Novel Methods for Diagnosis of Pulmonary Microangiopathy in Diabetes Mellitus

Lung microangiopathy is a little known negative influence of diabetes mellitus on the functioning of the lungs. In current medical practice lung microangiopathy is diagnosed by comparing two measurements of lung diffusing capacity – one with the subject standing and one with the subject lying. The necessity to take two measurements is inconvenient. In lung microangiopathy we observe a reduction of diffusing capacity, lung flow and volume. Diabetes is a chronic illness that can lead to diabetic angiopathy, a pathology of the blood vessels (arteries, veins and capillaries). There are two types of diabetic angiopathy: macroangiopathy (disease of the larger blood vessels) and microangiopathy (microvascular pathology). The examples of angiopathy include: neuropathy, nephropathy and retinopathy. Current knowledge regarding diabetic lung microangiopathy is limited. Histopathological examination of lung biopsy samples is not a conclusive test of the consequences of diabetes (Dalquen, 1999). Animal experiments and post-mortem examinations have disclosed the influence of diabetes on the lung capillaries and alveolar-capillary membranes (Kida et al, 1993), (Kodolova et al, 1982) and (Popov& Simionescu, 1997). Histopathological tests have revealed the thickening of the alveolar and venous capillary walls (Matsubara& Hara, 1991) and (Weynand et al, 1999). Lung diffusing capacity measurements illustrate the state of alveolar-capillary barrier (Goldman, 2003). These are measurements of diffusion across the alveolar-capillary membrane. For diagnosing microangiopathy, the lung diffusing capacity is measured in two body positions: standing ( standing L D ) and lying on the back ( lying L D ), (Strojek et al, 1993) and (Kuziemski et al, 2008). On account of the human anatomic structure, diffusing capacity depends on the body’s position. For healthy subjects the diffusing capacity increases in the reclined position, lying standing L L D D  . The opposite is observed in the case of microangiopathic patients: the diffusing capacity decreases when the subject is lying lying standing L L D D  . This is the result of blood vessel damage and alveolar thickening caused by diabetes (Kuziemski et al, 2009). Only the fact that diffusing capacity increases or decreases in a given position is important as far as microangiopathy diagnosis is concerned. Spirometry (measurement of the volume and flow of inhaled and exhaled air) is the most common pulmonary function test. It is helpful in diagnosing asthma, pulmonary fibrosis, cystic fibrosis, and COPD (Chronic Obstructive Pulmonary Disease).


Introduction
Lung microangiopathy is a little known negative influence of diabetes mellitus on the functioning of the lungs. In current medical practice lung microangiopathy is diagnosed by comparing two measurements of lung diffusing capacity -one with the subject standing and one with the subject lying. The necessity to take two measurements is inconvenient. In lung microangiopathy we observe a reduction of diffusing capacity, lung flow and volume. Diabetes is a chronic illness that can lead to diabetic angiopathy, a pathology of the blood vessels (arteries, veins and capillaries). There are two types of diabetic angiopathy: macroangiopathy (disease of the larger blood vessels) and microangiopathy (microvascular pathology). The examples of angiopathy include: neuropathy, nephropathy and retinopathy. Current knowledge regarding diabetic lung microangiopathy is limited. Histopathological examination of lung biopsy samples is not a conclusive test of the consequences of diabetes (Dalquen, 1999). Animal experiments and post-mortem examinations have disclosed the influence of diabetes on the lung capillaries and alveolar-capillary membranes (Kida et al, 1993), (Kodolova et al, 1982) and (Popov& Simionescu, 1997). Histopathological tests have revealed the thickening of the alveolar and venous capillary walls (Matsubara& Hara, 1991) and (Weynand et al, 1999). Lung diffusing capacity measurements illustrate the state of alveolar-capillary barrier (Goldman, 2003). These are measurements of diffusion across the alveolar-capillary membrane. For diagnosing microangiopathy, the lung diffusing capacity is measured in two body positions: standing ( standing ) and lying on the back ( lying L D ), (Strojek et al, 1993) and (Kuziemski et al, 2008). On account of the human anatomic structure, diffusing capacity depends on the body's position. For healthy subjects the diffusing capacity increases in the reclined position, lying Research (Goldman, 2003), (Kaminski, 2004) has shown that a single spirometry test does not provide sufficient information for a diagnosis of lung flow limitation and lung volume reduction caused by diabetes mellitus. Instead we need long-term observations of spirometry results to diagnose lung microangiopathy. Microangiopathy lung impairment is characterised by a decrease in spirometric parameters FVC (the maximum volume of air that can be exhaled or inspired during a forced manoeuvre) and 1 FEV (the forced expiratory volume during the first second of expiration) (Davis et al, 2004), (Litonjua et al, 2005), (Yeh et al, 2008). A later study (Kuziemski et al, 2009) proved that a single spirometric test is insufficient to identify lung microangiopathy on account of functional reserve breathing. The reserves compensate, to some extent, the lung dysfunction, and thus the negative effect of diabetes is concealed until the effect reaches a higher level of impairment. Since the direct measurements of pulmonary function are not conclusive and post-mortem autopsies do not serve the given patient, the diagnostics have to be based on such indirect results as spirometric measurements and modelling. It is probably also capable of revealing lung microangiopathy if the results are very carefully analysed. Perfusion computed tomography (pCT) is a diagnostic method that enables the imaging of the organ and tissue. This is a powerful tool for diagnosing perfusion of internal organs such as the brain, liver, pancreas, spleen, kidneys and lungs (Alonzi & Hoskin, 2006), (Blomley et al, 1993), (Cao, 2011), (Eichinger et al, 2010), (Kuziemski et al, 2011). The method enables quantitative evaluation of circulation by determining changes in tissue during the flow of a contrast agent in the blood vessels. Changes, when compared to a normal tissue, can indicate tissue pathology. Lung pCT measurements can help in the diagnosis of cystic fibrosis (Eichinger et al, 2010) and diabetes (Kuziemski et al, 2011). It can also help to differentiate benign pulmonary nodules from lung cancer (Alonsi & Hoskin, 2006). The technique of lung pCT involves the intravenous injection of a non-iodinated contrast agent and the sequential scanning of the diagnosed region of the chest. There are a number of pulmonary perfusion parameters, calculated pixel by pixel on the basis of raw CT data, which are next analysed in order to reveal differences between the normal and altered tissue. The most useful perfusion parameters are: blood volume (BV), blood flow (BF), mean transit time (MTT), time to peak (TTP) and permeability surface (PS) (Kuziemski et al, 2011). The diagnosis of pulmonary function has to be based on indirect results, such as diffusing capacity, spirometry and pCT because direct assessment is difficult.

Materials
The tests were performed on a group of 72 diabetics. People with cardiovascular disease and smokers were not included in this group. First the diabetics were tested for microangiopathy, by comparing diffusing capacity in standing and lying position, and on this basis they were split into two groups: The pCT test was performed on a group of 18 subjects: 10 diabetics and 8 non-diabetic volunteers. The local perfusion parameters BF (blood flow), BV (blood volume), MTT (mean transit time) and PS (permeability surface) were obtained in selected regions of interest (ROI) in the artery and parenchyma. The patients classified as suffering from microangiopathy had breathing impairment symptoms cause only by diabetes. They were all non-smokers and had not been diagnosed with any other acute or chronic respiratory disease.

Diffusing capacity
The quality of gas exchange in the lungs depends on the diffusing capacity . During the measurement of L D (American Thoracic Society, 1995) a person takes a full inhalation of air mixed with small amounts of carbon monoxide and helium. The mixture is held in the lungs for a few seconds and then exhaled. The first part of the expired gas is discarded. The next portion, which includes gas from the alveoli, is collected. The L D is determined by analyzing and comparing the concentrations of carbon monoxide and helium in the samples of the inhaled gas and the exhaled gas. The alveolar volume A V is also determined in this test by using the single-breath helium dilution technique. Fig. 1 shows oxygen diffusion. Oxygen transportation from the alveoli to erythrocytes, through the alveoli-capillary barrier, is presented as the flow  (Dalquen, 1999) , West, 2008) and (Taylor et al, 1989  ] g s  depends on the organism's metabolic rate ( MR ), i.e. the organism's need for oxygen. In medical practice, blood saturation S gives the basic information concerning the quantity of oxygen transported from the lungs to all the other organs. The kinetics of oxygen association with haemoglobin is described in Hill's equation (Khee-Shing, 2007 and so the 0 H was rejected. These parameters enable a distinction to be made between patients with and without microangiopathy. In a standing body position ( 0.05 p  ) the parameters 1 p and 2 p are not useful for microangiopathy diagnosis.

www.intechopen.com
The results in Table 1 and Table 2 indicate the possibility of using a diagnostic test based on  a single measurement lying   L   D , and also on the basis of modelling results 1 p .and 2 p . To select the best potential classification algorithm, statistical measures, sensitivity and specificity (Panzer, Blach & Griner, 1991) were considered. Sensitivity Sens is the ability of a test to detect a disease when it is really present. Specificity Spec is the ability to confirm the absence of a disease in patients when it is really absent.

True Positive Sens
True The theoretically optimal prediction is: 100% Sens  ( all the sick patients were identified as sick) and 100% Spec  (none of the healthy patients was identified as sick).  Therefore the diagnostic procedure shown in Fig. 3 is recommend. H . For binary classification the more conclusive algorithms I) (with respect to Spec ) and II) (with respect to Sens ) are recommended.
The binary classification was performed for the 72 M  subjects. The results were compared with already known medical diagnoses of 44 microangiopathic patients and 18 patients with no microangiopathy. Then the sensitivity and the specificity were calculated according to equations (8) and (9), ( Table 3). The larger the Sens and Spec , the better. It seems reasonable to assume that results larger than 75% are satisfactory. Therefore an algorithm should be selected to fulfil this requirement, and not every entry in Table 3  H diagnostic thresholds to obtain a binary classification result. This procedure produces one of two possible results: 1) high probability of lung microangiopathy or 2) high probability of no lung microangiopathy. This can serve as useful confirmation in a doctor's diagnosis. The final decision is made by the doctor conducting the diagnosis, who can take into account this classification result along with other diagnostic data.

Spirometry
The most popular spirometry method is a dynamic one. The pneumotachometer (Miller, Harkinson&Brusasco, 2005)

defines the volume () []
Vt l and airflow 1 () [ ] Qt l s  during the inhalation and exhalation. The flow-volume curve () QV is received on the basis of the spirometry measurements of () Vt and () Qt . Specific ventilation manoeuvres are required in the spirometric test. The measurement is preceded by a period of quiet breathing-in and out (Fry, Hyatt, 1960). Next the maximal breath-in and the maximal forced breath-out manoeuvres are performed as the important parts of the spirometric test. The respiration parameters are defined on the basis of the volume-time curve () Vt and the flow-volume curves () QV (see Fig. 4 and Fig. 5), (Miller, Harkinson&Brusasco, 2005). In medical practice two terms, volume and capacity, are used. The difference between them is based on the assumption that volume refers to lung volumes, while capacity refers to different combinations of lung volumes, usually in relation to inhalation and exhalation. The relations between volumes and capacities are presented in equations (10).     (Miller, Harkinson & Brusasco, 2005) has published the spirometric norms and the reference values for these spirometric parameters. The reference values of the respiration parameters, obtained for a representative healthy population, are used for the interpreting the spirometry data and making the diagnosis. The reference values depend on the patient's anthropometric data, such as age, height and gender. In some medical cases (asthma, pulmonary fibrosis, cystic fibrosis and COPD) the spirometric parameter boundary values have already been defined. However, as far as lung microangiopathy is concerned, this has not yet been done. According to medical practice experience, lung microangiopathy is diagnosed on the basis of the following spirometric parameters:

MEF
, IC and 1 % FEV FVC . These spirometric parameters were collected from the 72 M  diabetic patients. However, a single spirometric test is not sufficient to diagnose the lung flow and volume reduction caused by microangiopathy; long term observation is necessary. Therefore one should consider an alternative method. By using the spirometric results in a statistical significance test, we can find the spirometric parameters that are most responsive to lung microangiopathy.

Spirometric parameters. Statistical comparison of results for microangiopathic and non-microangiopathic patients
Traditional spirometric parameters were tested for their ability to detect lung microangiopathy. The mean, standard deviation and the statistical significance level p for spirometric parameters are presented in Table 4  , which means that 0 H concerning mean parameter value equality is not rejected. The parameters do not allow for a distinction to be made between patients with and without microangiopathy.
The parameters do not show statistical significance in detecting microangiopathy: ex post significance level p is larger than ex ante significance level 0.05

 
, which confirms the null hypothesis concerning equality of mean values for both groups. Therefore, none of spirometric parameters are useful in diagnosing microangiopathy.

Spirometry modelling
The maximal breathing-in flow was modelled by means of the sine function. The forced breathing-out flow was finally mimicked using the gamma variate function (Askey R.A.& Roy R,2007) -after testing a large group of prospective regression functions, such as exponentials, polynomials, etc (Kalicka et al, 2007). The maximal inflow () in QV is modelled as follows:  Fig. 6 the () QV spirometric test results have been added, for comparison.

Statistical comparison of modelling results for microangiopathic and nonmicroangiopathic patients
The mean parameter value and standard deviation were calculated separately for microangiopathic and non-microangiopathic patients ( Table 5). The aim was to find out www.intechopen.com   subjects were now divided into two groups on the basis of these discrimination thresholds. Two algorithms, I and II (which gave the best results), were applied for binary classification. The algorithms are presented in Fig. 7. The binary classification decision was compared with previously known medical diagnoses, and then the Sens and Spec were calculated according to equations (8) and (9), (see Table 6). None of the tested algorithms could produce simultaneously a large Sens and a large Spec . Algorithm II gave the best 82% Sens  and algorithm I gave the best 62% Spec  . Algorithm II accurately detected the majority of the microangiopathy cases ( 82% Sens  ), and only missed out on a few. Yet 50% Spec  means algorithm II identified only 50% of the nonmicroangiopathic patients as non-microangiopathic and gave too many false alarms. This is an unacceptable result. Somewhat better 62% Spec  was obtained with algorithm I but this was still not satisfactory. The most likely explanation for these poor results is the fact that the thresholds K H and a H are too general and do not take into consideration important anthropometric features. Therefore the range of anthropometric features was made more specific ( Table 6.

T h i s m o r e a c c u r a t e c h o i c e o f t h e d i a g n o s t i c thresholds gave much better results. The
100% Sens  means that algorithm II identified all the sick patients as sick. This excellent result has been obtained for the sample 1 22 M  selected from the general population and therefore should be treated with care. An another sample may give a somewhat different result but the great improvement is obvious. The diagnostic procedure, utilising the algorithms I and II, is shown in Fig. 8.   Applied with a H and K H , algorithm I gives the best (most conclusive) Sens classification result, while the algorithm II gives the best (most conclusive) Spec result. This procedure gives one of two alternative results: 1) high probability of lung microangiopathy or 2) high probability of no lung microangiopathy. The final decision is made by the doctor.

Perfusion computed tomography (pCT)
Tests commonly used in clinical practice such as spirometry and lung diffusion capacity measurements are not considered precise enough to identify lung microangiopathy. There are very few publications in this field and most of them concern the decrease of diffusion www.intechopen.com capacity in diabetics (Villa et al, 2004), (Goldman, 2003). Autopsies of patients who died of diabetes complications revealed a thickening in the basement membrane of alveolar capillaries and arterioles in the lungs as well as of the entire walls of pulmonary arterioles together with fibroblast proliferation (Weynand et al, 1999). Impaired gas exchange in pulmonary alveoli results in the progressive reduction of reserves in small pulmonary vessels. For this reason lung perfusion in regions of lung microangiopathy may be diminished. Computed tomography of chest perfusion allows imaging of pulmonary vessels and parenchyma and may be useful in diagnosing pulmonary microangiopathy (Kuziemski et al, 2011).

Patients and study protocol
A group of 10 never-smoking diabetic adults and a control group of 8 non-diabetic volunteers were chosen. All participants had a similar body mass index and none of them suffered any disease that affected pulmonary function. The mean time from when diabetes mellitus was first diagnosed was 15.5 years. Nephropathy was diagnosed in 3 diabetics, retinopathy in 6 and polyneuropathy in 4. The question was whether or not they also suffered from lung microangiopathy. In the control group lung and cardio-vascular diseases were not diagnosed. For all patients, spirometry and diffusing capacity tests were performed. For the pCT tests a 64-row CT scanner GE -Light Speed VCT (GE Healthcare USA) was used. Pulmonary perfusion was evaluated after the intravenous administration of 40 ml of contrast at the rate of 4 ml/s. CT images were taken with 1 s resolution in the period of 40 s of a selected lung section (4 cm thick layer situated 2 cm below carina). Within this 4 cm thick segment, three cross-sections, anterior, medium and posterior, 5 mm thick each, were selected at uniform distances from each other. 6 ROIs were chosen in each of the crosssections (18 ROIs in all, numbered ROI 3-ROI 20). The ROIs 3-20 were situated in the upper, medial and lower parts of the left and right lungs. The most vascularised areas of the lungs were left beyond the ROIs so as not to distort the evaluation of lung parenchyma. The pCT measurements were performed using standard CT Perfusion 4 (GE Healthcare USA). The following perfusion parameters were calculated: - [ 100 ] BV ml g -blood volume in 100g of lung tissue.
-[ / 100 / min] BF ml g -blood flow through 100g of lung tissue in 1 minute. - [min] MTT -mean transit time through the vascular system in selected region. -[ / 100 / min] PS ml g -permeability surface, the permeability of blood from intravascular to extravascular space, observed in 100g of lung tissue during 1 min.

pCT results
The example results of pCT scans, obtained from a 62 year old diabetic male, are shown in Fig. 9 a), b), c) and d). Perfusion parameters BF , BV , MTT and PS were calculated and mapped out on the lung cross-section. Analysis was made of the pulmonary artery -ROI 1, which represents the arterial input function, i.e. the flow blood into the pulmonary system. In Fig. 9 the ROI 1 is marked with a circle. The calculated values of perfusion parameters (mean value for all pixels in the region and standard deviation of the mean) are displayed in the right lower corner of the scan. Table 7 shows mean perfusion parameters Avg and standard deviation Dev in the ROI 1. for the diabetic subject. The measurements were obtained in the same location of the lung cross-section. Fig. 10 shows PS maps obtained from a healthy patient and a diabetic patient. The qualitative comparison (amount of the red, permeable area) as well as the quantitative comparison (healthy patient 3.31 PS  and diabetic patient 79.80 PS  ) indicate damage to capillaries and pulmonary arterioles of the diabetic patient. All perfusion parameters were calculated in 18 ROIs (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20) in the parenchyma to discover whether parameter values depend on the location of tested lung cross-section. Statistical analysis revealed no correlation between perfusion parameter value and ROI location in the upper central or lower part of the lungs (Kuziemski et al, 2011). Therefore mean perfusion www.intechopen.com parameters were calculated for all 18 ROIs. The results are shown in Table 8 (Kuziemski et al, 2011). Here a noticeable increase was observed in the parenchyma perfusion parameter values of the diabetics with respect to control group, which may be due to diabetic mellitus.  Table 8. Mean parameters BF , BV and PS in the artery and parenchyma of the diabetics and of the control group. The calculated standard deviation Dev illustrates the distribution of test results in 18 ROIs (ROI 3 -20, in upper, central and lower part of lungs) with respect to the mean of the eighteen results. No significant MTT differences were found.
Chest perfusion CT allows imaging of lung vessels and parenchyma, which is essential in diagnosing pulmonary angiopathy. The analyzed results are very new and currently not all that numerous. While this does not allow for a complex statistical analysis, the above preliminary results indicate the great potential of pCT tests in diagnosing lung microangiopathy.

Lung area extraction
Each pCT image (see Fig. 9) consists of the body cross-section, including the lungs, essential for medical diagnosis, the rest of the body and the background, which is unnecessary for lung diagnostics but was nevertheless included in the perfusion map calculations. As computations of perfusion parameters involve complicated methods of signal processing, it is desirable that the distinction between diagnostically important regions (the lung crosssection) and other areas (background and the rest of body cross-section) is conducted before the process starts. This shortens the time needed to obtain parametric maps by omitting data processing from the area outside the lungs. It can also help avoiding false lung malfunction diagnoses, which can occur when the automatic methods of pathological changes wrongly interpret signals from areas outside the lungs as abnormal lung signals. Literature describes many methods of lung area extraction from CT chest scans (Homma et al, 2009), (de Nunzio et al, 2011, (Vinhais & Campilho, 2006), (Zhou et al, 2004). Most methods are devoted to extracting the lung area from a single image. Some of these methods allow for neighbouring (upper and lower) slices to be analysed (Homma et al, 2009). The lung extraction presented here was obtained using the two-step procedure. The objective of the preliminary processing was to define the rough mask, i.e. the border within which the lung contour is always contained during the pCT test while the lungs are breathing and the heart is beating. The lung and heart region contour pixels are subjected to the greatest changes. This behaviour can be utilised to detect the contours. It can be achieved by using a variability factor VF , which is defined as follows: Then the final lung mask was obtained with the use of morphological operators (dilation, erosion, closing and opening) to extract lung cross-sections from the chest pCT. Fig. 12 shows the BF , BV , MTT and PS maps of just the lungs, extracted using the mask. The lung mask was superimposed onto the calculated maps in Fig. 9. Visual comparison of Fig. 9 and Fig. 12 shows just how much unnecessary computations are usually performed in areas outside the lungs. What is more, a clearly defined lung area allows for a much more accurate and reliable medical diagnosis.

Conclusion
Pulmonary diabetic microangiopathy has not yet been sufficiently studied and is difficult to diagnose. In clinical practice, lung microangiopathy is diagnosed by comparing two diffusing capacity measurements, taken with the patient standing up and lying down. Other methods to assist diagnosis of lung microangiopathy are needed.
The three methods presented here seem good enough to be competitive options to complement those currently used in medicine today. Diffusing capacity measurement and modelling allows for the development of a procedure to enhance the diagnostic process. This procedure is based on a single diffusing capacity measurement taken from a patient lying down, together with a model of oxygen diffusion from the alveoli to the blood. The model parameters, calculated using routine medical test results and physiological constants, turned out to be useful for binary classification in lung microangiopathy diagnosis. Diagnostic thresholds were calculated. Next binary classification was performed, based on model parameters, and, for comparison, on the diffusing capacity test result. These results were compared with already known medical diagnoses. This allowed for the sensitivity and specificity to be calculated. The best binary classification results were 79% Sens  and 87% Spec  , when based on model parameters, and, when based on diffusing capacity measurements, 62% Sens  and 71% Spec  . These results show that for microangiopathy diagnosis the model parameters were more sensitive and more specific than the diffusing capacity measurements. The next method to be examined for usefulness in lung microangiopathy diagnosis was spirometry. Statistical comparison of the spirometry parameters of microangiopathic and non-microangiopathic patients, revealed that such parameters do not allow for a distinction to be made between patients with and without microangiopathy. Therefore a spirometry model was developed and its parameters were tested for their ability to distinguish between microangiopathic and non-microangiopathic subjects. Two, out of five, model parameters were statistically significant and helpful in lung microangiopathy diagnosis. Diagnostic thresholds were calculated and binary classification was performed, based on the statistically significant model parameters. The best results were 100% Sens  and 83% Spec  . Both diagnostic procedures, the one based on the diffusing capacity and the one based on spirometric measurements and modelling, produce one of two possible results: 1) high probability of lung microangiopathy or 2) high probability of no lung microangiopathy. Perfusion CT measurement seems to have considerable diagnostic potential with regard to lung microangiopathy. A study of lung tissue perfusion in diabetic patients in comparison with a group of healthy subjects showed an increase in the value of some parameters. These increased values may result from pulmonary microangiopathy. Further studies on the application of pCT in the detection of diabetes mellitus may not only improve diagnosis but also help us better understand pulmonary microangiopathy. The developments in molecular medicine are transforming respiratory medicine. Leading clinicians and scientists in the world have brought their knowledge and experience in their contributions to this book. Clinicians and researchers will learn about the most recent advances in a variety of lung diseases that will better enable them to understand respiratory disorders. This treatise presents state of the art essays on airways disease, neoplastic diseases, and pediatric respiratory conditions. Additionally, aspects of immune regulation, respiratory infections, acute lung injury/ARDS, pulmonary edema, functional evaluation in respiratory disorders, and a variety of other conditions are also discussed. The book will be invaluable to clinicians who keep up with the current concepts, improve their diagnostic skills, and understand potential new therapeutic applications in lung diseases, while scientists can contemplate a plethora of new research avenues for exploration.