Sage Journals: Discover world-class research

Abstract

Objective

Mild cognitive impairment (MCI) due to Alzheimer's disease (AD), as an early stage of AD, is an important point for early warning of AD. Neuropathological studies have shown that AD pathology in pre-dementia patients involves the hippocampus and caudate nucleus, which are responsible for controlling cognitive mechanisms such as the spatial executive process (SEP). The aim of this study is to design a new method for early warning of MCI due to AD by dynamically evaluating SEP.

Methods

We designed fingertip interaction handwriting digital evaluation paradigms and analyzed the dynamic trajectory of fingertip interaction and image data during “clock drawing” and “repetitive writing” tasks. Extracted fingertip interaction digital biomarkers were used to assess participants’ SEP disorders, ultimately enabling intelligent diagnosis of MCI due to AD. A cross-sectional study demonstrated the predictive performance of this new method.

Results

We enrolled 30 normal cognitive (NC) elderly and 30 MCI due to AD patients, and clinical research results showed that there may be neurobehavioral differences between the two groups in digital biomarkers captured during SEP. The early warning performance for MCI due to AD of this new method (areas under the curve (AUC) = 0.880) is better than that of the Minimum Mental State Examination (MMSE) neuropsychological scale (AUC = 0.856) assessed by physicians.

Conclusion

Patients with MCI due to AD may have SEP disorders, and this new method based on dynamic evaluation of SEP will provide a novel human–computer interaction and intelligent early warning method for home and community screening of MCI due to AD.

Keywords

Alzheimer's disease mild cognitive impairment spatial executive process alerting in the community fingertip human–computer interaction intelligent analysis algorithm

Introduction

Alzheimer's disease (AD) is one of the most common neurodegenerative diseases.^1,2 A large number of patients, increasing prevalence with an aging population, high social costs, no effective cure, and other factors have made AD a major medical and social concern worldwide.³ Mild cognitive impairment (MCI) due to AD is defined as the symptomatic predementia phase of AD.⁴ Some studies have found that appropriate intervention at the MCI due to AD stage has the potential to halt or even reverse AD progression.⁵ Therefore, MCI due to AD is an important point for early warning of AD.^6,7 However, the onset of AD is insidious and the latent phase before the onset of clinical syndrome can last for decades. Accurate diagnosis and large-scale screening of MCI due to AD, as the symptomatic predementia phase of AD, is also a recognized challenge in the field of medical health. The current clinical screening methods with high diagnostic efficacy, such as cerebrospinal fluid screening (CSF), blood biomarker testing, positron emission computed tomography (PET), and magnetic resonance imaging (MRI), have the bottlenecks of being expensive and difficult to promote large-scale early screening in the community.^8–11 In addition, neuropsychological scales such as the Minimum Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA), which are widely used in AD screening, have bottlenecks of relying on doctors’ on-the-spot evaluation, intense subjectivity, time-consuming, and insufficient early warning efficacy for early-stage patients with a highly educated background.^12–14 Therefore, the above bottlenecks in clinical examination methods such as imaging, biology, and neuropsychological scales have led to a low diagnosis and a high underdiagnosis rate of MCI due to AD, resulting in many patients at this stage losing the best time to intervene. They also limit large-scale early screening for MCI due to AD at home and in the community, especially in developing countries, where medical resources are significantly inadequate.¹⁵ Therefore, there is an urgent need to develop new intelligent early warning methods suitable for large-scale early detection of MCI due to AD at home and in the community.

Current neuropathological studies suggest that the hippocampus is the first site to be eroded by pathological changes in AD.^16–18 The hippocampus, as the most important subcortical brain region involved in learning memory functions, plays an important role in spatial and episodic memory. Spatial and episodic memory impairments are also the most prominent manifestations of cognitive impairment in the early stages of AD. In addition, some studies have shown that the caudate nucleus, which plays a key role in supporting the planning and execution of behaviors needed to achieve goal-directed tasks, will also be involved in AD pathology.^19,20 Executive function is defined as a cognitive mechanism, and impairments in executive processes resulting from damage to the caudate nucleus are also increasingly considered to have potential as a marker for early screening for AD.^20,21 In addition to its memory function, the hippocampus also plays a guiding role. The ability to store and process spatial information involved in the hippocampus is crucial for humans to complete goal-oriented spatial execution tasks. And the core of many tests used to evaluate memory and learning ability is the evaluation of the spatial executive process (SEP).^21–23

Handwriting is a complex act of spatial execution in humans that requires a complex mixture of cognitive, kinesthetic, and perceptual movements.²⁴ The characteristics of writing behavior during handwriting task execution can respond not only to participants’ executive functions but also to their higher-order neurocognitive decision making, information processing speed, memory, and other cognitive domain functions.²⁵ Tests related to “handwriting behavior pattern” have been shown to be sensitive to the cognitive abilities of patients with MCI due to AD.^26,27 Several studies have also demonstrated the correlation between “handwriting behavior pattern” and hippocampal and caudate pathological changes through medical imaging.^28–30 Moreover, because of its low cost, high acceptance, and short completion time,^25,31 the “handwriting behavior pattern” has been widely used in rapid screening for major neurological disorders such as AD and Parkinson's disease (PD).^24,32 Several studies have demonstrated the feasibility and effectiveness of handwriting tasks for cognitive detection based on human–computer interaction (HCI) early warning technology and digital devices.^26,27,33,34 Physiologically and behaviorally relevant “digital biomarkers” captured by digital and intelligent devices are objective and quantifiable, and they can compensate for the shortcomings of traditional diagnostic methods that are clinician-dependent, subjective, and difficult to quantify diagnostic results.^35–37 In addition, digital biomarkers have higher diagnostic accuracy and a lower rate of missed screening in the early diagnosis of neurological diseases such as AD, which are difficult to be early warned due to their insidious onset, and therefore have a good potential for clinical application.^35–41

However, most of the current commonly used AD handwriting assessment methods require the use of digital writing tools (such as digital pens or electromagnetic pens) to write on the touch screen or digital board.^42,43 These digital writing tool-based assessments increase the cost of screening equipment and have some unavoidable drawbacks. For example, handwriting evaluation results are influenced by individual differences in pencil grip and individual familiarity with electronic devices.^43,44 What's more, many studies followed the traditional “brain partitioning” theory to rigidly divide individuals’ integrated information processing abilities into different modules in the cognitive process. Moreover, they did not reflect the entire cognitive assessment process in dynamic time series.^45–47

Recently, Peter Stern presented a new idea in Science that brain neurons do not function in isolation but rather in connection and communication between different regions. The emergence of behavior and cognition also stems from the interaction between cortical areas.⁴⁸ The above views are in line with Villemagne et al.,⁴⁹ who argued that AD pathology affects a wide range of brain functions.

In our previous research work, we found a comprehensive capability indicator that can dynamically measure the entire spatial task execution process: the spatial execution process.⁵⁰ It is related to both hippocampus-dependent memory and orientation abilities and executive processes deficit (including full-process optimal decision-making and fingertip executive dynamic abilities) controlled by the caudate nucleus.^{20,21,50–52} Based on this, the hypothesis that “there are neurobehavioral differences between normal cognitive (NC) elderly and MCI due to AD patients in digital biomarkers captured during the spatial execution” was proposed. And we designed a smart 2-min mobile alerting method for MCI due to AD in the community. This 2-min alerting method used a mobile tablet device to dynamically and intelligently assess participants’ SEP under a new spatial interaction task we designed, with the advantages of fast screening and low economic cost. However, we found that the paradigm in this method (control of the target sphere around obstacles by fingertip interaction to eliminate target squares) had poorer warning efficacy for low-education MCI due to the AD population with less than 9 years of educational experience. This may be due to the lower acceptance and operational proficiency of smart devices in the less educated elderly population.

Therefore, we have improved on our previous work and developed a new method for screening MCI due to AD patients with higher accuracy, age-friendly, and more generalizability. We continued to use the fingertip interaction format and designed a new early warning method for MCI due to AD based on dynamic evaluation of the SEP by incorporating two handwriting tasks, “clock drawing” and “repetitive writing.” These are more commonly used in neuropsychological tests and usually used in the clinical diagnosis of MCI due to AD. By intelligently capturing the task performance of MCI due to AD in fingertip HCI tasks, the new early warning method achieves objective quantitative analysis of the entire data, thus achieving effective intelligent early warning for MCI due to AD in the home, community, and other nonphysician environments.

Methods

HCI software design

Fingertip interaction handwriting digital evaluation paradigm design. Since “clock drawing” and “repetitive writing” are the two most common measures used to screen for MCI due to AD patients, we have designed the “fingertip interaction clock drawing paradigm” and “fingertip interaction repetitive writing paradigm” for MCI due to AD screening. Based on these two paradigms of fingertip interaction, we will collect spatial, temporal, and kinematic data on handwriting behavior patterns that can reflect MCI due to AD patients’ memory capacity, orientation ability, optimal decision-making, and fingertip executive dynamic abilities.^16–23 We then used these data to assess participants’ SEP impairment during finger-writing tasks, enabling the assessment of cognitive function in MCI due to AD patients. The paradigms we have designed will replace the traditional “pen-holding writing” on paper media with “fingertip touch screen sensing.” And finger writing directly on the trackpad is in line with the concept of suitable aging and is more suitable for the intelligent assessment of the elderly (see Figure 1).

Figure 1.

(a) Fingertip interaction capacitive touch screen, (b) fingertip interaction clock drawing paradigm, and (c) fingertip interaction repetitive writing paradigm.

Hardware and software architecture and data storage. We collect data through the capacitive touch screen of the fingertip interaction, the relevant parameters of which are shown in Figure 1(a). In designing the fingertip interaction software system, we used Visual Studio Code, Electron, and Vue front-end framework to build the desktop client. And the backend system was built based on the Fast API backend framework and scientific computing libraries such as NumPy and Pandas. We built a MySQL system database to store and manage the writing process data and the resulting image feature data (including finger position coordinates and corresponding time) of the finger interaction clock drawing paradigm and the repetitive finger interaction writing paradigm operated by the participants. We can draw paradigms’ “process-result” dynamic spatial trajectory maps based on the Echart mapping library (see Figure 2).

Figure 2.

(a) Process-result dynamic trajectory multi-dimensional rotation visualization of repetitive writing, the black is the 2D trajectory, the grey is the 3D timing trajectory, the X and Y axes are the screen coordinates of the participant's handwriting, and the Z axis is the time, (b) Process-result dynamic trajectory multi-dimensional rotation visualization of clock drawing.

Fingertip interaction digital biomarkers extraction

With the use of digital biomarker extraction for fingertip interaction, we obtained data on drawing time, speed, and pause time during the two paradigms of finger interaction clock drawing evaluation and finger interaction repetitive writing evaluation via the finger interaction capacitive touch screen. In conjunction with the previous study, we set up three dimensions of digital biomarkers to provide intelligent warning to MCI due to AD patients from the perspective of spatial execution processes: (a) the digital biomarker dimension of execution speed signature and (b) the digital biomarker dimension of execution pause signature for evaluating optimal decision planning and fingertip execution dynamics dependent on the caudate nucleus^19–21 and (c) the digital biomarker dimension of execution image signature for evaluating hippocampus-dependent memory capacity and orientation ability.^16–18,22 Based on the above dimensions of digital biomarkers, we extracted fourteen digital biomarkers from two models of numerical fingerprint interaction handwriting evaluation, respectively.

Table 1 shows the digital biomarker dimension of execution speed signature, including four digital biomarkers: mean drawing speed ( $M_{speed}$ ), maximum drawing speed ( $S_{Max}$ ), minimum drawing speed ( $S_{Min}$ ), and drawing speed variability ( $C V_{speed}$ ).

Table 1.

The digital biomarker dimension of execution speed signature.

Digital biomarker names	Description
Mean drawing speed ( $M_{speed}$ ), px/s	Average drawing speed
Maximum drawing speed ( $S_{Max}$ ), px/s	Maximum value of drawing speed
Minimum drawing speed ( $S_{Min}$ ), px/s	Minimum value of drawing speed
Drawing speed variability ( $C V_{speed}$ )	Coefficient of Variation of Drawing Speed

Table 2 shows the digital biomarker dimension of execution pause signature, including nine digital biomarkers: total task time (T), start writing time ( $T_{start}$ ), total drawing time ( $T_{drawing}$ ), total pause time ( $T_{pause}$ ), mean pause time ( $M_{pause time}$ ), maximum pause time ( $T_{pause Max}$ ), variability of pause time ( $C V_{pause time}$ ), drawing pause rate ( $R_{pause}$ ), and fingertip retention rate ( $R_{retention}$ ).

Table 2.

The digital biomarker dimension of execution pause signature.

Digital biomarker names	Description
Total task time (T), s	Total time to complete a digital evaluation paradigm
Start writing time ( $T_{start}$ ), s	Time from the start of the digital evaluation until the first touch point occurs
Total drawing time ( $T_{drawing}$ ), s	Time between the appearance of the first touch point and the end of the last touch stroke
Total pause time ( $T_{pause}$ ), s	The sum of the time the fingertips are off the touch screen after the first touch point occurs
Maximum pause time ( $T_{pause Max}$ ), s	The maximum amount of fingertip time off the screen between each adjacent stroke after the first touch point occurs
Mean pause time ( $M_{pause time}$ ), s	The average time between fingertips leaving the touch screen after the first touch point occurs
Variability of pause time ( $C V_{pause time}$ )	Coefficient of variation of fingertip off touch screen time
Drawing pause rate ( $R_{pause}$ )	The ratio of total pause time ( $T_{pause}$ ) to total drawing time ( $T_{drawing}$ )
Fingertip retention rate ( $R_{retention}$ )	The ratio of the total fingertip contact time to the total task time (T)

Table 3 shows the digital biomarker dimension of execution image signature, including two digital biomarkers: the clock drawing process and resulting image feature score ( $S c o r e_{clock}$ ) and the repetitive writing process and resulting image feature score ( $S c o r e_{writing}$ ).

Table 3.

The digital biomarker dimension of execution image signature.

Digital biomarker names	Description
Clock drawing process and resulting image feature score ( $S c o r e_{clock}$ )	The total value of the score given by the system to the clock image results, including the clock profile score, clock number score, and clock time score
Repetitive writing process and resulting image feature score ( $S c o r e_{writing}$ )	Combined rating value given by the system for the sequence of the Chinese character stroke writing process and the image results

Intelligent analysis algorithms

We designed an intelligent analysis algorithm for fingertip interactive digital biomarkers analysis. It is suitable for analyzing the writing process and the resulting image features during the fingertip interaction clock drawing paradigm and the fingertip interaction repetitive writing paradigm. The following algorithms are general algorithms for digital biomarker analysis in two paradigms. We set the total task time as T, the start writing time as $T_{start}$ , the maximum number of strokes as n, the set of all positive integers as N*, the stroke length and stroke time of the ith stroke as $S_{i}$ and $T_{i}$ respectively, the pause time between the ith stroke and the i + 1th stroke as $T_{(i, i + 1)}$ , and the total pause time as $T_{pause}$ . The calculation methods are as follows:

T_{pause} = \sum_{i = 1}^{n - 1} T_{(i, i + 1)} (1 \leq i \leq n, i \in N^{*}) .

(1)

The mean pause time of clock is

M_{pause time}

, which is calculated as follows:

M_{pause time} = \frac{1}{n - 1} \sum_{i = 1}^{n - 1} T_{(i, i + 1)} (1 \leq i \leq n, i \in N^{*}) .

(2)

The variability of pause time of clock is

C V_{pause time}

, and its calculation method is as follows:

σ_{pause} = \sqrt{\frac{\sum_{i = 1}^{n - 1} {(T_{(i, i + 1)} - M_{pause time})}^{2}}{n - 1}} (1 \leq i \leq n, i \in N^{*}) .

(3)

C V_{pause time} = σ_{pause} / M_{pause time} .

(4)

σ_{pause}

in equation (3) is the standard deviation of stay time in the air.

The drawing pause rate of clock is $R_{pause}$ , which is calculated as follows:

R_{pause} = \frac{T_{pause}}{T - T_{start}} (1 \leq i \leq n, i \in N^{*}) .

(5)

The fingertip retention rate of clock is

R_{retention}

, which is calculated as follows:

R_{retention} = \frac{\sum_{i = 1}^{n} T_{i}}{T} (1 \leq i \leq n, i \in N^{*}) .

(6)

The mean drawing speed of clock is

M_{speed}

, which is calculated as follows:

M_{speed} = \frac{1}{n} \sum_{i = 1}^{n} \frac{S_{i}}{T_{i}} (1 \leq i \leq n, i \in N^{*}) .

(7)

The drawing speed variability of clock is

C V_{speed}

, which is calculated as follows:

σ_{speed} = \sqrt{\frac{\sum_{i = 1}^{n} \frac{S_{i}}{T_{i}} - M_{speed}^{2}}{n}} (1 \leq i \leq n, i \in N^{*}) .

(8)

C V_{speed} = σ_{speed} / M_{speed} .

(9)

σ_{speed}

in equation (8) is the standard deviation of the stroke speed.

Second, we evaluated the resulting image features of the fingertip interaction clock drawing evaluation paradigm through the clock outline score, clock digit score, and clock hand score. The specific analysis method is as follows:

Based on the Open Source Computer Vision Library (OpenCV2), we realize the automatic preprocessing of the original clock image through the “open operation” in morphology. We “eroded” the clock images drawn by the participants to remove “burrs” and then “expanded” the clock images to prevent subsequent influence on the judgment of the closure of the outer contour of the clock. The formula for the “opening operation” is as follows:

f \circ a = (f ⊖ a) \oplus a .

(10)

In the formula, f is the clock image, a is the structure element,

f ⊖ a

is the “corrosion” of the image f with a, and

(f ⊖ a) \oplus a

is the “expansion” of the image f after the “corrosion” with a.

We recorded binary clock profile scores based on contour edge detection techniques. If the clock profile is closed, the score is 1 or 0 otherwise. We used optical character recognition (OCR) and recorded the binary clock number score. If the clock numbers are complete and distributed clockwise, the score is 1 or 0 otherwise, as shown in Figure 3(a). We used a clock identification architecture from the space converter network (STN)⁵³ and recorded clock time scores in binary. If the clock time is recognized as 11:10, the score is 1 or 0 otherwise, as shown in Figure 3(b). The clock time reading formula is as follows:

\hat{T} = Φ_{cls} (Picture s_{clock}) \in P^{720} .

(11)

Figure 3.

Digital biomarker analysis algorithms: (a) the schematic diagram of the recognition of clock numbers, (b) the schematic diagram of clock time reading, and (c) the schematic diagram of Chinese character recognition of “repetitive writing.”

In the formula, $\hat{T}$ is the time of the predicted clock; $Picture s_{clock}$ is the clock image drawn by the participant; $Φ_{cls}$ is the classification network. After we quantified the time, there will be 12 possible outcomes for hours and 60 possible outcomes for minutes. Finally, there are 720 ways to combine hours and minutes in total, and $P^{720}$ is 720 classification results.

Third, we evaluated the resulting image features of the fingertip interaction repetitive writing paradigm. Based on the turtle library's handwriting restoration algorithm and OCR technology, we analyzed the stroke order of the participant and the resulting image during the repetitive writing process, as shown in Figure 3(c). If the participants’ stroke order and results are correct, the writing score will be 1 or 0 otherwise.

Experimental settings

In order to verify the clinical efficacy of the new early warning method for MCI due to AD based on dynamic evaluation of SEP, we conducted a cross-sectional study in the neurology department of a tertiary level A hospital in Zhejiang Province.

Participant recruitment

From June 2022 to September 2022, we enrolled 60 participants (27 males and 33 females) aged 55–80 in a hospital, including 30 NC elderly and 30 MCI due to AD patients. They were divided into the NC group and the MCI group. All recruited participants underwent clinical assessment and diagnosis by a neurologist (including laboratory tests, cognitive ability assessment, daily and general functional assessment, and psycho-behavioral assessment, and some patients underwent imaging tests such as MRI and PET). Individuals in the NC group were confirmed by clinical interviews not to show cognitive impairment or any signs of neurological or mental illness. The individual diagnosis of the MCI group met the clinical MCI diagnostic criteria developed by the National Institutes of Aging and Alzheimer's Disease Association in 2011,⁴ including confirmed complaints of memory loss, objective memory impairment, retention of general cognitive function, complete daily activities, and no dementia.

Exclusion criteria included a history of mental illness such as schizophrenia, severe anxiety, and depression; meeting diagnostic criteria for PD, frontotemporal dementia, Lewy body dementia, or Huntington's disease; dementia caused by other causes, such as cerebrovascular disease and central nervous system trauma; aphasia, disturbance of consciousness, and other diseases affecting cognitive evaluation; severe arrhythmia; myocardial infarction occurring within 6 months; severe lung/liver/kidney dysfunction and severe gastrointestinal diseases; severe anemia; tumors; and a history of epilepsy or taking antiepileptic. In addition, participants in this study were right-handed.

The study design was approved by the Human Research Ethics Committee of Zhejiang Chinese Medical University (Approval Number: 20210806-1). All operations of this study were conducted in accordance with the relevant guidelines and regulations. After receiving a detailed description of the study, all signed informed consent. And all patient privacy data in this paper have been desensitized to protect patients’ privacy rights.

Experimental process

All participants were asked to sit at a table (see Figure 4(a) and (b)) and complete the digital evaluation of fingertip interaction handwriting with the index finger of the habitual hand in the blank area displayed on the touch screen according to the prompt of the fingertip HCI software system interface. In the fingertip interaction clock drawing paradigm, participants were prompted to draw an 11: 10 clock with their habitual hand index finger, including the contour, number, and pointer of the clock. Examples of drawing results are shown in Figure 4(c) and (d). In the fingertip interaction repetitive writing paradigm, participants were prompted to write two Chinese characters: “井” and “米” with their habitual hand and index finger. Each Chinese character must be written three times, as shown in Figure 4(e) and (f).

Figure 4.

Evaluation scenarios and example results: (a) fingertip interaction clock drawing evaluation scene, (b) fingertip interaction repetitive writing evaluation scene, (c) examples of clock drawing result of NC, (d) examples of clock drawing result of MCI, (e) examples of repetitive writing result of NC, and (f) examples of repetitive writing result of MCI. MCI: mild cognitive impairment; NC: normal cognitive.

Fingertip interaction digital biomarker screening based on semistructured interviews

To enhance the scientific and clinical applicability of our new early warning method, we further used semistructured interviews to select digital biomarkers that are more credible, more consistent with commonly used clinical MCI due to AD screening, and easier to quantify digitally.

First, we developed interview profiles with different focuses for the two groups of interviewees based on previous literature reviews and the summary of the previous study.^{27,44,54–59} Interview Group A included 20 neuroscience experts with extensive experience in the assessment and diagnosis of cognitive impairment. Interview Group B included 20 information experts in the field of digital medicine. Then, we provided the interview outlines to interviewees 3 days before the formal interview, so that they could prepare in advance and thus improve the effectiveness and quality of the interview. In addition, we informed the interviewer of the interview process and explained the use of the information obtained before the interview was formally conducted, so that interviewees would have no worries. Tables 4 and 5 show the details of the interview outlines.

Table 4.

Interview outlines for clinical medicine specialists.

Group A (clinical medicine specialists)
Interview outlines	Source
1. Which of the two interaction modes, “fingertip interaction” or “digital pen interaction,” do you think is more suitable for MCI due to AD early warning in the elderly population in actual clinical, home and large scale community settings?	Ishikawa et al.⁵⁶
2. Do you think that the new MCI due to AD early warning method proposed in this study is medically scientific? Is it applicable to MCI due to AD early warning screening in the elderly population?	Kobayashi et al.⁵⁸
3. Combined with the main clinical features of pre-AD, do you think that fingertip interaction digital biomarker profiles can reflect the impairment of executive processes during interaction tasks in the MCI due to AD population?	Garre-Olmo et al.⁵⁹
4. From a clinical diagnostic perspective, which of the fingertip interaction digital biomarkers extracted from this study do you think are more specific and sensitive in MCI due to AD early warning screening?	Rentz et al.²⁷
5. Which digital biomarkers do you think are more appropriate for disease screening in clinical, home, and large-scale community scenarios?	Müeller et al.⁵⁴

AD: Alzheimer's disease; MCI: mild cognitive impairment; NC: normal cognitive.

Table 5.

Interview outline for computer science experts.

Group B (computer science experts)
Interview outlines	Source
1. Do you think that the new MCI due to AD early warning method proposed in this study can achieve dynamic capture of the whole process of multidimensional fingertip interaction digital biomarkers?	Williams et al.⁵⁷
2. Among the fingertip interaction digital biomarkers extracted in this study, which markers do you think are more suitable to capture using digital smart devices?	Ghaderyan et al.⁵⁵
3. Among the fingertip interaction digital biomarkers extracted in this study, which markers do you think are more computable, easy to analyze, and quantitatively accurate?	Rentz et al.²⁷
4. Among the digital biomarkers of fingertip interaction extracted in this study, which markers do you think are more responsive to the participants’ performance of the interaction task in its entirety in a time series?	Sliwinski et al.⁴⁴
5. Among the fingertip interaction digital biomarkers extracted in this study, which ones do you think are more objective in screening subjects for cognitive function?	Ghaderyan et al.⁵⁵
6. Which digital biomarkers do you think can better reflect the universality and ubiquity of this new MCI due to AD early warning method?	Müeller et al.⁵⁴

AD: Alzheimer's disease; MCI: mild cognitive impairment; NC: normal cognitive.

Data analysis

All statistical analyses were performed using the general-purpose data analysis software SPSS 25.0. The homogeneity of variances was checked using Levene's test. Measures conforming to a normal distribution were expressed as `x ± s using independent t-test samples. While skewed distributions were expressed as M (interquartile range (IQR)) using the Mann–Whitney U test. The χ² test was used for the count data. Receiver operating characteristic (ROC) curves were generated and the areas under the curve (AUC) were compared for the performance of single fingertip interaction digital biomarkers in detection of MCI due to AD. For the joint analysis of multiple digital biomarkers, a binary logistic regression model was used for multivariate analysis. p values < 0.05 were considered statistically significant.

Results

We included a total of 60 participants aged 55–80, of whom 30 were NC elderly and 30 were MCI due to AD patients. They were divided into the NC group and the MCI group. First, we performed a differential analysis of baseline information between the two groups, and there were no significant differences in age, sex, and years of education between the two groups (p > 0.05), as shown in Figure 5(a) and (b). Detailed results are presented in Table 6. We organized senior physicians to score the participants on the MMSE neuropsychological scale (one of the commonly used clinical diagnostic tools for AD) at the time participants enrolled (MMSE score ≥ 27 considered NC function). We analyzed the differences between the two groups in the diagnostic results of the MMSE neuropsychological scale. Results showed that the MMSE score of the NC group was higher than that of the MCI group (p < 0.05), and there was a significant difference between the two groups, as shown in Figure 5(c).

Figure 5.

Data distribution of (a) age of the NC group and the MCI group, (b) years of education of the NC group and the MCI group, and (c) MMSE score of the NC group and the MCI group. MCI: mild cognitive impairment; MMSE: minimum mental state examination; NC: normal cognitive.

Table 6.

Comparative results of demographic and clinical characteristics of the NC and MCI groups.

	NC group n = 30	MCI group n = 30	p value
Age, years	66.07 ± 7.93	69.23 ± 5.88	0.076
Female (n, %)	14 (46.67)	13 (43.33)	0.795
Years of education	12.03 ± 3.61	10.40 ± 4.06	0.111
MMSE score**	29.00 (2.00)	26.00 (4.25)	< 0.001

MCI: mild cognitive impairment patients; MMSE: minimum mental state examination; NC: normal cognitive.

∗∗ Two-sample rank sum test with significance of p < 0.01.

Second, we compared all digital biomarkers involved in completing the fingertip interaction clock drawing paradigm and the fingertip interaction repetitive writing paradigm between two groups. We found that a total of 19 digital biomarkers were different between the NC and MCI groups (p < 0.05). In the fingertip interaction clock drawing paradigm, T clock, $T_{drawing} clock$ , $T_{pause} clock$ , $T_{pause Max} clock$ , $M_{pause time} clock$ , $C V_{pause time} clock$ , and $R_{pause} clock$ of the MCI group were higher than that of NC group, but $R_{retention} clock$ and $S c o r e_{clock}$ were lower than that of the NC group; in the fingertip interaction repetitive writing paradigm, $C V_{speed} writing$ , T writing, $T_{start} writing$ , $T_{drawing} writing$ , $T_{pause} writing$ , $T_{pause Max} writing$ , $M_{pause time} writing$ , and $C V_{pause time} writing$ of the MCI group were higher than that of the NC group, but $S_{Min} writing$ and $R_{retention} writing$ were lower than that of the NC group. The complete difference test results between the two groups are shown in Table 7. Figure 6 illustrates 11 digital biomarkers with significant intergroup variability (p < 0.01).

Figure 6.

Digital biomarkers with intergroup variability p < 0.01 between the NC and MCI groups. ** indicates significant difference between the two groups, at p < 0.01. MCI: mild cognitive impairment; NC: normal cognitive; $M_{speed}$ : mean drawing speed; $S_{Max}$ : maximum drawing speed; $S_{Min}$ : minimum drawing speed; $C V_{speed}$ : drawing speed variability; T: total task time; $T_{start}$ : start writing time; $T_{drawing}$ : total drawing time; $T_{pause}$ : total pause time; $T_{pause Max}$ : maximum pause time; $M_{pause time}$ : mean pause time; $C V_{pause time}$ : variability of pause time; $R_{pause}$ : drawing pause rate; $R_{retention}$ : fingertip retention rate; $S c o r e_{clock}$ : clock drawing process and resulting image feature score; $S c o r e_{writing}$ : repetitive writing process and resulting image feature score.

Table 7.

Results of the test of variability between the NC and MCI groups for all fingertip interaction digital biomarkers.

	NC group n = 30	MCI group n = 30	p value
Fingertip interaction clock drawing paradigm
The digital biomarker dimension of execution speed signature
$M_{\binom{speed}{}}$ clock, px/s	260.04 (127.07)	211.46 (141.45)	0.143
$S_{Max}$ clock, px/s	790.64 (385.83)	649.55 (266.09)	0.188
$S_{Min}$ clock, px/s	0.00 (80.07)	0.00 (0.00)	0.110
$C V_{speed}$ clock	105.62 (83.89)	97.87 (78.19)	0.813
The digital biomarker dimension of execution pause signature
T clock, s^∗∗	42.29 (23.45)	65.33 (33.56)	0.001
$T_{start}$ clock, s	3.66 (2.21)	3.86 (4.32)	0.348
$T_{drawing}$ clock, s^∗∗	38.23 (25.63)	62.55 (33.28)	0.001
$T_{pause}$ clock, s^∗∗	34.92 (22.44)	60.83 (32.38)	0.001
$T_{pause Max}$ clock, s^∗∗	3.86 (4.18)	9.78 (9.73)	0.001
$M_{pause time}$ clock, s^∗∗	0.78 (0.66)	1.27 (1.33)	0.003
$C V_{pause time}$ clock^∗∗	1.04 (1.47)	3.47 (4.54)	0.001
$R_{pause}$ clock^∗	0.92 (0.07)	0.95 (0.04)	0.011
$R_{retention}$ clock^∗∗	0.39 (0.15)	0.28 (0.17)	0.001
The digital biomarker dimension of execution image signature
$S c o r e_{clock}$ ^∗∗	2.00 (1.00)	1.00 (1.00)	0.001
Fingertip interaction repetitive writing paradigm
The digital biomarker dimension of execution speed signature
$M_{speed}$ writing, px/s	692.77 (332.86)	595.85 (276.15)	0.089
$S_{Max}$ writing, px/s	1219.94 (724.86)	1050.69 (435.90)	0.564
$S_{Min}$ writing, px/s^∗	267.45 (197.47)	142.81 (235.96)	0.018
$C V_{speed}$ writing^∗	68.00 (41.50)	92.26 (59.11)	0.040
The digital biomarker dimension of execution pause signature
T writing, s^∗∗	16.45 (11.84)	23.01 (12.96)	0.001
$T_{s t a r t}$ writing, s^∗	2.44 (1.71)	3.54 (4.00)	0.032
$T_{drawing}$ writing, s^∗	13.59 (8.12)	18.04 (11.79)	0.011
$T_{pause}$ writing, s^∗∗	13.12 (8.10)	17.48 (11.49)	0.008
$T_{pause Max}$ writing, s^∗	1.09 (1.73)	1.94 (2.94)	0.015
$M_{pause time}$ writing, s^∗	0.25 (0.17)	0.33 (0.22)	0.011
$C V_{pause time}$ writing^∗	0.19 (0.54)	0.37 (1.18)	0.029
$R_{pause}$ writing	0.97 (0.02)	0.97 (0.01)	0.871
$R_{retention}$ writing^∗∗	0.44 (0.08)	0.36 (0.10)	0.001
The digital biomarker dimension of execution image signature
$S c o r e_{writing}$	1.00 (1.00)	1.00 (1.00)	0.792

MCI: mild cognitive impairment; NC: normal cognitive; $M_{speed}$ : mean drawing speed; $S_{Max}$ : maximum drawing speed; $S_{Min}$ : minimum drawing speed; $C V_{speed}$ : drawing speed variability; T: total task time; $T_{start}$ : start writing time; $T_{drawing}$ : total drawing time; $T_{pause}$ : total pause time; $T_{pause Max}$ : maximum pause time; $M_{pause time}$ : mean pause time; $C V_{pause time}$ : variability of pause time; $R_{pause}$ : drawing pause rate; $R_{retention}$ : fingertip retention rate; $S c o r e_{clock}$ : clock drawing process and resulting image feature score; $S c o r e_{writing}$ : repetitive writing process and resulting image feature score.

∗ Two-sample rank sum test with significance of p < 0.05; ∗∗ p < 0.01.

Fingertip interaction digital biomarker screening results based on semistructured interviews

We conducted semistructured interviews with 20 clinical medicine experts and 20 computer science experts, and the results were as follows:

Conclusion 1. 75% of clinical medicine experts believe that the “fingertip interaction” model is more suitable for MCI due to AD early warning screening of the elderly population in clinical, home, and community settings; and all clinical medical and computer science experts from their professional fields affirmed the scientific, rational, and universal applicability of our new MCI due to AD early warning method;

Conclusion 2. In interview Group A, 60% of clinical medicine experts believed that the digital biomarkers extracted in this study were indicative of SEP impairment during interactive tasks in the MCI due to AD population; however, 70% of clinical medicine experts believe that $C V_{pause time} clock$ , $R_{retention} clock$ , $S c o r e_{clock}$ , $C V_{speed} writing$ , $S_{Min} writing$ , $C V_{pause time} writing$ , $R_{retention} writing$ , and $S c o r e_{writing}$ are more specific and sensitive, and 65% of clinical medicine experts believe that T clock, $T_{drawing} clock$ , $T_{pause} clock$ , $T_{pause Max} clock$ , T writing, $T_{drawing} writing$ , and $T_{pause} writing$ are more suitable for disease screening in clinical, home and community settings;

Conclusion 3. In interview Group B, 90% of computer science experts affirmed the ability of this new early warning method to dynamically capture the entire process of multidimensional fingertip interaction digital biomarkers; 80% of computer science experts believe that T clock, $T_{pause} clock$ , $T_{pause Max} clock$ , $T_{drawing} clock$ , $S_{Min} clock$ , T writing, $T_{pause} writing$ , $T_{pause Max} writing$ , $T_{drawing} clock$ , and $S_{Min} clock$ are more convenient to capture by digital intelligence devices, and they are more computable, easy to analyze, and quantitatively accurate; 75% of computer science experts believe that $C V_{speed} clock$ , $C V_{pause time} clock$ , $R_{retention} clock$ , $C V_{speed} writing$ , $C V_{pause time} writing$ , and $R_{retention} writing$ can better objectively reflect the dynamics of subjects during the whole process of interactive task execution; 65% of computer science experts believe that $T_{drawing} clock$ , $T_{pause} clock$ , $R_{retention} clock$ , $C V_{pause time} clock$ , $T_{drawing} writing$ , $T_{pause} writing$ , $R_{retention} writing$ , and $C V_{pause time} writing$ can better reflect the universality of this new early warning method.

Based on the results of the above semistructured interviews, we mapped the recommendation index radar of clinical medicine experts and computer science experts for 19 digital biomarkers with intergroup variability in the above two paradigms (see Figure 7).

Figure 7.

Fingertip interactive digital biomarkers selection results. $S_{Min}$ : minimum drawing speed; $C V_{speed}$ : drawing speed variability; T: total task time; $T_{start}$ : start writing time; $T_{drawing}$ : total drawing time; $T_{pause}$ : total pause time; $T_{pause Max}$ : maximum pause time; $M_{pause time}$ : mean pause time; $C V_{pause time}$ : variability of pause time; $R_{pause}$ : drawing pause rate; $R_{retention}$ : fingertip retention rate; $S c o r e_{clock}$ : clock drawing process and resulting image feature score; $S c o r e_{writing}$ : repetitive writing process and resulting image feature score.

Based on the above results, we downscaled the original digital biomarkers and screened six digital biomarkers. They may reflect MCI due to AD patients’ SEP barriers and are also commonly used for clinical screening and diagnosis of MCI due to AD. At the same time, these digital biomarkers have good performance in objective quantification and calculation. These digital biomarkers include $T_{pause} clock$ , $T_{pause Max} clock$ , $M_{pause time} clock$ , $C V_{pause time} clock$ , $R_{retention} clock$ , $C V_{pause time} writing$ , and $R_{retention} writing$ .

We then draw the combined ROC curve of MCI due to AD of these six preferred digital biomarkers and the AUC reached 0.880. Its early warning effectiveness exceeded that of the MMSE, a neuropsychological scale judged by clinicians (AUC = 0.856). The results demonstrate the clinical validity of this new early warning method for MCI due to AD based on dynamic evaluation of SEP (see Figure 8).

Figure 8.

Receiver operating characteristic curves and the areas under the curve for MCI alerted by MMSE and six fingertip interaction digital biomarkers. $C V_{pause time}$ : variability of pause time; $R_{retention}$ : fingertip retention rate; $S c o r e_{clock}$ : clock drawing process and resulting image feature score; $S_{Min}$ : minimum drawing speed; T: total task time.

Exploring the cowarning efficacy of multidimensional fingertip interaction digital biomarkers analysis results

In addition to the early warning effectiveness analysis of the above six digital biomarkers, we also made an exploratory analysis of other digital biomarkers that were well evaluated by experts, although they did not show significant differences between NC and MCI groups (p > 0.05). The reason why there is no significant difference between groups may be due to insufficient sample size. We analyzed the dimensions of all digital biomarkers with intergroup differences. In the fingertip interaction clock drawing paradigm, they are aggregated in execution pause signature and execution image signature; in fingertip interaction repetitive writing paradigm, they are aggregated in execution speed signature and execution pause signature (see Figure 9).

Figure 9.

Dimension distribution of biomarkers with intergroup differences. $S_{Min}$ : minimum drawing speed; $C V_{speed}$ : drawing speed variability; T: total task time; $T_{start}$ : start writing time; $T_{drawing}$ : total drawing time; $T_{pause}$ : total pause time; $T_{pause Max}$ : maximum pause time; $M_{pause time}$ : mean pause time; $C V_{pause time}$ : variability of pause time; $R_{pause}$ : drawing pause rate; $R_{retention}$ : fingertip retention rate; $S c o r e_{clock}$ : clock drawing process and resulting image feature score; $S c o r e_{writing}$ : repetitive writing process and resulting image feature score.

We included digital biomarkers that have not yet shown intergroup differences and explored the cowarning efficacy of multidimensional fingertip interaction digital biomarkers incorporating in execution pause, speed, and image signature digital biomarkers. Based on six previously screened digital biomarkers, we added an execution speed digital biomarker in fingertip interaction clock drawing paradigm ( $S_{Min} writing$ ) and an execution image digital biomarker in interaction repetitive writing paradigm ( $S c o r e_{writing}$ ). The final results showed that compared to the previous early warning performance (AUC = 0.880), the early warning performance of our new method was further improved after the inclusion of two exploratory markers (AUC = 0.902), as shown in Figure 10. Therefore, we speculate that when the sample size is large enough in the future, further fusion of the multidimensional data of the execution pause, speed, and image signature digital biomarkers will improve the early warning effectiveness of this new method for MCI due to AD.

Figure 10.

The combined receiver operating characteristic curve and the area under the curve after adding two exploratory markers. $C V_{pause time}$ : variability of pause time; $R_{retention}$ : fingertip retention rate; $S c o r e_{clock}$ : clock drawing process and resulting image feature score; $S_{Min}$ : minimum drawing speed; T: total task time; $S c o r e_{writing}$ : repetitive writing process and resulting image feature score.

Discussion

Using the new early warning method for MCI due to AD based on dynamic evaluation of SEP, we extracted six digital biomarkers indicating spatial executive dysfunction in MCI due to AD patients. And the new method has an early warning performance of 0.880 for MCI due to AD patients.

First, compared to our previous 2.5-Minute Human-computer Interactive Rapid Digital Early Warning Technology,³⁴ we used fingertip interaction mode instead of eye tracking devices and gamepads. The use of fingertip interaction can minimize individual differences in familiarity with digital writing tools and pencil grip and avoid “writing anomalies” when participants change their usual writing style due to the writing tool. It not only reduces the cost of screening equipment and improves the mobility of the screening system but also improves the warning effectiveness of the screening method (the early warning effectiveness of eye-tracking device-based 2.5-min rapid digital early warning technology for MCI due to AD patients being 0.824). In addition to this, compared to our previous 2-min mobile screening method also based on fingertip interaction,⁵⁰ our new method takes into account differences in learning ability and acceptance of smart devices and digital assessment paradigms across older individuals. We chose the handwriting task paradigms of “drawing clocks” and “repetitive writing,” which are more commonly used in clinical settings and are more familiar to older adults, to minimize variability in individual familiarity with the paradigms and to reflect the interactive task performance characteristics of the subjects more objectively. Through the clinical research, our new method (AUC = 0.880) was found to be more sensitive than our previous 2-min mobile fingertip interaction-based screening method (AUC = 0.830). The above results suggested that this fingertip interactive handwriting assessment method based on “drawing clock” and “repetitive writing” is more suitable for large-scale initial screening of MCI due to AD in developing countries with uneven education and lack of medical resources due to uneven economic development, as well as for cognitive function assessment in home and community settings without physicians.

Second, according to the theory of neuropsychology, we interpreted the early warning mechanism of the new early warning method of screening MCI due to AD based on the dynamic evaluation of SEP at the medical level. Because the level of writing difficulties in patients with MCI due to AD is closely related to the severity of the disease and the level of concomitant cognitive impairment, indicators related to the handwriting of writing results and the process of performing interactive tasks in patients with MCI due to AD may reveal deficits in their ability to perform the process. The handwritten evaluation process is essentially a dynamic evaluation process based on a time series. And participants’ multidimensional characteristics such as touch-sensitive pressure, writing speed and acceleration, writing pause duration, and hand movements while performing interactive tasks can be used as MCI due to AD discriminators. Therefore, the fingertip interaction clock drawing paradigm and the fingertip interaction repetitive writing paradigm we designed could show participants’ SEP impairment in spatial interactive tasks in dynamic time series, which may reflect cognitive dysfunction in MCI due to AD.

Third, we explored the model of community-based intelligent primary screening for MCI due to AD in non-physical conditions. This new community-based early warning method for MCI due to AD could be used to achieve intelligent assessment across regions in developing countries where physicians are scarce and medical resources are unevenly distributed, using technologies such as the Internet and the Internet of Things. And it could also be repeated multiple times at different times and locations in various everyday life settings, such as home and community.

Finally, we explored the limitations of the study. Our relatively small sample size may result in insufficient test efficacy, and the actual warning efficacy of some digital biomarkers may not be demonstrated. In addition, the subjects of this study only include Chinese, so the features of “repetitive writing” may only be valid for individuals using Chinese characters. Furthermore, due to study conditions, we conducted only a cross-sectional study, not a longitudinal study, and only a comparative study with MMSE scores in the clinical applicability, lacking a comparison with more diagnostic imaging data. In the future, we will conduct longitudinal studies and explore the “repetitive writing” characteristics of individuals using different characters. And we will continue to test the clinical applicability and incorporate other clinical diagnostic data and further optimize the early warning system and evaluation paradigm based on large sample size data.

Conclusions

We believe that MCI due to AD patients has a SEP impairment during the execution of spatial interaction tasks. Therefore, we have developed a novel and innovative MCI due to AD screening method that can objectively and dynamically evaluate the whole process of SEP disorders in a nonphysician environment. Two digital evaluation paradigms of fingertip interaction, the fingertip interaction clock drawing paradigm and the fingertip interaction repetitive writing paradigm, were also designed. The method can intelligently extract and analyze the multidimensional digital biomarkers of fingertip interaction during the execution of the above two digital evaluation paradigms of fingertip interaction and realize effective early warning of MCI due to AD population by evaluating subjects’ ability to dynamically perform the whole process. After clinical research validation, our novel early warning method based on dynamic evaluation of SEP has a good warning performance for MCI due to AD (AUC = 0.880). And the new early warning method for MCI due to AD based on dynamic evaluation of SEP has the advantages of simple operation, high degree of aging suitability, high degree of intelligence, and high early warning effectiveness and will provide a reliable tool for early mass screening and warning of AD in a home environment without a doctor in developing countries with large populations, high demand for disease screening, uneven economic development, and varying levels of education.

Footnotes

Acknowledgments

The authors would like to thank all participants for enrollment in this study.

Contributorship

KL and RX: conceptualization and methodology. JX and SL: software. XM, TC, AO, BW, and YW: validation and formal analysis. SZ, YW, SH, ZC, and JL: investigation. KL, TC, and RX: resources; XM, YW, ZC, and CW: data curation. KL, XM, JX, and CW: writing—original draft preparation; KL RX, and TC: writing—review and editing; XM and CW: visualization; RX: supervision; KL: project administration and funding acquisition. All authors reviewed and edited the manuscript and approved the final version of the manuscript.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Ethical approval

The experimental protocol was designed in accordance with the ethical guidelines of the Helsinki Declaration. Ethical permissions were granted by the Human Ethics Committee of the Zhejiang Chinese Medical University (Approval Number: 20210806-1). Written informed consents were obtained from patients or their guardians.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Province Key Research and Development Program of Zhejiang (grant numbers 2021C03116 and 2022C03064), and scientific research project of Zhejiang Chinese Medical University (grant number 2021JKZDZC04).

Guarantor

KL.

ORCID iDs

Kai Li

Xiaowen Ma

Siyu Zhou

References

Scheltens

De Strooper

Kivipelto

, et al. Alzheimer's disease. Lancet 2021; 397: 1577–1590.

Jia

Quan

, et al. Dementia in China: epidemiology, clinical management, and research advances. Lancet Neurol 2020; 19: 81–92.

Wimo

Seeher

Cataldi

, et al. The worldwide costs of dementia in 2019. Alzheimers Dement 2023; 19: 2865–2873.

Albert

DeKosky

Dickson

, et al. The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011; 7: 270–279.

Lindbergh

Dishman

Miller

. Functional disability in mild cognitive impairment: a systematic review and meta-analysis. Neuropsychol Rev 2016; 26: 129–159.

Petersen

Lopez

Armstrong

, et al. Practice guideline update summary: mild cognitive impairment: report of the guideline development, dissemination, and implementation subcommittee of the American academy of neurology. Neurology 2018; 90: 126–135.

Knopman

Amieva

Petersen

, et al. Alzheimer disease. Nat Rev Dis Primers 2021; 7: 33.

Lian

Liu

Zhang

, et al. Hierarchical fully convolutional network for joint atrophy localization and Alzheimer's disease diagnosis using structural MRI. IEEE Trans Pattern Anal Mach Intell 2020; 42: 880–893.

Carneiro

Faria

Coutinho

, et al. Evaluation of 10-minute post-injection 11C-PiB PET and its correlation with 18F-FDG PET in older adults who are cognitively healthy, mildly impaired, or with probable Alzheimer's disease. Braz J Psychiatry 2022; 44: 495–506.

10.

Karikari

Pascoal

Ashton

, et al. Blood phosphorylated tau 181 as a biomarker for Alzheimer's disease: a diagnostic performance and prediction modelling study using data from four prospective cohorts. Lancet Neurol 2020; 19: 422–433.

11.

Blennow

Stomrud

Zetterberg

, et al. Second-generation Elecsys cerebrospinal fluid immunoassays aid diagnosis of early Alzheimer's disease. Clin Chem Lab Med 2023; 61: 234–244.

12.

Karikari

Benedet

Ashton

, et al. Diagnostic performance and prediction of clinical progression of plasma phospho-tau181 in the Alzheimer's disease neuroimaging initiative. Mol Psychiatry 2021; 26: 429–442.

13.

Freitas

Simões

Alves

, et al. Montreal Cognitive Assessment (MoCA): validation study for vascular dementia. J Int Neuropsychol Soc 2012; 18: 1031–1040.

14.

Pinto

TCC

Machado

Bulgacov

, et al.

Is the Montreal Cognitive Assessment (MoCA) screening superior to the Mini-Mental State Examination (MMSE) in the detection of mild cognitive impairment (MCI) and Alzheimer's disease (AD) in the elderly?

Int Psychogeriatr 2019; 31: 491–504.

15.

Kimani

Mwangi

Kwasa

, et al. Rethinking the design of low-cost point-of-care diagnostic devices. Micromachines (Basel) 2017; 8: 317.

16.

Parizkova

Lerch

Moffat

, et al. The effect of Alzheimer's disease on spatial navigation strategies. Neurobiol Aging 2018; 64: 107–115.

17.

Backman

Small

Fratiglioni

. Stability of the preclinical episodic memory deficit in Alzheimer's disease. Brain 2001; 124: 96–102.

18.

Heggland

Kvello

Witter

. Electrophysiological characterization of networks and single cells in the hippocampal region of a transgenic rat model of Alzheimer's disease. eNeuro 2019; 6: 0448-17.

19.

Smith

Chakrabarti

Mowery

, et al. Convergence of forepaw somatosensory and motor cortical projections in the striatum, claustrum, thalamus, and pontine nuclei of cats. Brain Struct Funct 2022; 227: 361–379.

20.

Grahn

Parkinson

Owen

. The role of the basal ganglia in learning and memory: neuropsychological studies. Behav Brain Res 2009; 199: 53–60.

21.

McCabe

Roediger III

McDaniel

, et al. The relationship between working memory capacity and executive functioning: evidence for a common executive attention construct. Neuropsychology 2010; 24: 222–243.

22.

Allain

Chaudet

Nicoleau

, et al. A study of action planning in patients with Alzheimer's disease using the zoo map test. Rev Neurol (Paris) 2007; 163: 222–230.

23.

Schutte

Jacoby

. Individual differences in executive attention and inhibitory control are related to spatial memory biases in adults. Q J Exp Psychol (Colchester) 2023; 76: 391–403.

24.

De Stefano

Fontanella

Impedovo

, et al. Handwriting analysis to support neurodegenerative diseases diagnosis: a review. Pattern Recognit Lett 2019; 121: 37–45.

25.

Piers

Devlin

Ning

, et al. Age and graphomotor decision making assessed with the digital clock drawing test: the Framingham Heart Study. J Alzheimers Dis 2017; 60: 1611–1620.

26.

Sabbagh

Boada

Borson

, et al. Rationale for early diagnosis of mild cognitive impairment (MCI) supported by emerging digital technologies. Jpad-J Prev Alzheimers Dis 2020; 7: 158–164.

27.

Rentz

Papp

Mayblyum

, et al. Association of digital clock drawing with PET amyloid and tau pathology in normal older adults. Neurology 2021; 96: E1844–E1854.

28.

Peter

Sandkamp

Minkova

, et al. Real-world navigation in amnestic mild cognitive impairment: the relation to visuospatial memory and volume of hippocampal subregions. Neuropsychologia 2018; 109: 86–94.

29.

Hirjak

Sambataro

Remmele

, et al. The relevance of hippocampal subfield integrity and clock drawing test performance for the diagnosis of Alzheimer's disease and mild cognitive impairment. World J Biol Psychiatry 2019; 20: 197–208.

30.

Duro

Cerveira

Santiago

, et al. Clock drawing test in mild cognitive impairment: correlation with cerebral perfusion in single-photon emission computed tomography. Neuropsychology 2019; 33: 617–632.

31.

Cheah

W-T

Hwang

J-J

Hong

S-Y

, et al. A digital screening system for Alzheimer disease based on a neuropsychological test and a convolutional neural network: system development and validation. JMIR Med Inform 2022; 10: e31106.

32.

Delazer

Zamarian

Djamshidian

. Handwriting in Alzheimer's disease. J Alzheimers Dis 2021; 82: 727–735.

33.

Kourtis

Regele

Wright

, et al. Digital biomarkers for Alzheimer's disease: the mobile/wearable devices opportunity. Npj Digital Med 2019; 2: 9.

34.

Yang

, et al. Exploratory research on key technology of human–computer interactive 2.5-minute fast digital early warning for mild cognitive impairment. Comput Intell Neurosci 2022; 2022: 2495330.

35.

Lio

Fadda

Doneddu

, et al. Digit-tracking as a new tactile interface for visual perception analysis. Nat Commun 2019; 10: 5392.

36.

Bera

Schalper

Rimm

, et al. Artificial intelligence in digital pathology—new tools for diagnosis and precision oncology. Nat Rev Clin Oncol 2019; 16: 703–715.

37.

Bosl

Tager-Flusberg

. EEG analytics for early detection of autism spectrum disorder: a data-driven approach. Sci Rep 2018; 8: 6828.

38.

Kasasbeh

Christensen

Parsons

, et al. Artificial neural network computer tomography perfusion prediction of ischemic core. Stroke 2019; 50: 1578–1581.

39.

Zhan

Mohan

Tarolli

, et al. Using smartphones and machine learning to quantify Parkinson disease severity: the mobile Parkinson disease score. JAMA Neurol 2018; 75: 876–880.

40.

Shoeibi

Ghassemi

Alizadehsani

, et al. A comprehensive comparison of handcrafted features and convolutional autoencoders for epileptic seizures detection in EEG signals. Expert Syst Appl 2021; 163: 113788.

41.

Gershman

Horvitz

Tenenbaum

. Computational rationality: a converging paradigm for intelligence in brains, minds, and machines. Science 2015; 349: 273–278.

42.

Prattichizzo

Meli

Malvezzi

. Digital handwriting with a finger or a stylus: a biomechanical comparison. IEEE Trans Haptic 2015; 8: 356–370.

43.

Marullo

Pozzi

Malvezzi

, et al. Analysis of postures for handwriting on touch screens without using tools. Sci Rep 2022; 12: 296.

44.

Sliwinski

Mogle

Hyun

, et al. Reliability and validity of ambulatory cognitive assessments. Assessment 2018; 25: 14–30.

45.

Pearce

JMS

. The “split brain” and Roger Wolcott Sperry (1913-1994). Rev Neurol (Paris) 2019; 175: 217–220.

46.

Rodriguez-Hidalgo

Garcia-Alba

Buxo

, et al. The pictorial screening memory test (P-MIS) for adults with moderate intellectual disability and Alzheimer's disease. Int J Environ Res Public Health 2022; 19: 10780.

47.

Breton

Casey

Arnaoutoglou

. Cognitive tests for the detection of mild cognitive impairment (MCI), the prodromal stage of dementia: meta-analysis of diagnostic accuracy studies. Int J Geriatr Psychiatry 2019; 34: 233–242.

48.

Stern

. No neuron is an island. Science (New York, NY) 2022; 378: 486–487.

49.

Villemagne

Doré

Burnham

, et al. Imaging tau and amyloid-β proteinopathies in Alzheimer disease and other conditions. Nat Rev Neurol 2018; 14: 225–236.

50.

Wang

Chen

Wang

, et al. A new smart 2-min mobile alerting method for mild cognitive impairment due to Alzheimer’s disease in the community. Brain Sci 2023; 13: 244.

51.

Hartley

Lever

Burgess

, et al. Space in the brain: how the hippocampal formation supports spatial cognition. Philos Trans R Soc B-Biol Sci 2014; 369: 20120510.

52.

deIpolyi

Rankin

Mucke

, et al. Spatial cognition and the human navigation network in AD and MCI. Neurology 2007; 69: 986–997.

53.

Yang

Xie

Zisserman

. It's about time: analog clock reading in the wild. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition 2022, pp.2508-2517.

54.

Mueller

Preische

Heymann

, et al. Diagnostic value of a tablet-based drawing task for discrimination of patients in the early course of Alzheimer's disease from healthy individuals. J Alzheimers Dis 2017; 55: 1463–1469.

55.

Ghaderyan

Abbasi

Saber

. A new algorithm for kinematic analysis of handwriting data; towards a reliable handwriting-based tool for early detection of Alzheimer's disease. Expert Syst Appl 2018; 114: 428–440.

56.

Ishikawa

Nemoto

, et al. Handwriting features of multiple drawing tests for early detection of Alzheimer's disease: a preliminary result. Stud Health Technol Inform 2019; 264: 168–172.

57.

Williams

O'Connell

Leonard

, et al. Development of the four-item letter and shape drawing test (LSD-4): a brief bedside test of visuospatial function. Psychiatry Res 2017; 247: 317–322.

58.

Kobayashi

Yamada

Shinkawa

, et al. Automated early detection of Alzheimer's disease by capturing impairments in multiple cognitive domains with multiple drawing tasks. J Alzheimers Dis 2022; 88: 1075–1089.

59.

Garre-Olmo

Faundez-Zanuy

Lopez-de-Ipina

, et al. Kinematic and pressure features of handwriting and drawing: preliminary results between patients with mild cognitive impairment, Alzheimer disease and healthy controls. Curr Alzheimer Res 2017; 14: 960–968.

A new early warning method for mild cognitive impairment due to Alzheimer's disease based on dynamic evaluation of the “spatial executive process”

Abstract

Objective

Methods

Results

Conclusion

Keywords

Introduction

Methods

HCI software design

Fingertip interaction digital biomarkers extraction

Intelligent analysis algorithms

Experimental settings

Participant recruitment

Experimental process

Fingertip interaction digital biomarker screening based on semistructured interviews

Data analysis

Results

Fingertip interaction digital biomarker screening results based on semistructured interviews

Exploring the cowarning efficacy of multidimensional fingertip interaction digital biomarkers analysis results

Discussion

Conclusions

Footnotes

Acknowledgments

Contributorship

Declaration of conflicting interests

Ethical approval

Funding

Guarantor

ORCID iDs

References