1. Introduction

Hemoglobin (Hb), a protein in mammalian blood that contains iron, plays a crucial role in transporting oxygen (O₂) [1]. Anemia manifests in a range of symptoms from mild, including weakness, fatigue, and dizziness, to severe complications like life-threatening cardiovascular collapse in acute cases [2]. Conversely, elevated hematocrit levels lead to increased blood viscosity, slowing blood flow and increasing the likelihood of thrombus formation, especially in smaller arterial vessels [3]. Consequently, the concentration of hemoglobin is a critical biomarker for both health assessment and diagnostic evaluation [4–6]. Traditionally, hemoglobin concentration is assessed using venous blood samples analyzed by hematology analyzers, or capillary blood samples with handheld Point of Care Testing (POCT) devices [7]. However, these invasive methodologies are associated with significant costs, and generate considerable medical waste [8]. Recently, the market has seen proactive development of non-invasive hemoglobin measurement devices, like the Masimo Pronto-7 (Masimo Corporation, CA), a non-invasive POCT hemoglobin analyzer. The manufacturer's description of the technology states that the SpHb sensor measures light absorption through the finger at 7 wavelengths to determine hemoglobin concentrations. Its unique signal processing algorithms translate the spectroscopic information to hemoglobin concentrations based on an empirically derived look-up table, analogous to conventional pulse oximetry [5,7]. In 2014, the Patino research team conducted measurements on 46 pediatric subjects, ranging in age from 2 months to 17 years (encompassing 140 data points), at Cincinnati Children’s Hospital Medical Center, finding a Pearson correlation coefficient of 0.76 between their measurements and the hemoglobin values from hospital blood draws [5]. In a similar vein, the Jost research team performed clinical measurements on 218 adult subjects, who had a median age of 35. They reported a Pearson correlation coefficient of 0.55 between their measurements and hospital blood values [9]. From the cited literature [5,9], it is evident that the accuracy of this empirically derived formula may be influenced by variations in the test populations. In addition, smartphones offer an ideal platform for noninvasive POCT applications, with emerging research on their potential role in assisting with hemoglobin estimation. Notably, a novel, noninvasive smartphone application developed by Mannino, R.G., et al. has demonstrated initial promising results in measuring hemoglobin concentrations using an unmodified smartphone [10]. The app employs an image algorithm that accurately analyzes the RGB composition of fingernail bed images for estimating hemoglobin concentration. Although initial proof-of-concept studies showed promising results, later evaluations by other teams revealed this noninvasive smartphone method's lower correlation, accuracy, and classification bias compared to invasive devices [11]. For example, noninvasive hemoglobin measurements by the smartphone application were poorly correlated with reference hemoglobin concentrations obtained from a hematology analyzer (Apple r²= 0.08; Android r² = 0.11). It was found that variations in smartphone software and hardware, such as differing operating systems between iOS and Android devices, as well as discrepancies in camera sensor specifications, are suspected contributors to the observed measurement inconsistencies between different smartphone models.

Diffuse Reflectance Spectroscopy (DRS), as a technology with potential clinical diagnostic applications, offers advantages such as rapid and non-invasive testing, making it widely used in measuring skin tissue characteristics. Among its variations, Spatially Resolved Diffuse Reflectance Spectroscopy (SRDRS) is particularly notable for its simple structure, potential for miniaturization, and lower cost, leading to its extensive application. Over the past decade, our team has successfully utilized SRDRS systems in numerous studies related to skin tissue component quantification. These include non-invasive measurement of collagen [12], evaluation of therapeutic response in keloid scar [13], detection of neonatal jaundice [14], and assessment of skin hydration [15], among others.

Wearable devices have gained increasing popularity in recent years due to their convenience and portability. Currently available wearable medical monitoring devices, such as smartwatches, can provide limited physiological information like heart rate and blood oxygen saturation. Although our current benchtop SRDRS prototype demonstrated a simple architecture and could provide ample information, it was not yet amenable to wearable applications. In this study, we aimed to develop a wearable SRDRS device capable of measuring the hemoglobin concentration of subjects. Validation across different populations will be carried out to assess the practicality of this novel hemoglobin monitoring system.

2. Material and methods

2.1 Wearable SRDRS measurement system

Our team has built a handheld DRS device for clinical measurements, utilizing mini-spectrometers (C12880 MA, Hamamatsu, Japan) and white light LEDs [16]. Building on this progress, we further miniaturized the system in this study by using LEDs that emit six different wavelengths and replacing the mini-spectrometer with photodiodes (PDs), thus evolving the device into a wearable form factor. Figure 1(a) depicts the block diagram of the wearable DRS device, which primarily consists of a control board and a sensor board, and is operated via a personal computer (PC). The control board contains a microcontroller unit (MCU) (LPC5500, NXP) to control the LED drivers and analog-to-digital converters (ADCs). The LEDs are activated sequentially with a driving current ranging from 20 to 60 mA, and the data collected by the PDs are relayed back to the MCU through the ADCs. As shown in Fig. 1(b), the size of the control board enclosure is 12.5 cm × 5.5 cm × 5.6 cm, and the sensor board is housed in a wrist device whose size is 3.2 cm × 3 cm × 1.7 cm.

 

Fig. 1. (a) Block diagram of the wearable DRS device. (b) Appearance of the wearable DRS system showing the control board enclosure and sensor board wrist device. (c) Close-up photo of the sensor board showing the arrangement of the LEDs and PDs. (PD: photodiodes, LED: light emitting diodes, MCU: microcontroller unit, ADC: analog-to-digital converters, PC: personal computer).

Download Full Size | PDF

Figure 1(c) illustrates the layout of the sensor board. We selected six custom LEDs (Epileds Technologies) with emission center wavelengths of 500 nm, 521 nm, 541 nm, 561 nm, 572 nm, and 599 nm, corresponding to the hemoglobin absorption features. The dimensions of the chosen surface mount LED chips were 330 µm × 275 µm. Additionally, the PDs utilized were 380 µm × 380 µm in active area (Tyntek). As depicted in Fig. 1(c), the sensor board contains 6 PDs and 12 LEDs, with a diameter of 24 mm. The leftmost vertical array consists of the 6 PDs, while the other two vertical arrays consist of the 12 LEDs. The LEDs in the same horizontal row have the same emission wavelength. The source-detector separations (SDS) between the PD array and each of the two LED arrays are 1 mm and 2 mm, respectively. These SDS values are consistent with those used in our previous work, where the utility of the 1 mm and 2 mm source-detector separations has been validated across our previous studies [17–19].

2.2 Recovery of skin optical properties and hemoglobin concentration

We incorporated an artificial neural network (ANN) which was adept at processing the data obtained from skin reflectance for interpreting skin optical properties, as detailed in prior literature [20]. In brief, the diffuse reflectance collected at various tissue optical properties was calculated employing the Monte Carlo method [21]. Although modeling a large number of photon trajectories with the Monte Carlo method is accurate, it is also computationally intensive. Therefore, we constructed artificial neural network (ANN) models, trained on Monte Carlo simulation results, to enable more efficient extraction of tissue optical properties from diffuse reflectance measured on skin. The models were trained with two inputs and two outputs. The forward model took the absorption coefficient and reduced scattering coefficient as inputs, and output the diffuse reflectance at 2 distances. The inverse model took the diffuse reflectance at 2 distances as input, and output the absorption coefficient and reduced scattering coefficient. For the Monte Carlo simulations, we presumed the tissue sample to be homogeneous, with a refractive index of 1.45. The Henyey-Greenstein phase function with an anisotropy factor of 0.8 was used in all simulations. In this study, the absorption coefficient (µ_a) range of interest was defined to be from 0.001 mm⁻¹ to 1 mm⁻¹, spanning a total range of 0.999 mm⁻¹. The reduced scattering coefficient (µ_s’) range was set between 0.1 mm⁻¹ and 3 mm⁻¹, covering 2.9 mm⁻¹. To construct an accurate training database for the ANN within this parameter space, the µ_a range was discretized into 100 equal bins with a step size of 0.001 mm⁻¹ per bin, while the µ_s’ range was divided into 30 equal bins with an incremental step of 0.10 mm⁻¹. This resulted in 100 µ_a values and 30 µ_s’ values, forming 3000 combinations of optical properties when paired together. The ANN was trained using the calculated diffuse reflectance database and the ‘trainbr’ MATLAB (MathWorks, MA) function, which employs Bayesian regularization to update weight and bias values based on Levenberg-Marquardt optimization. The Log-sigmoid transfer function “logsig” was chosen for the hidden layer to train these network models. Furthermore, based on the validation results of our team's ANN models [20], our optimized ANN models at 1 mm and 2 mm source-detector separations ensures accurate mapping of the measured diffuse reflectance back to the absorption and reduced scattering coefficients without issues of multiple indistinguishable solutions. When calculating the absorption and scattering of the test object, the process includes a calibration step as follows: First, input the known absorption coefficient and reduced scattering coefficient of the phantom into the forward ANN model to obtain the theoretical diffuse reflectance spectrum of the phantom. Measure the diffuse reflectance spectrum of the phantom, and divide it by the theoretical spectrum of the phantom to obtain the system response Next, divide the measured skin diffuse reflectance spectrum by the system response to obtain the true skin diffuse reflectance spectrum. Input this into the inverse ANN model to output the absorption coefficient and reduced scattering coefficient of skin. In this study, these ANN models were utilized to compute skin absorption and scattering spectra at wavelengths of 500 nm, 521 nm, 541 nm, 561 nm, 572 nm, and 599 nm, enabling detailed calculation of hemoglobin concentration in the examined subjects. The recovered absorption spectra were subsequently fitted to the established absorption spectra of the principal chromophores. The absorption spectra of the principal chromophores - hemoglobin, oxygenated hemoglobin, and melanin - were obtained from the Oregon Medical Laser Center database. This process allowed for the quantification of tissue chromophore concentrations using the following equation:

Additionally, our DRS system measures skin tissue, which generally contains less than 5% blood by volume in the probed region. This implies that the raw hemoglobin values derived from our DRS are inherently unequal to the hemoglobin values obtained from blood samples. Hence, a conversion process is typically required to map the raw hemoglobin values to those from invasive methods. We mapped the optically-derived hemoglobin concentrations (µM) to the hemoglobin concentrations (g/dL) obtained through blood sampling by performing a linear regression between the optical and invasive methods.

2.3 In-vivo human study

To validate our wearable DRS device's accuracy, we conducted experiments with two cohorts, approved by the Institutional Review Boards at National Cheng Kung University (Approval No. NCKU HREC-E-112-569-2) and National Cheng Kung University Hospital (No. B-ER-109-354).

Cohort-1 comprised 21 healthy participants, including 8 males and 13 females, aged 19-26 years. Prior to optical measurements, the wearable DRS device was disinfected following alcohol sterilization procedures. We then non-invasively acquired diffuse reflectance spectra at the wrist site. Afterwards, fingertip capillary blood was collected for obtaining benchmark hemoglobin concentrations using a HemoCue Hb 201+ (HemoCue AB, Sweden). Cohort-2 consisted of 9 elder participants, with 6 males and 3 females, aged between 50 and 83 years, recruited from the Department of Internal Medicine at National Cheng Kung University Hospital. Diffuse reflectance spectra were non-invasively measured using our wearable device. Blood samples from cohort-2 participants were collected to determine their hemoglobin concentrations using the hospital's complete blood count instrument. Optical measurements were directly followed by invasive blood draws for cohort-1 within a 30-minute window. For cohort-2 elderly patients, the optical measurements were carried out during a clinical visit, while venous blood samples were analyzed on a subsequent day within a 1-2 day span, with both assessments being conducted in a supine position.

This study adhered to the latest revision of the Declaration of Helsinki. For both cohorts, informed consent was obtained prior to measurements. Participants’ age, gender and other details were also documented. Each optical measurement contained 5 readings, and the average value was taken as the measurement value. We examined correlations between invasive and non-invasive hemoglobin levels to validate our wearable device's accuracy.

3. Results and discussion

3.1 Measurement results from the young participants group

As detailed in Section 2.3, cohort-1 comprised 21 young, healthy participants: 8 males and 13 females. Their ages ranged from 19 to 26 years, with a mean age of 21.71 years. Height distribution ranged from 153 to 181 cm, with an average of 166.8 cm. The body weight of participants varied between 40 and 107 kg, with a mean weight of 62.1 kg. Systolic/Diastolic blood pressure measurements ranged from 83/49 to 141/95 mmHg, with mean values of 109.0/72.1 mmHg. Heart rates of the participants varied between 49 and 99 beats per minute, with an average rate of 75.4 beats per minute. The HemoCue Hb 201 + invasive POCT device was utilized as the reference standard. The Hb levels from the 21 participants extended from 9.6 to 17.2 g/dL, with a mean level of 13.6 g/dL. The HemoCue system and wearable DRS device hemoglobin values were collected for each subject and plotted as shown in the scatter plot in Fig. 2(a), resulting a high Pearson correlation coefficient (r) of 0.91 with a linear regression equation of y = 0.4212x + 0.9809. The grey shaded area around the regression line represents the 95% confidence interval. If the measured tissue were uniformly distributed with blood and other skin components were also evenly distributed, the regression line's intercept obtained from our measurements would be zero. However, the living skin tissue measured by our DRS system does not constitute a homogeneously diluted structure of whole blood. Additionally, the distribution of blood vessels varies among individuals, and the vascular distribution within the measured tissue differs with each measurement. Therefore, the regression curve obtained through our diffuse technique exhibits a non-zero intercept, as also observed in other similar studies [9].

 

Fig. 2. (a) Correlation analysis between hemoglobin concentration measured by our non-invasive wearable DRS device and Hemocue method (n = 21) in young participant groups (cohort-1). (b) Correlation analysis between hemoglobin concentration measured by our non-invasive wearable DRS device and hospital venous blood CBC test (n = 9) in internal medicine patient groups (cohort-2).

Download Full Size | PDF

3.2 Measurement results from the elder participants group

Cohort-2, consisting of 9 elder participants (6 males and 3 females) from the Department of Internal Medicine at National Cheng Kung University Hospital, were aged 50 to 83 years. The distribution of height ranged from 150 to 173 cm, with a mean height of 163.5 cm. Body weight varied between 31 and 107 kg, and the mean weight was 71.8 kg. Systolic/Diastolic blood pressure measurements ranged from 107/64 to 157/102 mmHg, with mean values of 134.6/82.1 mmHg. Heart rates ranged from 63 to 101 beats per minute, with an average rate of 82.4 beats per minute. We employed venous blood complete blood count (CBC) tests conducted by the hospital as a reference for Hb levels, which ranged from 9.2 to 15.8 g/dL, with a mean level of 11.9 g/dL. We collected hemoglobin levels measured by our wearable DRS device alongside values obtained from the hospital's venous blood CBC tests and depicted the data in a scatter plot, as illustrated in Fig. 2(b). A Pearson correlation coefficient of 0.89 was obtained, demonstrating a strong correlation between the two measurement methods. The derived regression equation is y = 0.3721x + 1.5835. The grey shaded area encompassing the regression line represents the 95% confidence interval.

3.3 Combined analysis of measurements from young and elder participants

According to the normal hemoglobin standards defined by the World Health Organization (WHO) (13-18 g/dL for males, 12-16 g/dL for females) [24], the hemoglobin levels of our two study cohorts spanned 9.2 to 17.2 g/dL, fully encompassing the standard concentration boundaries. As depicted in Fig. 2, our wearable DRS device exhibits strong correlations between the estimated tissue hemoglobin concentration and the hemoglobin values measured invasively both in the younger group and in the elderly group. The slope of the regression curve cannot directly represent the volume fraction of blood in tissue, but it should be related. Our study indeed identifies subtle differences in the slopes of regression equations between two cohorts, potentially due to variations in blood volume fractions, with older individuals possibly having lower blood volume fractions. Investigating this relationship poses an interesting question with significant implications, meriting further research.

To examine potential differences across populations, we conducted a combined analysis by plotting the scatter points from Figs. 2(a) and 2(b) together, as shown in Fig. 3(a). Importantly, we combined the data points from both cohorts for an overall correlation analysis in Fig. 3(a). The optical hemoglobin values from the two cohorts were merged directly, with no mathematical transformation applied to either dataset. This combined correlation analysis yielded a Pearson correlation coefficient of 0.90, indicating a high degree of correlation. The regression line is described by the equation y = 0.4052x + 1.1852. The 95% confidence bounds are signified by the shaded grey region surrounding the regression line. A key benefit of Pearson correlation analysis is in revealing the strength of linear association between the hemoglobin values measured by our optical method and the invasive reference standard. However, correlation coefficients alone do not confirm interchangeability or absolute agreement between two measurement techniques. Importantly, correlation curves can also be disproportionately influenced by outliers and may fail to expose subgroup differences if heterogeneity exists within the collective dataset. Therefore, we supplemented the Pearson correlation testing with Bland-Altman analysis, which directly quantifies mean bias and limits of agreement through graphical means, irrespective of correlation values. In contrast to methods involving blood samples, our DRS system primarily assesses skin tissue, where the blood volume typically constitutes less than 5% of the probed region. As a result, the raw hemoglobin values obtained from our DRS system are not directly equivalent to those derived from blood samples. To facilitate a Bland-Altman analysis, it is essential to convert these values through a linear regression, establishing a correlation between hemoglobin concentrations measured optically (in µM) and those obtained from blood sampling (in g/dL). Utilizing the equation derived from Fig. 3(a), the optically measured hemoglobin values (in µM) were converted into g/dL by subtracting 1.1852 and then dividing by 0.4052 to yield the converted optical hemoglobin values. Bland-Altman analysis was conducted on the converted data between the wearable DRS device and blood sample assay results, with the findings plotted in Fig. 3(b). The analysis revealed a mean bias of 11.55 × 10⁻¹⁴g/dL, indicated by the central solid red line in Fig. 3(b). The upper and lower 95% limits of agreement (LOA) are depicted by the dashed blue lines, calculated to be -1.98 to 1.98 g/dL. The evenly dispersed data points clustered around the mean bias line and residing within the LOA boundaries demonstrate good concordance between our optical spectroscopy approach and the invasive hemoglobin standard at a wide range of hemoglobin values.

 

Fig. 3. (a) A correlation analysis that combines two cohorts of participants (n = 30) between hemoglobin concentration measured by our wearable DRS device and the invasive method. The points highlighted in blue indicate deviations from the trend. The points highlighted in red indicate the two participants with the highest and lowest mean absorption coefficients. (b) A Bland-Altman plot to assess the agreement between two methods of hemoglobin measurement.

Download Full Size | PDF

Upon inspection of all participants, we identified an 81-year-old individual in cohort-2, who weighed only 31 kilograms with a BMI of 13.8, markedly lower than the average BMI of 23.9. This participant's data, highlighted by a blue circle in Fig. 3(a), deviates from the regression line trend. We observed that this individual's lower BMI resulted in poor device contact due to the skinny wrist and prominent bone curvature. A common criterion of wearable optical devices is to provide good device-tissue contact. Future developments aim to miniaturize the device further, ensuring a good sensor-to-skin contact, thereby mitigating issues related to optical contact.

Previous studies utilizing optical methods to measure arterial blood oxygen levels have demonstrated that dark skin decreases the accuracy of pulse oximeters at low oxygen saturation levels [25]. To assess the impact of skin tones on our DRS device's accuracy, we selected two participants: one with the highest average µ_a, X002 (0.607 mm⁻¹), and the other with the lowest average µ_a, X012 (0.181 mm⁻¹), as depicted in the measurement site photographs in Fig. 4(a) and 4(b), respectively. The invasive hemoglobin values for X002 and X012 were 13.7 and 10.9 g/dL, respectively, while the converted optical hemoglobin values were 15.4 and 9.2 g/dL, as circled in red in Fig. 3(a). The mean µ_a across the six wavelengths was 0.426 mm⁻¹ higher for X002 compared to X012. In our analysis, the contribution of melanin, ascertained through chromophore fitting, was subtracted from the skin absorption spectrum. The resultant melanin-subtracted spectra of X002 and X012 are presented in Fig. 4(e) and 4(f), respectively. After subtracting the effect of melanin, the mean for X002 became 0.094 mm⁻¹ and the mean for X012 became 0.090 mm⁻¹, which had comparable levels. Our results suggest that the hemoglobin values of subjects can be properly recovered using our DRS device even when the subjects have different skin tones.

 

Fig. 4. (a) Photograph of participant X002 with highest lowest mean absorption coefficient. (b) Photograph of participant X012 with lowest mean absorption coefficient. (c) Absorption spectra of participant X002 and (d) participant X012 measured by our wearable DRS device. (e) Melanin-subtracted absorption spectra for participant X002 and (f) participant X012.

Download Full Size | PDF

The predominant non-invasive optical medical device currently available for hemoglobin monitoring is the Masimo Radical-7 Pulse CO-Oximeter. In 2014, the research team led by Al-Khabori utilized the Masimo Pronto-7 for clinical studies, collecting blood samples from 106 donors at a blood bank (average age 27 years, range: 18-49 years) for analysis at the hospital [26]. The correlation between the Masimo Pronto-7 readings and the hospital laboratory results yielded a Pearson correlation coefficient of 0.46. The Bland-Altman plot indicated a mean bias of 0.22 g/dL with limits of agreement (LOA) of 4.98 g/dL, ranging from -2.27 g/dL to 2.71 g/dL. Our preliminary results indicate an improved consistency, with a better Pearson correlation and a narrower LOA, despite our DRS study having a smaller sample size. We infer that the enhanced performance could originate from our incorporation of chromophore fitting to accurately isolate and quantify hemoglobin, unaffected by other major skin chromophores like melanin. On the other hand, according to previous research by Ashish Jain et al. [27], 115 volunteers were assessed using a HemoCue system compared to an automated cell counter for hemoglobin quantification. The HemoCue a301 POCT device yielded optimal results, demonstrating a bias of 0.24 g/dL, with lower and upper LOA at -0.57 g/dL and 1.06 g/dL, respectively. These results demonstrate that the HemoCue system also has a certain degree of error, but the tighter LOA range and higher measurement precision indicate that the accuracy of optical techniques is still inferior to that of invasive methods. To further enhance the performance of our device, we could add more wavelengths to the sensor board to facilitate more accurate spectral analysis or we could replace the LEDs with laser diodes to determine more precise spectral information. Besides, we currently use 3 chromophores in the fitting, we will investigate whether adding more chromophores in the fitting could further decrease LOA. Additionally, as mentioned previously, reducing the size of the device to allow it to conform more closely to the skin is another objective moving forward. Future efforts would also focus on validating our wearable DRS device across heterogeneous groups encompassing diverse age ranges, ethnicities, and genders under various clinical conditions. Broadening the participant demographics and skin phenotypes would allow us to rigorously validate that the precision of our technology is not affected by differences in individual characteristics like pigmentation, skin condition, age or gender.

4. Conclusion

This study involved the development and validation of a miniaturized, wearable DRS system designed for noninvasive hemoglobin measurement. The device underwent testing on two cohorts: one of young healthy adults and another of elder patients, together comprising 30 subjects with hemoglobin levels ranging from 9.2 to 17.2 g/dL. Our wearable DRS method exhibited a strong correlation with invasive reference methods, achieving an overall Pearson coefficient of 0.90 across both cohorts. The Bland-Altman analysis showed a tight 95% LOA ranging from -1.98 to 1.98 g/dL between the wearable DRS and invasive hemoglobin values, following the conversion of optical data. These results validate the ability of our wearable DRS device to accurately quantify a wide range of hemoglobin levels in-vivo.

Compared to existing optical hemoglobin monitoring technologies, our wearable DRS system achieved better Pearson correlation and tighter LOA. Future enhancements, including replacing LEDs with laser diodes, could further improve measurement accuracy. Replacing the current LED light sources with laser diodes offers potential advantages in quantitation accuracy due to superior beam quality that provides higher optical energy density and narrower emission bandwidths compared to LEDs. This enables more defined hemoglobin absorption spectra. However, laser diodes have substantially higher costs that may impede scalability. We aim to build upon our ongoing miniaturization efforts to further reduce the sensor footprint for better conformity to wrist anatomy. This can ameliorate device-skin interface variability that degrades measurement precision. We can also refine the geometrical layout to achieve more uniform optical path lengths across individual source-detector pairs. These concerted improvements will facilitate our goal of reliable, whole-blood hemoglobin monitoring. Additionally, continued miniaturization will help reduce contact variability. Validation studies encompassing a wide range of demographics and clinical settings, coupled with enhancements to create a fully integrated system, will facilitate the seamless adoption of this technology in both medical and consumer wellness markets. This will facilitate effortless tracking of personal hemoglobin values, along with other functional information, especially if additional wavelengths are incorporated. To enable consumer adoption of this hemoglobin sensing technology in wearable devices, engineering optimizations should address user comfort through ergonomic miniaturization, durability via ruggedization testing, and usability by refining the interface and integrating with smart-watches and smart-rings.

This noninvasive hemoglobin assessment technology offers significant advantages for patient groups needing regular hemoglobin monitoring. Compared to traditional blood tests, it provides improving management of chronic diseases such as kidney and cardiovascular diseases. It also eliminates the need for hospital visits and needle use in anemia screenings for the elderly and children, while pregnant women benefit from safer hemoglobin and iron level monitoring. Furthermore, wearable versions of this technology can aid in assessing donor suitability and informing transfusion decisions in surgical settings, streamlining clinical processes. Ultimately, the technology's capacity for continuous, noninvasive hemoglobin monitoring promises to enhance patient care across various medical contexts where accurate hemoglobin measurement is critical.