high cost associated with specialized equipment poses a considerable challenge especially in developing countries, thereby constraining comprehensive assessment of this memory domain in in vivo studies.
       
Impairments in hippocampal CA1 function have been closely associated with deficits in PA performance, reflecting its critical role in encoding and retrieval of aversive contextual information (Min-Kaung-Wint-Mon and Mitsushima, 2025). Scopolamine, a muscarinic acetylcholine receptor antagonist, can be used to induce a rodent model of cognitive impairment (Choudhury et al., 2022). Scopolamine has been extensively demonstrated to disrupt cholinergic neurotransmission and induce substantial neuronal loss in CA1 and CA3 regions, leading to hippocampal-dependent memory dysfunction (Ansari et al., 2025). Therefore, this study aims to develop a PA apparatus integrated with software for behavioral recording and analysis at a low cost. The system was designed to generate visualized heat maps of locomotor activity and automatically calculate step-through latency. Scopolamine-induced amnesia was then used as a pharmacological model to validate the sensitivity and performance of the constructed low-cost PA apparatus in assessing memory decline.

Materials
 
The materials used for each component, along with their respective dimensions, are summarized in Table 1. The apparatus comprises two main components, including the chamber assembly and control cabinet, both of which were fabricated using the machining procedures described in Table 2.




 
Apparatus design
 
The apparatus is a two-compartment system comprising a dark chamber fabricated from 3D-printed acrylonitrile butadiene styrene (ABS) and a light chamber constructed from 5 mm acrylic sheets, with overall dimensions of 40 × 20 × 30 cm (length × width × height).  The two compartments are separated by a 6 × 4 cm (length × height) sliding door made of Formex. The control cabinet houses an embedded computer system (Raspberry Pi 5, Fig 1a) integrated with a camera module for animal monitoring and behavioral recognition. The Raspberry Pi Camera Module 3 is specifically designed for use with Raspberry Pi devices (Fig 1b). It connects directly to the Raspberry Pi via the camera serial interface, enabling high-resolution image capture and video recording without requiring any intermediate hardware.


 
Stimulation circuit
 
A 5 V-10 A relay module is also integrated into the circuit to control the on-off switching of the current supplied to the load in response to control signals from the Arduino (Fig 1c). The relay operates at 5 VDC and is capable of switching high alternating current (AC) loads. The electrical stimulation circuit was designed with two main components. The power supply consists of a 24 VAC, 3 A adapter. Voltage regulation is achieved using a TRIAC BT136. The TRIAC modulates the AC voltage and current delivered to the load, thereby enabling precise control of the stimulation intensity.  The load comprises a fixed resistor and a variable resistor. The fixed resistor is incorporated to prevent the current from dropping to 0 A, ensuring stable circuit operation, whereas the variable resistor allows fine adjustment of the current intensity according to the experimental requirements (Fig 1d, e).
 
Data acquisition and annotation
 
Image data were acquired using a Raspberry Pi camera module 3 with a resolution of 1080p and a frame rate of 30 FPS, mounted above the two compartments. Data were collected through two experimental phases: training trial and test trial. Images were extracted at every third frame to reduce redundancy.
       
Image labelling was performed using RoboFlow with YOLO TXT format, with two object classes defined: (a) lab_mouse, representing the experimental animal and (b) border, representing the physical boundary between the two comparments. The animal’s location was determined based on its spatial relationship to the detected border and classified as being in the light chamber, dark chamber or at the doorway. Each bounding box was drawn to enclose the entire body of the animal.
 
Model training
 
The dataset was organized according to the standard Ultralytics YOLOv11 directory structure. Given the fixed top-mounted camera position, augmentation was applied to dataset to reflect realistic experimental variations. All images were resized to 640 × 640 pixels and split into training (70%), validation (20%) and test (10%) subsets using a stratified strategy to ensure balanced class distribution. Subsequently, the model was trained on Google Colab using the Ultralytics framework, with continuous monitoring of loss functions covering localization, classification and objectness. Model performance was evaluated using precision, recall and mean average precision (mAP) to assess the accuracy and reliability of the model.
 
Heatmap generation
 
A heatmap was constructed based on the locations of the detected bounding boxes to visualize the spatial distribution of animal movement within the apparatus. Each image was divided into a uniform grid and the frequency of bounding box occurrences within each cell was calculated and mapped onto a color scale. Red regions indicate higher occupancy whereas blue regions indicate lower occupancy.
 
Animals
 
Female Swiss albino mice (6-8 weeks old, weighing 20-25 g) were purchased from the Pasteur Institute, Ho Chi Minh City, Vietnam. Animals were acclimatized for 14 days prior to experimentation under controlled environmental conditions (temperature 23±2°C, relative humidity 30-40%, 12 h light/dark cycle) as previously described for our animal housing conditions (Trinh et al., 2026a). Mice were housed in polypropylene cages (36 × 30 × 20 cm), with 4-6 animals per cage and had access to standard laboratory chow and water ad libitum. All procedures involving animals were approved by the Animal Ethics Committee of Ton Duc Thang University (Approval No. 17/HDTVDD). The experiments were carried out at Ton Duc Thang University, Ho Chi Minh City, Vietnam, between July 2025 and July 2026.
 
Experimental validation
 
To validate the constructed PA apparatus as a tool to evaluate memory, scopolamine was administered intraperitoneally at 1 mg/kg once daily for six consecutive days, following a previously reported protocol for inducing memory impairment in mice (Kim et al., 2016). Following habituation, mice were randomly assigned into two groups: (a) control group (n = 8): mice received 0.9% saline (i.p.) and (b) scopolamine group (n = 8): mice received scopolamine at a dose of 1 mg/kg (i.p.).
 
Passive avoidance (PA) test
 
On days 5 and 6, mice were subjected to the PA test 30 mins after each injection. The training trial (day 5) commenced with placing each mouse individually in the light compartment. After a 5 s delay, the sliding door was opened, allowing the animals to voluntarily enter the dark compartment up to 300 s. Upon complete entry, the door was automatically closed and a mild electric foot shock (1.3 mA, 3 s) was delivered through the grid floor 2 s thereafter. Mice remained in the dark compartment for an additional 120s before being returned to their home cage. The test trial, conducted 24 hours later, followed the same procedure as the training day, however, no electric shock was delivered. Memory performance was assessed by measuring the step-through latency, defined as the time taken for the animal to enter the dark compartment, with a maximum cut-off time of 300 s.
 
Hematoxylin and eosin (H and E) staining
 
Mice were anesthetized with a single intraperitoneal injection of avertin (300 mg/kg). They were subsequently subjected to transcardial perfusion by initially flushing with phosphate-buffered saline (PBS, pH 7.4) followed by 4% (w/v) paraformaldehyde (PFA). Brains were carefully removed, sectioned into 3-mm blocks and post-fixed in 4% PFA for 24 hours at room temperature. Hippocampus-containing block was then processed for dehydration and embedded in paraffin and coronal serial 10 µm-thick sections were stained with H and E according to our previous protocols (Trinh et al., 2026b and Mai et al., 2026).
  
Immunofluorescence
 
Paraffin-embedded sections of 10 µm thickness were deparaffinized, rehydrated and then subjected to antigen retrieval. Sections were incubated in blocking solution [10% (v/v) fetal bovine serum, 0.1% Triton X-100 in PBS, pH 7.4] for 1 hour at room temperature. Rabbit anti-brain-derived neurotrophic factor (BDNF) primary antibody (1:100, Elabscience E-AB-18244) applied to sections overnight at 4oC. Sections were then incubated with goat anti-rabbit secondary antibody conjugated to AlexaFluor-488 (1:500, Servicebio) for 2 hours at room temperature. Nuclei were labelled using DAPI (1:1000, Invitrogen D1306).  After washing in PBS-Tween [0.1% (v/v) Tween-20 in PBS, pH 7.4] for 15 mins, sections were air-dried and coverslipped with fluorescent mounting medium (Invitrogen). 20× images were acquired using an Axiovert 5 microscope with Zen software (Carl Zeiss).

Statistical analysis
 
All statistical analyses were performed using GraphPad Prism 8.0 software (San Diego, CA, USA). Data are presented as mean ± SD. Step-through latency and BDNF expression data were analyzed using two-way ANOVA followed by Tukey’s post-hoc test. Statistical significance was defined as p<0.05 and is indicated in each figure legend.      

Construction of PA apparatus and model training
 
The PA test was constructed as a two-compartment system consisting of a light and a dark chamber separated by an automated sliding door. In this study, the apparatus was designed to closely resemble commercially available systems but at a low-cost, incorporating an automated gate and a computerized system to track the location of the mouse. The dark compartment was equipped with an electrifiable grid floor connected to a shock generator that delivered an electrical stimulus of 1.3 mA. A video camera was mounted above the chambers to support ethological analyses (Fig 2a, b). The majority of components were fabricated from acrylic or plastic to reduce the overall cost. Acrylic was selected for the white chamber due to its optical clarity, ease of laser cutting and low material cost. Meanwhile, it still provides adequate light transmission and enables animal monitoring. ABS plastic was used for the dark chamber, control cabinet, cabinet lid, camera mount bracket, power switch panel, light source housing and touchscreen frame given its lightweight nature, ease of machining and greater affordability than metal alternatives. The sliding door was made of Formex sheet material, facilitating smooth and frictionless movement within the guide rail grooves between two chambers. Fabrication primarily involved laser cutting, plastic cutting, drilling and screw-based assembly. These methods are straightforward and do not necessitate specialized machinery or high labor input.


       
Model development included data augmentation techniques to introduce variability in the experimental animal’s position inside the chamber. Horizontal flipping and scaling were applied to 50 raw images, resulting in a total of 200 images for model training (Fig 2c, d). During the early epochs of training, mAP was relatively low (approximately 0.1). As training progressed, localization loss (box_loss) and classification loss (cls_loss) steadily decreased, suggesting improved prediction reliability (Fig 2e). The model was considered stable and ready for deployment when loss values approached zero and mAP approached one.
 
Development of software
 
The resulting interface developed by Qt Designer, titled “Passive Avoidance Test,” was structured into three distinct sections (Fig 3). The input section, positioned at the top, included fields for mouse ID, status, drug, run time, voltage (VAC) and current (mA), allowing users to enter information about the subject and adjust the foot shock current. The operation section contained two real-time display panels. Tracking panel on the left displayed live video monitoring of animal movement, while heatmap panel on the right visualized the spatial location of the animal within the apparatus. The output section at the bottom comprised two components. Statistics panel displayed the date, start time and the time at which the animal entered the dark compartment. Meanwhile, the chart of position over time provided a graphical representation of the time spent by the animal in either the light or dark compartment throughout the trial. Upon completion of each trial, the generated interface file was exported to the main program and step-through latency data were automatically extracted and saved in an excel file.


       
This in-house PA apparatus combines affordable materials, ease of construction, software-assisted behavioral assessment and heatmap-based visualization. With an estimated cost of approximately 450 USD, it is around 50-fold less expensive than commercial systems. Automated recording may reduce human error and observer bias while improving procedural consistency and reproducibility. Moreover, the heatmap feature enables visualization of spatial occupancy patterns, providing additional information regarding behavioral responses during pre-shock, shock and post-shock periods. Overall, the developed apparatus offers a potentially cost-effective alternative for PA testing.
 
Memory test by PA
 
On the test trial, step-through latency recorded by automated software was significantly lower in scopolamine-treated group compared to saline controls (p = 0.0233), indicating that scopolamine induced memory impairment (Fig 4a). To validate the accuracy of automated recordings, this indicator of memory was simultaneously measured manually by a trained observer. The results obtained by manual recording were comparable with those generated by the software, supporting the reliability of automated recording system. Further assessment using Bland-Altman analysis showed that automated recording yielded step-through latencies approximately 3.375 seconds shorter than manual recording (Fig 4b). This negative bias may partly reflect observer response time as manual recording requires the observer to identify the exact moment of crossing and manually stop the timer, potentially introducing a delay in recording the latency.


       
A novel feature of the apparatus was the integration of a heatmap function to visually represent the cumulative spatial distribution of each animal across the two compartments throughout the trial. Warm colors (red/yellow) represented prolonged occupancy in a specific location. In the saline group, higher intensity regions were observed in the light chamber, indicating that controls remained there longer before stepping through the dark chamber. This corresponded to a longer step-through latency and better retrieval of the aversive memory (Fig 4c). By contrast, the dark chamber showed greater intensity in the scopolamine-treated mice, consistent with a shorter latency time and impaired memory (Fig 4d). These findings suggest that the heatmap function provides a complementary visual indicator of memory performance alongside latency measurements. 
 
Validation of scopolamine-induced amnesia and neurodegeneration
 
To confirm the successful induction of scopolamine-induced neurodegeneration associated with memory deficits, HandE staining was performed to examine histopathological changes in CA1, CA3 and dentate gyrus (DG) regions of the hippocampus. Saline-treated mice exhibited well-organized hippocampal layers with intact cell nuclei and no substantial cell loss was observed in all regions examined. In contrast, scopolamine treatment caused a significant cell loss and sparse cellular distribution, suggestive of hippocampal neuronal damage (Fig 5a-h). Scopolamine induces oxidative stress, neuroinflammation and apoptosis, both of which contribute to cellular death and neurodegeneration (Cheon et al., 2021; Gul et al., 2023; Kim et al., 2026; Mohammad et al., 2026) and consequently impairs memory that resembles aspects of cognition dysfunction in Alzheimer’s disease and other neurodegenerative disorders. Previous investigations used relatively prolonged scopolamine regimens to induce such comparable neuronal loss such as 3 mg/kg daily for 15 days (Huang et al., 2022; Park et al., 2024), 4 mg/kg for 8 days (Helmi et al., 2021) and 2 mg/kg for 14 days (Hou et al., 2014). In contrast, our protocol with a lower dose and shorter duration (1 mg/kg daily for 6 days) was sufficient to produce remarkable cells loss in all hippocampal regions. These findings confirm the successful establishment of neurodegeneration associated with memory impairment.


               
To further validate memory impairment induced by scopolamine, BDNF expression was quantified in the hippocampi of saline- and scopolamine-treated mice. The number of BDNF-positive cells co-labelled with DAPI was significantly reduced in the CA1 (p = 0.0134) and CA3 (p = 0.0022) regions of scopolamine mice compared with the saline group, suggesting that impaired neurotrophic signalling contributed to the observed memory deficits (Fig 5i-p). Neurotrophic factors also play a crucial role in maintaining neuronal survival and synaptic plasticity. Brain-derived neurotrophic factor (BDNF) has been demonstrated to involve long-term potentiation and memory consolidation (Miranda et al., 2019). Previous studies reported that injection with scopolamine reduces hippocampal BDNF expression, which compromises neuroplasticity and cognitive performance (Kim et al., 2023). In line with these investigations, we observed a significant reduction in BDNF-positive cells in the hippocampus following 6 days of scopolamine (1 mg/kg, i.p.) treatment. Altogether, with the observed behavioral deficits, the histopathological alterations and decreased BDNF expression further reinforced the validity of the constructed PA apparatus as a reliable tool for evaluating memory deficits. 

This study designed and constructed PA apparatus at low cost to support research conducted in developing countries, with total expenditure for materials and fabrication amounting to approximately 450 USD. This apparatus is intended for assessing memory performance, which is fundamental to the majority of basic research on memory-enhancing therapies. The neuropathological alterations and changes in BDNF expression in the hippocampus of scopolamine-treated mice were consistent with behavioural findings, further supporting the validity of the PA apparatus as a reliable tool for evaluating memory impairment.
 
Funding
 
This research is funded by Ton Duc Thang University under grant number FOSTECT.2025.09.
 
Disclaimers
 
The views and conclusions expressed in this article are solely those of the authors and do not necessarily represent the views of their affiliated institutions. The authors are responsible for the accuracy and completeness of the information provided, but do not accept any liability for any direct or indirect losses resulting from the use of this content.
 
Informed consent
 
All animal procedures for experiments were approved by the Animal Ethics Committee of Ton Duc Thang University (Approval No. 17/HDTVDD).

The authors declare that there are no conflicts of interest regarding the publication of this article. No funding or sponsorship influenced the design of the study, data collection, analysis, decision to publish, or preparation of the manuscript.