Non-invasive hemoglobin estimation from conjunctival images using deep learning.

Hemoglobin, a crucial protein found in erythrocytes, transports oxygen throughout the body. Deviations from optimal hemoglobin levels in the blood are linked to medical conditions, serving as diagnostic markers for certain diseases. The hemoglobin level is usually measured invasively with different devices using the blood sample. In the physical interpretation, some signs are traditionally used. These signs are the palms, face, nail beds, pallor of the conjunctiva, and palmar wrinkles. Studies have shown that conjunctival pallor can yield more effective results in detecting anemia than the pallor of the palms or nail beds. This study is aimed to predict the hemoglobin level by deep learning method, non-invasive, cheap, fast, high accuracy, and without creating medical waste. In this context, conjunctival images and age, weight, height, gender, and hemoglobin values were collected from 388 people who donated blood to the Turkish Red Crescent. A dataset was generated by augmenting the gathered data with body mass index data. Within the scope of this investigation, the limits of agreement (LoA) value at a 95% confidence interval was computed to be 1.23 g/dL, while the bias was established as 0.26 g/dL. The mean absolute percentage error (MAPE) values were determined to be 3.4%, and the root mean squared error (RMSE) was calculated to be 0.68 g/dL. These findings exhibit a successful outcome compared to similar investigations, signifying that this non-invasive method can be employed for hemoglobin level estimation. Furthermore, the estimated hemoglobin levels could aid in diagnosing several hemoglobin-related ailments.

An intelligent non-invasive system for automated diagnosis of anemia exploiting a novel dataset.

Anemia is a condition in which the oxygen-carrying capacity of red blood cells is insufficient to meet the body's physiological needs. It affects billions of people worldwide. An early diagnosis of this disease could prevent the advancement of other disorders. Traditional methods used to detect anemia consist of venipuncture, which requires a patient to frequently undergo laboratory tests. Therefore, anemia diagnosis using noninvasive and cost-effective methods is an open challenge. The pallor of the fingertips, palms, nail beds, and eye conjunctiva can be observed to establish whether a patient suffers from anemia. This article addresses the above challenges by presenting a novel intelligent system, based on machine learning, that supports the automated diagnosis of anemia. This system is innovative from different points of view. Specifically, it has been trained on a dataset that contains eye conjunctiva photos of Indian and Italian patients. This dataset, which was created using a very strict experimental set, is now made available to the Scientific Community. Moreover, compared to previous systems in the literature, the proposed system uses a low-cost device, which makes it suitable for widespread use. The performance of the learning algorithms utilizing two different areas of the mucous membrane of the eye is discussed. In particular, the RUSBoost algorithm, when appropriately trained on palpebral conjunctiva images, shows good performance in classifying anemic and nonanemic patients. The results are very robust, even when considering different ethnicities.

Brief Resolved Unexplained Events: Analysis of an Apparent Life Threatening Event Database.

OBJECTIVE: To identify the proportion of patients previously diagnosed with apparent life-threatening events (ALTE) who would meet criteria for brief resolved unexplained events (BRUE) and to identify rates of adverse outcomes in subgroups: ALTE not meeting criteria for BRUE, lower-risk BRUE, and higher-risk ALTE. METHODS: We performed a secondary analysis of a single-center prospective registry of patients diagnosed with ALTE in a tertiary care emergency department from March 1, 1997 to October 31, 2007. We identified the proportion of patients meeting criteria for BRUE, and the proportion of patients with BRUE meeting lower-risk criteria. We assessed outcomes of patients in subgroups. RESULTS: Seven hundred and sixty-two patients were included. Adverse outcomes included recurrent ALTE (n = 49), aspiration (n = 9), trauma (n = 8), and death (n = 4). Three hundred and twenty-six of 762 (42.8%) met criteria for BRUE. Seventy of 326 (21.5%) met criteria for lower-risk BRUE. Adverse outcomes occurred in 40 of 436 (9.2%) with ALTE not meeting criteria for BRUE, 2 of 70 (2.9%) with lower-risk BRUE, and 23 of 256 (9.0%) with higher-risk BRUE. Of 4 patients who died, 1 had an ALTE not meeting criteria for BRUE and 3 had non-lower-risk BRUE. The BRUE risk criteria identified all BRUE patients that died or had substantial morbidity as higher-risk. CONCLUSIONS: Less than half of patients with ALTE meet criteria for BRUE. Of those who do, one-fifth is lower-risk. In this series, the risk-stratification in the BRUE criteria identified those patients at highest risk of adverse outcomes. Further research is required to risk-stratify patients with BRUE.

A case of recurrent facial pain associated with a Pourfour du Petit syndrome: a new entity?

We report the case of a 38-year-old woman with a history of migraine who experienced an association of recurrent unilateral facial pain and Pourfour du Petit syndrome. The episodes occurred for between a few seconds and up to 3 minutes up to 6 times a day mimicking short-lasting unilateral neuralgiform headaches with cranial autonomic symptoms. No lesional cause was found and the use of topiramate led to a nearly complete disappearance of the episodes. This new entity raises the question of a novel autonomic dysfunction in short-lasting unilateral neuralgiform headaches with cranial autonomic symptoms or an unexpected presentation of migraine.

Visual diagnosis: newborn who has unilateral color change. Diagnosis: Harlequin color change.

HCC is a benign, transient, self-limited condition encountered in neonates. Prompt recognition is required to avoid unnecessary diagnostic testing. No treatment is needed, and parental reassurance is adequate.

The 2011 PEMpix photo competition award winner: Vanessa Perez, MD: An unusual case of pallor, the eyes have it!

This is a case report of 7-year-old boy who presented with diplopia. Physical examination was remarkable for pallor and right-eye proptosis. Differential diagnosis of this chief complaint and the management of the suspected condition are discussed. This case was presented at the Section of Emergency Medicine Meeting at the National Conference and Exhibition of the American Academy of Pediatrics in 2011 and was awarded first place in the PEMpix photo competition.

Evaluation of WHO haemoglobin colour scale & palmar pallor for screening of anaemia among children (6-35 months) in rural Wardha, India.

BACKGROUND & OBJECTIVE: Anaemia is a public health problem in India, particularly in women and children. Clinical diagnosis by health workers may be inaccurate. It is therefore important to evaluate simple methods for diagnosis of anaemia in rural set up with no or minimal laboratory facilities. We carried out this study to evaluate the performance of haemoglobin colour scale and palmar pallor against filter paper cyanmethaemoglobin method as screening methods for anaemia in children 6-35 months of age. METHODS: A total of 772 children between 6-35 months were studied from three primary health centres of Wardha district, central India, by house-to-house visit. The child was examined for presence of palmar pallor and haemoglobin estimated by haemoglobin colour scale (HCS) and filter paper cyanmethaemoglobin method independently by two persons. RESULTS: Haemoglobin colour scale had sensitivity of 89 per cent and specificity of 97 per cent in detecting anaemia in children. Pallor had sensitivity of 73 per cent and specificity of 98 per cent. The predictive values were 99 per cent and 69 per cent for positive and negative HCS result. Similarly, the predictive values were 99 and 48 per cent for positive and negative pallor result. Correct diagnosis of anaemia was 34 and 0.1 times common if HCS test result was positive (Hb < 110.0 g/l) and negative (Hb > 110.0 g/l) for anaemia. INTERPRETATION & CONCLUSION: HCS fulfills the requirement of field test for screening of anaemia in resource poor country. It also has acceptable precision and accuracy and hence can be used in national anaemia control programmes.

Evaluation of the WHO Haemoglobin Colour Scale for diagnosis of anaemia in children and pregnant women as used by primary health care nurses and community health workers in western Kenya.

OBJECTIVES: To evaluate the diagnostic accuracy of the WHO Haemoglobin Colour Scale (HCS) for anaemia in three groups of children aged 2 months to 2 years (sick children, those visiting an immunization clinic and a community-based random sample of children) and a sample of pregnant women. METHODS: Finger-prick blood samples were taken from all consenting participants. Haemoglobin (Hb) levels from the HCS were compared with results from a HemoCue portable haemoglobinometer. Sensitivity, specificity and positive and negative predictive values for the HCS were calculated. RESULTS: A total of 457 sick children, 336 children visiting immunization clinics, 454 children from the community at large and 643 pregnant women participated. The prevalence of anaemia (Hb<11 g/dl) in these groups was 87%, 79%, 74% and 52%, respectively. The prevalence of severe anaemia (Hb<7 g/dl) was 24%, 11%, 10% and 2%, respectively. The sensitivity of the HCS for anaemia ranged from 60% to 79% and specificity from 59% to 94%. The sensitivity of the HCS for severe anaemia ranged from 24% to 63% and the specificity from 97% to 100%. Through use of the HCS, the proportion of sick, anaemic children visiting peripheral health facilities diagnosed and treated for anaemia would increase from 3% to 65%. CONCLUSIONS: In an area with high prevalence of anaemia among sick children, use of the HCS has the potential to significantly increase the proportion of sick, anaemic children who are diagnosed with anaemia and given appropriate treatment. Further evaluations of the effect of the use of the HCS on treatment practices at the health facility level are required.

Anemia hemolítica autoinmune por anticuerpos calientes: reporte de un caso pediátrico / Autoimmune hemolytic anemia, warm antibody: a case report pediatric

Las anemias hemolíticas autoinmunitarias se caracterizan por la presencia de inmunoglobulinas en la superficie eritrocitaria dirigidas contra los determinantes antigénicos de los hematíes. La anemia hemolítica autoinmune por anticuerpos calientes se caracteriza porque los autoanticuerpos actúan a la temperatura del organismo (37°C), son de clase IgG y la hemólisis es predominantemente extravascular, siendo el tipo más frecuente de anemia hemolítica autoinmune en los niños de 2-12 años de edad. Sus manifestaciones clínicas son postración, palidez, ictericia, fiebre y hemoglobinuria. El diagnóstico de las AHAI se establece con la prueba de Coombs. La administración de corticoesteroides constituye el tratamiento inicial de elección. Se presenta el caso de una preescolar femenina de tres años de edad procedente del medio rural, quien exhibe las características clínicas, paraclínica y epidemiológicas de anemia hemolítica autoinmune por anticuerpos calientes, con respuesta satisfactoria a la terapia con esteroides.

Recognition of pallor associated with severe anaemia by primary caregivers in western Kenya.

OBJECTIVES: To explore which pallor signs and symptoms of severe anaemia could be recognized by primary caregivers following minimal instructions. METHODS: Data from three community-based cross-sectional surveys were used. Test characteristics to predict haemoglobin (Hb) concentrations < 5 and < 7 g/dl were compared for different combinations of pallor signs (eyelid, tongue, palmar and nailbed) and symptoms. RESULTS: Pallor signs and haemoglobin levels were available for 3782 children under 5 years of age from 2609 households. Comparisons of the sensitivity and specificity at a range of haemoglobin cut-offs showed that Hb < 5 g/dl was associated with the greatest combined sensitivity and specificity for pallor at any anatomical site (sensitivity = 75.6%, specificity = 63.0%, Youden index = 38.6). Higher or lower haemoglobin cut-offs resulted in more children being misclassified. Similar results were obtained for all individual pallor sites. Combining a history of soil eating with pallor at any site improved the sensitivity (87.8%) to detect Hb < 5 g/dl with a smaller reduction in specificity (53.3%; Youden index 41.1). Other combinations including respiratory signs or poor feeding resulted in lower accuracy. CONCLUSION: Primary caregivers can recognize severe anaemia (Hb < 5 g/dl) in their children, but only with moderate accuracy. Soil eating should be considered as an additional indicator of severe anaemia. The effect of training caretakers to improve recognition of severe anaemia and care-seeking behaviour at the household level should be assessed in prospective community-based studies.

Malaria diagnosis and treatment under the strategy of the integrated management of childhood illness (IMCI): relevance of laboratory support from the rapid immunochromatographic tests of ICT Malaria P.f/P.v and OptiMal.

The algorithm developed for the integrated management of childhood illness (IMCI) provides guidelines for the treatment of paediatric malaria. In areas where malaria is endemic, for example, the IMCI strategy may indicate that children who present with fever, a recent history of fever and/or pallor should receive antimalarial chemotherapy. In many holo-endemic areas, it is unclear whether laboratory tests to confirm that such signs are the result of malaria would be very relevant or useful. Children from a holo-endemic region of Tanzania were therefore checked for malarial parasites by microscopy and by using two rapid immunochromatographic tests (RIT) for the diagnosis of malaria (ICT Malaria P.f/P.v and OptiMal. At the time they were tested, each of these children had been targeted for antimalarial treatment (following the IMCI strategy) because of fever and/or pallor. Only 70% of the 395 children classified to receive antimalarial drugs by the IMCI algorithm had malarial parasitaemias (68.4% had Plasmodium falciparum trophozoites, 1.3% only P. falciparum gametocytes, 0.3% P. ovale and 0.3% P. malariae). As indicators of P. falciparum trophozoites in the peripheral blood, fever had a sensitivity of 93.0% and a specificity of 15.5% whereas pallor had a sensitivity of 72.2% and a specificity of 50.8%. The RIT both had very high corresponding sensitivities (of 100.0% for the ICT and 94.0% for OptiMal) but the specificity of the ICT (74.0%) was significantly lower than that for OptiMal (100.0%). Fever and pallor were significantly associated with the P. falciparum asexual parasitaemias that equalled or exceeded the threshold intensity (2000/microl) that has the optimum sensitivity and specificity for the definition of a malarial episode. Diagnostic likelihood ratios (DLR) showed that a positive result in the OptiMal test (DLR = infinity) was a better indication of malaria than a positive result in the ICT (DLR = 3.85). In fact, OptiMal had diagnostic reliability (0.93) which approached that of an ideal test and, since it only detects live parasites, OptiMal is superior to the ICT in monitoring therapeutic responses. Although the RIT may seem attractive for use in primary health facilities because relatively inexperienced staff can perform them, the high cost of these tests is prohibitive. In holo-endemic areas, use of RIT or microscopical examination of bloodsmears may only be relevant when malaria needs to be excluded as a cause of illness (e.g. prior to treatment with toxic or expensive drugs, or during malaria epidemics). Wherever the effective drugs for the first-line treatment of malaria are cheap (e.g. chloroquine and Fansidar), treatment based on clinical diagnosis alone should prove cost-saving in health facilities without microscopy.

The influence of experience and specialisation on the reliability of a common clinical sign.

OBJECTIVES: To explore the influence of experience and specialisation on clinical judgement by comparing accuracy in diagnosing anaemia between a consultant general surgeon, a consultant ophthalmologist and their registrars. PATIENTS AND METHODS: Conjunctival inspection of 101 patients, subsequent correlation with haemoglobin concentration. MAIN OUTCOME MEASURES: Number of correct and incorrect diagnoses of anaemia. RESULTS: 54 patients were anaemic and 47 were not. Overall accuracy in diagnosing anaemia ranged from 0.61-0.69, sensitivity 0.52-0.65 and specificity 0.62-0.83. Agreements between pairs of examiners were 0.68-0.81, with kappa values of 0.36-0.60 when adjusted for chance agreement. CONCLUSIONS: Neither experience nor specialisation significantly influenced our ability to diagnose anaemia, based on conjunctival inspection. Without critical analysis of clinical signs, we are unaware of their diagnostic limitations.

Field trial of a haemoglobin colour scale: an effective tool to detect anaemia in preschool children.

The objective of this study was to evaluate the performance of the Haemoglobin Colour Scale, developed by Stott and Lewis, to diagnose anaemia in a primary health care setting where anaemia was prevalent and severe. Three measures of anaemia were compared in 535 preschool children: haemoglobin based on the Haemoglobin Colour Scale, clinical assessment in three sites (conjunctiva, palm and nail bed) and haemoglobin based on a digital haemoglobinometer (HemoCue method) taken as gold standard. A statistically significant correlation (r = 0.80, coefficient = 0.77 and Y intercept = 2.33) was obtained between the results of the Haemoglobin Colour Scale and the HemoCue. In more than 80% of cases, the difference between the colour scale readings and the results of the HemoCue was within 1 g/dl. Of 415 anaemic children (Hb < 11 g/dl by HemoCue), 85.2% were so identified by the Haemoglobin Colour Scale and 19.7% were classified anaemic by clinical pallor. Of 19 severely anaemic children (Hb < 7 g/dl by HemoCue), 73.6% were identified as severely anaemic and 100% were classified as anaemic by the colour scale, 61.1% were classified as anaemic using clinical pallor. We found the Haemoglobin Colour Scale to be a useful tool in identifying anaemic and severely anaemic children. Efficiencies in term of cost, accuracy and time make it an important resource in primary health care settings in developing countries. Further testing with other staff in other settings is recommended to determine the usefulness of large-scale distribution.

Clinical pallor is useful to detect severe anemia in populations where anemia is prevalent and severe.

Clinical pallor is recommended as a simple way to detect severe anemia, but more data are needed on its accuracy and usefulness when assessed by nonphysicians in diverse settings. We measured hemoglobin and trained non-physician health workers to assess clinical pallor of the conjunctiva, palm and nail beds in five population samples in Nepal and Zanzibar, where severe anemia is common. In total, 5,760 individuals were examined, 3,072 of whom were anemic and 192 of whom had severe anemia (hemoglobin <70 g/L). The prevalence of pallor did not correspond to the prevalence of anemia or severe anemia in the groups studied. However, in all studies, pallor at each anatomical site was associated with a significantly lower hemoglobin concentration. The relative performance of different anatomical sites was not consistent among studies, and we recommend that multiple sites be assessed. Pallor at any of the three sites detected severe anemia with >84% specificity. However, the sensitivity varied from 81% in Nepalese postpartum women to 29% in Zanzibari preschoolers in 1996. Overall estimates for sensitivity and specificity were 50 and 92%, respectively. Although imperfect, use of pallor to screen and treat severe anemia by primary care providers is feasible and worthwhile where severe anemia is common. Usually, the majority of persons with severe anemia will be detected at practically no cost. Many people who are not severely anemic will also receive treatment, but the costs of this error are low compared to the benefits.

Evaluation of clinical signs to diagnose anaemia in Uganda and Bangladesh, in areas with and without malaria.

The object of this study was to assess the ability of pallor and other clinical signs, including those in the Integrated Management of Childhood Illness (IMCI) guidelines developed by WHO and UNICEF, to identify severe anaemia and some anaemia in developing country settings with and without malaria. A total of 1226 and 668 children aged 2 months to 5 years were prospectively sampled from patients presenting at, respectively, a district hospital in rural Uganda and a children's hospital in Dhaka, Bangladesh. The study physicians obtained a standardized history and carried out a physical examination that included pallor, signs of respiratory distress, and the remaining IMCI referral signs. The haematocrit or haemoglobin level was determined in all children with conjunctival or palmar pallor, and in a sample of the rest. Children with a blood level measurement and assessment of pallor at both sites were included in the anaemia analysis. Using the haematocrit or haemoglobin level as the reference standard, the correctness of assessments using severe and some pallor and other clinical signs in classifying severe and some anaemia was determined. While the full IMCI process would have referred most of the children in Uganda and nearly all the children in Bangladesh with severe anaemia to hospital, few would have received a diagnosis of severe anaemia. Severe palmar and conjunctival pallor, individually and together, had 10-50% sensitivity and 99% specificity for severe anaemia; the addition of grunting increased the sensitivity to 37-80% while maintaining a reasonable positive predictive value. Palmar pallor did not work as well as conjunctival pallor in Bangladesh for the detection for severe or some anaemia. Combining "conjunctival or palmar pallor" detected 71-87% of moderate anaemia and half or more of mild anaemia. About half the children with no anaemia were incorrectly classified as having "moderate or mild" anaemia. Anaemia was more easily diagnosed in Uganda in children with malaria. Our results show that simple clinical signs can correctly classify the anaemia status of most children. Grunting may serve as a useful adjunct to pallor in the diagnosis of severe anaemia. Conjunctival pallor should be added to the IMCI anaemia box, or the guidelines need to be adapted in regions where palmar pallor may not readily be detected.

PIP: The capability of pallor and other clinical signs to identify anemia was evaluated in developing country settings with malaria (Uganda) and without malaria (Bangladesh). Enrolled were 1226 children, 2 months to 5 years of age, who presented to a rural district hospital in Fort Portal, Uganda, and 668 children from a children's hospital in Dhaka, Bangladesh. Physicians examined the children for conjunctival pallor, respiratory distress, and other clinical signs of anemia (i.e., palmar pallor) included in the World Health Organization's Integrated Management of Childhood Illness (IMCI) guidelines. 58% of Ugandan and 47% of Bangladeshi children had pallor of at least one site. Hematocrit or hemoglobin levels were measured in all children with pallor of the conjunctiva or palms and in a sample of the remaining children. In both locations, 2% of children had severe anemia; 13% and 17%, respectively, in Uganda and Bangladesh had moderate anemia, and 42% and 62%, respectively, had mild anemia. Comparison of these findings with the clinical signs indicated that use of the full IMCI process would have referred 68-90% of children with severe anemia to the hospital. Severe palmar and conjunctival pallor, individually and together, had a sensitivity of 10-50% and a specificity of 99% for severe anemia; the addition of grunting increased the sensitivity to 37-80% while maintaining a reasonable positive predictive value. Palmar pallor was not as effective as conjunctival pallor for detecting anemia in Bangladesh. Combining conjunctival and palmar pallor detected 71-87% of cases of moderate anemia and half the cases of mild anemia. Anemia was more easily diagnosed in Uganda in children with malaria. These findings confirm the usefulness of clinical signs in the detection of anemia in developing countries, but suggest the feasibility of adding conjunctival pallor, and possibly grunting, to the IMCI algorithm.

Pallor as a clinical sign of severe anaemia in children: an investigation in the Gambia.

Anaemia associated with malaria is a major public health problem in African countries. Since most primary health facilities have to rely on physical signs and not laboratory tests to detect anaemic patients who need referral for blood transfusion, we have assessed the reliability of simple clinical signs to predict severe anaemia. A trained field assistant examined 368 children admitted to a tertiary care hospital, assessing the pallor of their eyelids (conjunctiva), palms and nailbeds, counting the respiratory rate, and looking for signs of respiratory distress. After the children's admission, their packed cell volume (PCV) was measured, and the need for transfusion and the outcomes were noted. A second observer examined 173 of these children so that interobserver variability in the detection of clinical signs could be assessed. A total of 27% of the 368 children had a PCV of < 15%. In a multiple regression analysis, definite pallor of the conjunctiva, definite pallor of the palms, and a "sick" appearance of the child were identified as independent significant predictors of a PCV of < 15%. The best predictor was a combination of definite pallor of the conjunctiva and pallor of the palms, with a sensitivity of 80% and a specificity of 85%. Inclusion of signs of respiratory distress did not improve the prediction. Pallor was a reproducible sign (weighted kappa statistic for the comparison between two observers: kappa = 0.6 for conjunctival pallor). We conclude that pallor can be used as a sign for referring children who may require blood transfusion.

PIP: Most primary health care facilities in Africa, must rely on physical signs, rather than expensive laboratory tests, to detect anemic patients in need of referral for blood transfusion. This study assessed the reliability of simple clinical signs to predict severe anemia in 368 children, admitted to a tertiary care hospital (Royal Victoria) in the Gambia, primarily with malaria. A trained field assistant examined each child, assessing pallor of the conjunctiva, palms, and nailbeds, on a scale of 0 (no pallor) to 3 (definite severe pallor), and looked for signs of respiratory distress. Examination of 173 of these children by a second observer confirmed inter-observer consistency in the detection of pallor. After the children's admission, their packed cell volume (PCV) was measured. 27% of the children had a PCV under 15%, indicative of severe anemia. Multiple regression analysis identified definite pallor of the conjunctiva, definite pallor of the palms, and a "sick" appearance as significant independent predictors of a PCV under 15%. The best predictor, a combination of definite pallor of the conjunctiva and definite pallor of the palms, had a positive predictive value of 67%, a sensitivity of 80%, and a specificity of 85%. Inclusion of signs of respiratory distress did not improve the prediction. These findings indicate that definite pallor of the conjunctiva, alone or in combination with definite pallor of the palms, can be used in primary health care centers to predict severe anemia with high precision.