PROCEEDING Of The 2017 International Multidisciplinary - UKRIDA

CONFERENCE PROCEEDING2017 International Multidisciplinary Conferences onProductivity and SustainabilityJakarta, 5-7 December 2017Staining Adjustment of Dye Amount to Clarify the Appearance of Fiber, Nuclei,and Cytoplasm in HE-stained Pathological Kidney Tissue ImageLina Septiana1*, Hiroyuki Suzuki2, Masahiro Ishikawa3, Takashi Obi2,Naoki Kobayashi3, Nagaaki Ohyama21Schoolof Engineering, Tokyo Institute of Technology, Tokyo, Japanof Innovative Research, Tokyo Institute of Technology, Tokyo, Japan3Faculty of Health and Medical Care, Saitama Medical University, Saitama, Japan*Corresponding author email: lina.septiana@gmail.com; septiana.l.aa@m.titech.ac.jp2InstitutePathological images are used to examine the microscopic patterns of tissue. However, the color of pathologicalimages may have variation depend on the staining laboratory facility or the pathologist. The color variationbecomes a problem specifically in the telepathology which will integrate some of pathological imaging systems.To overcome this problem, this study focused on a method of color adjustment in HE (Hematoxylin and Eosin)stained slides with case study on kidney tissue, and compared the spectral differences of tissue component i.e.fibrotic, nucleus, and cytoplasm structures. This work was applied by using a 61-bands microscopic multispectralcamera. The statistical analysis was calculated to get the characteristic of these three tissue components from dyeamount images.Keywords: multispectral image, pathology, dye amount, staining adjustment1. IntroductionIn pathological diagnosis, tissue staining becomes one of the necessary processes. One of the mostimportant staining methods is Hematoxylin and Eosin (H & E) stain. However, the chemical reaction of theHE stain is very sensitive depending on the dyeing time, temperature and pH of the solution, and also thecharacteristic of microscope and camera. Therefore, it is difficult to maintain practical staining undervarious conditions in a standardized one, and this variety results the pathological imaging color variation.In telepathology, this color variation becomes a problem, because some pathological images generated byvarious facilities should be examined. To overcome the above problem, Abe et al1 had proposed colorcorrection method by adjusting dye amount of Hematoxylin and Eosin in the pathological liver image. Inthis paper, we investigate the feasibility of applying the color control method based on dye amountadjustment into kidney images, and observe the statistic characteristics of fiber, nuclei, and cytoplasm ofthe dye amount image.2. MethodThe HE slide is stained by several dyes and processed by some of the staining mechanisms. Hence, thecolor adjustment was performed corresponding to each staining procedure separately. The correlationbetween a multispectral image pixel value and the spectral transmittance of a stained slide can be expressedas follows1:(1)where Sk(λ) is the camera sensitivity spectral product, and the optical filter transmittance at kth band. E(λ)represents the light source radiance, nk is the additive noise. t(λ) is spectral transmittance of a pixel in theslide which be estimated by using the Wiener technique.Page 1

Beer-Lambert law was applied to show the relation between the spectral transmittance and the dyeamount of a stained slide. It can be seen as follows1:(2)Where 𝑡(𝜆)is a spectral transmittance, 𝑡𝑔 (𝜆) is the glass slide spectral transmittance, Ɛ𝑖 (𝜆) is the spectralabsorption coefficient, 𝐶𝑖 is the amount of dye, with i 1,2,3 respectively for haematoxylin, eosin, red bloodcell.Since currently the HE staining standard has not available yet, a considered ideal HE staining samplewas used as reference image. The spectral absorption coefficient Ɛ𝑖 (𝜆) was measured by three pure stainedsamples respectively for haematoxylin, eosin, and unstained image which shows only erythrocyte pattern.The referenced image (𝐶𝑖 𝑟𝑒𝑓 ) was applied to equation (2) to get the referenced of the multispectral signal.Usually 𝐶𝑖 𝑟𝑒𝑓 depends on tissue structure such as nuclei, fibrosis, and cytoplasm, etc.(3)In order to estimate the weighting coefficient(𝜔𝑖 ), equation (4) was used.(4)By using this estimated weight 𝜔𝑖 , the corrected spectral transmittance 𝑡 ′ (𝜆) is calculated by equation (5).(5)3. ExperimentsHuman kidney tissue HE-stained specimens from BioChain Institute was used in this experiment, whichcollection and handling process conformed to Good Clinical Practice (GCP) regulations with theInstitutional Review Board (IRB) process and informed patient consent forms being required for eachsample. The spectral features were extracted from images captured by a hyperspectral camera (EBA Japan,NH-3) and Olympus BX-53. The HSIs images are converted to RGB images. The size of these images is752 x 480 pixels, the wavelength range is 420 – 720 nm, and 61 bands of 5 nm interval. Light source useLED lamp. Statistical analysis is done by using Matlab R2016b.Figure 1. Flow of the Staining Adjustment MethodFigure 1 is the flow of the staining adjustment method in this experiment. First, dye amount image (𝐶𝑖 )was estimated from the measured multi spectral image and spectral absorption coefficient using equation(2). Second, we calculated a referenced multispectral signal 𝑙𝑜𝑔{𝑡(𝜆)/𝑡𝑔 (𝜆)} based on three of pure𝑟𝑒𝑓Page 2

dye amount images 𝐶𝑖 𝑟𝑒𝑓 using equation (3). Third, we calculated the weighting value (𝜔𝑖 ) based on thisreferenced multispectral signal using equation (4). Fourth, the corrected multispectral signal 𝑙𝑜𝑔{𝑡′(𝜆)/𝑡𝑔 (𝜆)} was determined by using equation (5).The absorption coefficient Ɛ𝑖 (𝜆) of H, E, and red blood cell, could be seen in the figure 2, blue, pink, andred line respectively represent hematoxylin, eosin, and red blood cell. Figure 3 shows the controlled RGBHE kidney images by adjusting the weighting values, those obtained by simulating the changing of weightfactor of the hematoxylin and eosin components. The increasing of weight coefficient for Hematoxylinmakes the blue color more dominant, and the increasing of weight coefficient eosin make pink color moredominant. Figure 4a, 4b respectively are an ideal and a corrected RGB (red, green, blue) HE kidney.Figure 2. Spectral absorption coefficient(a)(b)Figure 3. RGB images reproduced by simulating theweighted dye amount of Hematoxylin and EosinFigure 4.a.Referenced imageFigure 4.b.Corrected image.(a)(b)(c)Figure 5. Sampling point of fiber in absorbance image (a)hematoxylin, (b)eosin, (c)red-blood-cell(a)(b)(c)Figure 6. Sampling point of nuclei in absorbance image (a)hematoxylin,(b)eosin, (c)red-blood-cellPage 3

(a)(b)(c)Figure 7. Sampling point of cytoplasm in absorbance )(c)(d)Figure 8. Dye amount intensity: (a) fiber, (b) nuclei, (c) cytoplasm, (d) all regionsWe investigated the distribution of pixel intensity values of dye amount from the selected candidatepoint of fiber, nuclei, cytoplasm, and all regions. Figure 8a, 8b, 8c, 8d respectively show the dye amountvalues of fiber, nucleus, cytoplasm, and all regions. The statistical calculation from these four data can beseen in table 1 for fibers, table 2 for the nucleus, table 3 for cytoplasm, and table 4 for all regions.Table 1. Statistical result of fiberTable 3. Statistical result of cytoplasmTable 2. Statistical result of nucleiTable 4. Statistical result of all regionsBased on these statistical data, the mean of fibers for hematoxylin was 0.2, 0.9 for eosin, and 0.25 forRBC. The mean of the nucleus was in 1.09 for hematoxylin, 1.43 for eosin, and 0.25 for RBC. The meanPage 4

of cytoplasm was in 0.3 for hematoxylin, in 1.24 for eosin, and 0.28 in RBC. The mean for all regions was0.4 for hematoxylin, 0.97 for the nucleus, and 0.26 for RBC.4. ConclusionThis work has shown the characteristic of the candidate of fiber, nuclei and cytoplasm in HE stained kidneyimage based on dye amount image intensity. This study could be used to clarify the appearance of fiber,nuclei, and cytoplasm in the HE stained image. In the future work, this work will be confirmed bypathologist, and will be applied to other organs for further classifications.AcknowledgementThe authors appreciate to Indonesia Endowment Fund for Education and Japan Society for the Promotionof Science, for providing financial support.References1. T. Abe, Y. Murakami, M. Yamaguchi, N. Ohyama, Y. Yagi, Color Correction of Pathological ImagesBased on Dye Amount Quantification, OPTICAL REVIEW Vol. 12, No. 4 (2005).2. N. Hashimoto, Y. Murakami, P. A. Bautista, M. Yamaguchi, T. Obi, N. Ohyama, K. Uto, Y. Kosugi,Multispectral Image Enhancement for Effective Visualization, OPTICS EXPRESS, Vol. 19, No. 10 / 9May 2011.3. E. Hashimoto, M. Ishikawa, K. Shinoda, M. Hasegawa, H. Komagata, N. Kobayashi, N. Mochidomo,Y. Oda, C. Iwamoto, K. Ohuchida, M. Hashizume, Tissue Classification of liver Pathological TissueSpeciment Image using Spectral Features, Proc. Of SPIE Vol.10140 (2017).Page 5

CONFERENCE PROCEEDING2017 International Multidisciplinary Conferences onProductivity and SustainabilityJakarta, 5-7 December 2017Cascade PID-FUZZY Controller for Temperature System withLong Dead-TimeRendra Dwi Firmansyah1, Oyas Wahyunggoro2, Adha Imam Cahyadi2, Rella Mareta11Departmentof Electrical Engineering and Informatics, Vocational College, Universitas Gadjah Mada, Yogyakarta, Indonesiaof Electrical Engineering and Information Technology, Universitas Gadjah Mada, Yogyakarta, Indonesia*Corresponding author email: firmansyah 1412@ugm.ac.id2DepartmentIn industrial practice, there are many production processes involving the temperature system. The product qualitydepends on how good the temperature system is controlled. The temperature system is one of challenging topics incontrol system because of its long dead time. The dead-time makes the control of temperature system more difficultsince it produces additional lag in the phase of system which decreasing gain margin and phase of the transfer function.Dead-time is caused by the time needed to transfer heat from heart source to the sensor because sensor position is farfrom heat source. To control the temperature system with long dead-time, a cascade PID-Fuzzy control is proposed.The proposed control system combines the Zieglers Nichols PID with 3x3x3 fuzzy control system. The proposedcontrol system gives fast rise time and little dead time around 6 seconds without overshoot.Keywords: temperature system, PID, Ziegler Nichols, fuzzy controller1. IntroductionTemperature control system is very important for production process in industry because it affects theproduct quality (1) . The stability of temperature on combustion process must be maintained to producegood quality product (2). Temperature in combustion process is very high up to thousand degree Celcius.Temperature sensor (3) cannot be placed on the heat source in combustion process because it will melt. Thesolution of this problem is placing temperature sensor far from the heat source. But, this solution causes atransport delay (4) or dead-time (5). Because the temperature sensor cannot be used directly at heat source,it waits until the heat is transported from the source to the sensor. The transport delay causes several effectsas mentioned in (6). The value of transport delay depends on distance of sensor and heat source. It alsodepends on the air velocity (7) to transport heat from source to sensor.The transport delay needs to be minimized by using control system. The common control systemwhich is used on temperature system are PID and Fuzzy Logic Controller (8). PID Controller usually usesbasic experimental Ziegler-Nichols tuning rules (9). This conventional PID controller is the most used inindustries to control non-linear processes like temperature. (10). Tuning method (11, 12) is developed toovercome weakness of previous conventional PID controller. Fuzzy control system is also developed toovercome transport delay problem (13, 14). Both fuzzy control system and PID controller have their ownweaknesses and advantage. To get best temperature control system, a hybrid controller is developed recentlysuch as Smith-fuzzy control system (15), Feedforward DMC-PID (16), Switch PID-Fuzzy (1, 17). Thispaper propose a control system that can minimize transport delay. A proposed control system is cascadePID-Fuzzy controller.2. Experimental DetailsThe experiment is conducted by using temperature system plant inspired from (18). The plant is made frompipe with Diameter 5 cm, Length 100 cm, Heater 100W, Fan, temperature sensor LM35. There are twosensors used in this plant. The first sensor, which is placed near the heat source, is called monitor sensor.The second sensor, which is placed far from heat source, is called feedback sensor. The distance betweenPage 6

the feedback sensor and the heat source is 95cm. Monitor sensor measures temperature around the heater.Feedback sensor is used to measure temperature of area that is far from heater, which is sent to controller.This plant uses DAQ 6008 to collect temperature data and Labview to control the temperature system. Theheat from heater is controlled by using PWM signal from DAQ 6008. The fan is used to transport heat fromsource to feedback sensor. The plant is showed by Error! Reference source not found.This experiment combines Ziegler Nichols PID with 3x3x3 fuzzy controller. This controller is madeby using block programing from Labview. There are two main blocks in the program, PID block and fuzzyblock. The block diagram of this controller is showed by figure 2 .The Ziegler Nichols method 2 is used totuning PID block. Tuning PID Ziegler Nicols Performed by searching for Kp constants in critical statescalled KCr. KCr is the value of the Kp constant which results in the system response being critical. A criticalstate is a state in which the system responds constantly to oscillation. The value of KCr is obtained byincreasing the value of Kp bit by bit until it gets critical condition. In this critical state also obtained acritical period called PCR that is used to calculate the constant Ki and Kd Constant. PCr is the critical periodobtained by calculating the distance between the oscillation waves of the system in a critical state.The resultof this tuning is showed by table 1.Table 1 Zigler Nicols Second Method 6Kcr0,5Pcr0,125PcrThe fuzzy block has two inputs and one output. The inputs are error and rate error. The output is PWMsignal. Fuzzy controller used in this experiment is 3x3x3. The membership of the fuzzy controller is showedby figure 2. The rule of the fuzzy controller is showed by table 2.e/ eNZPTable 2 Rules of Fuzzy ControllerNZPMSSMMLLLMPID and fuzzy controller are combined by summing the gain of each controller. The experiment wasconducted by setting the set point and then tuning on the plant. The GUI will show the result of controller.The controller will keep the temperature on the setting point.3. Results and DiscussionThis Experiment start with conventional PID controller with Zigler Nicols Tuning Second Method. In thiscontrol system gain constant we used are Kp 90, Ti 5,5 and Td 1,375. The result of this control system isshowed in figure 4. From this result we can see transport delay of the system is 3 second. There is noovershoot in this system but there is oscillation. This oscillation shows that the system is unstable. TheITAE of this control System is 6.346,1.Experiment of proposed controller was conducted with two tuning. The first one used Ziegler Nichols,the second one used trial error based on Ziegler Nichols result. The result of proposed controller is showedby figure 5a. Figure 5a shows the result of Ziegler Nichols PID and Fuzzy Controller experiment with setpoint 50 C. In this experiment the transport delay or Dead-Time is 6 seconds. This controller gives fast risetime without overshoot but it has oscillation around the set point. The rise time of this controller is 12seconds. The ITAE of this controller is 6.978,5.The figure 5b shows the result of Trial Error PID and fuzzy Controller experiment with set point 50 C.In this experiment the dead-time is 5 seconds. This controller has slower rise time but minimum oscillation.Page 7

The rise time of this system is 18 seconds. There is no overshoot either in this controller. The ITAE of thiscontroller is 6.171,9.If we compare the result of Conventional PID controller with proposed controller, we can see thetransport delay of conventional PID controller is better than proposed controller. But if we compare theoscillation that is occurred in the system, proposed controller is better than conventional PID controller.The ITAE of proposed controller also smaller than conventional PID controller so it is mean the proposedcontroller better than conventional PID controller in handling error. So overall the proposed controller givesbetter result in controlling temperature system with transport delay.4. ConclusionThe proposed controller works well on temperature system plant with long dead-time. It can remove theovershoot which is usually produced by PID controller. Also, this proposed system can minimize the risetime of temperature control system which is the weakness of fuzzy controller. So, this control system canimprove performance of previous control system for on temperature system plant with long dead-time. Thiscontrol system is gives better result than conventional PID control System.References1. Cao W, Meng Q, editors. Based on PLC temperature PID - fuzzy control system design and simulation.Information Networking and Automation (ICINA), 2010 International Conference on; 2010 18-19 Oct.2010.2. Xue-Ling S, Chao-Ying L, Zhe-Ying S, Xue-Fen S, editors. Robust PID control for steam superheater.Machine Learning and Cybernetics, 2004 Proceedings of 2004 International Conference on; 2004 2629 Aug. 2004.3. Purcaru D, Purcaru A, Rădulescu V, editors. Temperature measurement and control system forengineering education. 2017 18th International Carpathian Control Conference (ICCC); 2017 28-31May 2017.4. Hearns G, Grimble MJ, editors. Temperature control in transport delay systems. Proceedings of the2010 American Control Conference; 2010 June 30 2010-July 2 2010.5. Ramanathan P, Sukanya KC, Mishra S, Ramasamy S, editors. Study on Fuzzy Logic and PID Controllerfor temperature regulation of a system with time delay. Energy Efficient Technologies for Sustainability(ICEETS), 2013 International Conference on; 2013 10-12 April 2013.6. Lima DM, Santos TLM, Normey-Rico JE. Robust nonlinear predictor for dead-time systems with inputnonlinearities. Journal of Process Control. 2015;27:1-14.7. Jo J, Kim SJ, editors. Effect of velocity non-uniformity on heat transfer in oscillating flow. 2017 16thIEEE Intersociety Conference on Thermal and Thermomechanical Phenomena in Electronic Systems(ITherm); 2017 May 30 2017-June 2 2017.8. Elnour M, Taha WIM, editors. PID and fuzzy logic in temperature control system. 2013 InternationalConference on Computing, Electrical and Electronic Engineering (ICCEEE); 2013 26-28 Aug. 2013.9. Ikhlef A, Kihel M, Boukhezzar B, Guerrouj A, Mansouri N, editors. Online temperature control system.2014 International Conference on Interactive Mobile Communication Technologies and Learning(IMCL2014); 2014 13-14 Nov. 2014.10. Mugisha JC, Munyazikwiye B, Karími HR, editors. Design of temperature control system usingconventional PID and Intelligent Fuzzy Logic controller. 2015 International Conference on FuzzyTheory and Its Applications (iFUZZY); 2015 18-20 Nov. 2015.11. Deshmukh GL, Kadu CB, editors. Design of two degree of freedom PID controller for temperaturecontrol system. 2016 International Conference on Automatic Control and Dynamic OptimizationTechniques (ICACDOT); 2016 9-10 Sept. 2016.12. Nikita S, Chidambaram M. Tuning of PID Controllers for time Delay Unstable Systems with TwoUnstable Poles. IFAC-PapersOnLine. 2016;49(1):801-6.13. Wang H, editor Main Ste

Dr. Didik Wahjudi (Indonesia) Dr. Oki Sunardi (Indonesia) Dr. Ishak Ramli (Indonesia) Dr. Moeljono Widjaja (Indonesia) Dr. Iwan Aang Soenandi (Indonesia) Ignasia Yuyun, M.Pd. (Indonesia) Dr. Evans Garey (Indonesia) Yuseva Ariyani Iswandari, M.Pd. (Indonesia) PUBLISHED FIRST TIME BY: UKRIDA PRESS