The document presents a BSc thesis report on the design and development of a low-cost insulin pump prototype. It discusses: 1) The objectives of designing a prototype that achieves the accuracy and precision of commercial pumps at a lower cost for developing countries. 2) The literature review on insulin pump design and algorithms for predictive suspension to prevent hypoglycemia. 3) The materials and methodology used, including mathematical modeling of insulin dosing, 3D printing of parts, and software development. 4) The results of in vitro testing that show the prototype achieved accuracy and precision comparable to commercial pumps.

1. Robele Gemechu, Belayneh Mamush, Surafel Tehulu, Biruk Genene, Milky Guyassa
Advisor : Dr. Dawit A.
BSc. Thesis Report presentation
September 28, 2021

2. 2
1. INTRODUCTION
1.1. Background
1.2. Statement Of The Problem
1.3 Research Objective
1.3.1 General Objective
1.3.2 Specific Objective
1.4 Scope Of The Project
1.5 Significance Of The Project
2. LITERATURE REVIEW
3.MATERIALAND METHDOLOGY
3.1. Material

3. 3.2. Methodology
3.2.1 Mathematical Modeling Of Basal-bolus Insulin Dosing
3.2.2 Design Of The Low-cost Insulin Pump Prototype
3.2.3 Design And 3D Printing Of Parts Of The Low-cost Insulin Pump Prototype
3.2.4 Interface Of The Embedded Control System With 3D Printed Parts Of The Insulin Pump
3.2.5 Making A Case For The Prediction-based Attenuation Features
3.2.6 Kalman Filter Algorithm Modeling
3.2.7 Software Architecture
3.2.8 Android Development
3.2.10 In Vitro Performance Test
4. RESULT
5. DISCUSSION
6. CONCLUSION
7. RECOMENDATION
3

4. 1.1. Background
 Diabetes is a serious, chronic disease that occurs either because of insulin deficiency or
when the body cannot effectively use the insulin it produces.
 According to the World Health Organization (WHO), one in eleven people lives with
this disease, indicating 422 million people in the world.[1, 2, 3].
 Ethiopia recorded the highest numbers of people with diabetes in Africa with an
estimated 2.6 million diabetic patients [4].
4

5.  There are two types of diabetes: type 1 and type 2.
 Type 1 diabetes is an autoimmune disease .
 They can be treated by exogenous administration of insulin.
 Type 2 diabetes occurs as a result of your cells become resistant to the action of insulin.
 With type 2 diabetes, insulin is not always necessary.
5

6.  The characteristic of type 1 diabetic is a permanent picture of hyperglycemia,
and reliance on exogenous insulin for survival.
 insulin is daily administered by two methods which are conventional and intensive
insulin therapy.
 Intensive insulin treatment can be given by multiple daily injections or by
continuous subcutaneous insulin infusion.
6

7.  the Insulin doses must be finely tuned. Otherwise:
 Insulin under dosing can drive to hyperglycemia (BG > 180 mg/dl) and
over dosing of insulin can cause hypoglycemia (BG < 70 mg/dl)
 World spending with the disease reaches $327 billion and in 2018 only, around $51
billion was spent with the disease type 1 diabetes accounts for 5 to 10% of all diabetes
cases.
 insulin pumps have a high cost of acquisition for developing countries
like Ethiopia since all available models are imported.
7

8. 1.3.1 General Objective
Our main objective is to design and build a prototype of insulin
pump which can keep a blood glucose as close to normal as possible.
8

9. 1.3.2 Specific Objective
To attain the accuracy of insulin pump prototype within the range of
commercial ones,
To achieve the precision of the insulin pump within the range of
commercial ones,
To develop predictive based algorithm with automatic dosing of basal
delivery, and.
To build a low-cost insulin pump with commercially available hardware
setup.
9

10. Our project’s aim is to deliver a technology for insulin delivery
systems which helps as treatment for type 1 patient mainly.
10

11. we focused on building a system which comprised of an infusion pump
and user interface.
11

12. This study will benefit the low-income community who can not
purchase an insulin pump.
This technology has also the potential to decrease the burden of
diabetes management on the patients themselves.
In addition, this project will give additional insight for future
studies.
12

13. Yi Zhang et al. article introduced a generic insulin pump model and a
preliminary hazard analysis based on this model.
We used some of the system level safety issues
in our prototype design, to the extent our scope allowed us.
13

14. CONT….
 we choose to use a methodology developed by Coskun et al. [20] which
is based on the measurement of displacement syringe plunger (cm) over
time.
14

15. CONT…
Work by Veterotti et al. [21] reviewed the literature on methods for
CGM-based automatic attenuation or suspension of basal insulin with a
focus on algorithms, their implementation in commercial devices and
clinical evidence of their effectiveness and safety.
15

16. CONT…
An analysis of MiniMed 640G real-world data (data uploaded on
Carelink from January 2015 to January 2016), showed that
 “suspend before low” which uses prediction-based suspension was used on
83% of user days,
 “suspend on low” which uses threshold-based suspension was used on 11%
of user-days, while in the remaining
 6% of user-days, neither “suspend before low” or “suspend on low” were
activated.
16

17. Buckingham and coauthor [22] developed their prediction-based
suspension method using a single Kalman filter prediction
algorithm with a prediction horizon of 70 min.
this study involved the artificial induction of near-hypoglycemia by
increased basal insulin delivery, the algorithm with pump shutoff
prevented hypoglycemia 73% of the nights.
17

18. A work by Spaic et al. [27] goes further by using the prediction-based
suspension algorithm developed by Buckingham et al. [22] combined
with an automatic insulin-dosing component, forming the Predictive
Hyperglycemia and Hypoglycemia Minimization system for overnight
control which demonstrated increased time in range, lower mean
glucose level, and reduced hyperglycemia without increased
hypoglycemia compared with “suspend before low” features alone.
18
CONT….

19. 3.1. Materials
Syringe pump assembly hardware
 3D-printed Carcass
 3D-printed pusher block
 3D-printed Stepper motor connector
 Ball bearings
 Micromachined Spindle nut
 Micromachined Threaded spindle
 28BYJ-48 Stepper Motor
 Luecheck 3ml syringe
19

20. Electronic hardware
A4988 Stepper motor diver board
Microcontroller
MaxDay 9v battery
Rotary encoder
RTC
EEPROM
16*2 character LCD display module screen LCM HD44780
Button
Buzzer
Keypad
20

21. New wireless Bluetooth RF transceiver module serial RS232 HC-05
MAXDAY 9v battery
Insulated wires, USB cable and stationary material
Screw Driver
Soldering iron and flux core solder
Software resources
MATLAB
SOLIDWORKS
FRITZING
TINKERCARD
MIT APP INVENTOR
21

22.  Basal-bolus insulin dosing is widely used method of care for persons with
diabetes.
 The insulin pump usually feeds insulin to the body in two formats. The first one
is bolus dose and the second one is basal dose.
 bolus dose which pumped to cover food eaten or to correct a high blood glucose
level.
 basal dose which pumped continuously at an adjustable basal rate to deliver
insulin needed between meals and at night
22

23. 23
The insulin pump usually feeds insulin to the body in two formats,
known as bolus dose and basal dose.
23

24.  A. Calculation of total daily insulin requirement for 24 hours
 Method 1
Total Daily Dose (TDD) for insulin infusion = 0.75 X total daily insulin dose
prior to starting the insulin pump. Equation 3.1
 Method 2
Total Daily Dose (TDD) for insulin infusion = 0.5 X weight (kg) Equation 3.2
 Method 3
Total Daily Dose (TDD) for insulin infusion = (Method 1 +Method)/2
Equation 3.3
24

25.  B. Calculation of carbohydrate to insulin ratio
 The carbohydrate-to-insulin ratio (CIR) is the number of grams of
carbohydrate that are covered by 1 unit of insulin
CIR= 450 / TDD Equation 3.4
25

26.  C. Calculation of correction factor
 Correction factor (insulin sensitivity factor) is the amount of blood glucose is
lowered by the injection of 1 unit of insulin.
 Patients sensitivity for 1 unit of insulin.
 This is depend of the type of insulin they uses
 Insulin Sensitivity Factor = 1700 / TDD >>> for rapid acting insulin
Equation 3.5
 Insulin Sensitivity Factor = 1500 / TDD >>> for short acting insulin
Equation 3.6
26

27. If the post meal blood sugar is above the targeted blood sugar range
for 2 to 3 days then consider decreasing the CIR by 15 percent.
 If the post meal blood sugar is less than the targeted blood sugar
range for 2 to 3 days then consider increasing the CIR by 15
percent
27

28. D. Calculation of Correction Dose
 If the premeal blood sugar is out of the targeted range, the meal related insulin
dose may need to be adjusted accordingly
Correction dose = (Current blood sugar -Target blood sugar) / CF
28

29.  Basal insulin is the supply of insulin that is needed to maintain good blood sugar
control without taking into account eating any food.
 The basal insulin accounts for about 40 to 50% of the daily insulin requirement.
 Total Basal Insulin Requirement = 1/2 * Total Daily Dose (TDD) Equation 3.8
 Hourly basal rate = 1/24 * Total Basal Insulin Equation 3.9
29

30. Adjustment of Basal Rate
 The overnight basal rate is adjusted by checking the blood sugar at 12 AM, 3 AM and 7AM.
 If the glucose level rises more than 30 mg/dL between readings, the basal rate should be increased
by 15 percent.
 If the glucose level decreases by more than 30 mg/dL (or falls below target) between readings,
treat the low blood sugar and decrease the basal rate by 10 to 20 percent .
 To adjust other daytime basal rates the patient is instructed to not to eat between meals and not to
correct post-meal high blood sugars. The two-hour post-meal blood sugar is then compared to the
next pre-meal blood glucose.
 If the blood glucose decreases more than 60 mg /dL or falls below blood glucose target
decrease basal rate by 15 percent.
 If the blood glucose decreases less than 30 mg/dL or stays the same , or rises : increase the
basal rate by 15 percent [3].
30

31.  A ‘bolus dose’ is the term used for an additional insulin dose that can be given at
any time, usually to either match carbohydrate intake or to correct a high blood
glucose level.
 Dose of Insulin for meal = (CHO/ CIR) + (Gc -Gt) / CF Equation 3.10
 CHO (g) is the estimated amount of carbohydrates in the meal
 carbohydrate-to-insulin ratio (CIR)
 Gc is the current blood glucose level
 Gt is the target blood glucose level.
 correction factor (CF)
31

32.  The insulin pump prototype has a syringe infusion mechanism, whose flow
control is volumetric and micro-controlled by an electronic system.
 In the following step, the mechanical transmission converts the rotational
movement of the stepper motor into linear displacement of the syringe plunger
32

33. 33

34. 34

35. 35

36. Stepper motor
 We select Stepper motor because, accuracy and precision are our two main
objectives we need to fulfill at the end of the project.
 The stepper motor is a motor that move in slow precise and discrete steps.
 They excel other motor in application where precise positioning needed which is the
case in our project.
 The other reason it is easily accessible.
36

37. Syringe pump
 Low cost is a very important requirement in this project.
 During the insulin pump based T1D treatment, patients need to change insulin
reservoir every two days on average.
 The adopted syringe is the model Luecheck syringe 3mL.
 Such solution seems adequate because this syringe adopts the same diameter of
the standard syringe 10 mL, which keeps it short enough for pump’s case design
and demands low torque to push the plunger.
37

38. 38

39. 39

40. 40

41. 1) Main Module
2) Data logging and timer Module
3) Business Module
4) Bluetooth module
5) Motor Module
6) Display Module
7) Button control Module
8) Kalman module
41

42. 42

43.  We built an app for the user to have optional way to interface and access the
insulin pump.
 We have used an MIT app inventor for the development of app. App inventor
let you develop applications for android phones using a web browser and either
a connected phone or emulator.
43

44. We focused on the main hazard situation which was “insulin overdose”
that can lead the user to death.
The sources to identify the hazards involving insulin infusion pump
was Yi Zhang et al. literature.
44

45. 45

46. 46

47. 47

48.  Currently, the gold standard for insulin pump assessment precision is based on the
IEC 60601‐2‐2421 standard, which uses the so‐called time‐stamped micro
gravimetric method.
 For this analysis, we adapted the methodology proposed by Coskun et al.,7 and
evaluated the traveled distance (cm) of the syringe plunger, during infusion.
48

49. 49

50. 50

51.  𝐷𝑖 = (𝐷𝑖𝑚 ∗ 𝐷𝑟𝑚)/𝐷𝑟𝑝 Equation 3.26
 The experimental error was determined as well as the percentage of the samples
within ±5, ±10, ±15,
and ±20% deviation. The precision of the low-cost CSII prototype was statistically
analyzed by
one-way ANOVA, using MATLAB 2020a.
51

52.  To analyze the accuracy of the low-cost CSII prototype based on syringe plunger
displacement,we determined the target distance related to 5.0 IU infusion using a syringe
(Descarpack 3 mL) and a calibrated calipere determined the accuracy of data from a score
of samples within a precision deviation limit (± 5, ± 10, ± 15 and ±20%).
 Precision thresholds were defined as the percentage deviation from the target dose
volume.
52

53. 53

54. 3.2. Demo of Prototypes Menu System
54

55. 55
ONE WAY ANOVA analysis of method 2

56. 56
ONE WAY ANOVA analysis of method 1

57.  The measure Dt resulted in 1.03mm, and the measure Dm is the sample displacement
average, given by 1.0711mm for method 1 and 1.1056mm method 2. The error was
determined according to the equation below.
 ErrorSPD method 1 = 𝐷𝑚 − 𝐷𝑡 ∗ 100/𝐷𝑡 equation 4.27
= (1.1056-1.035) *100/1.035
=6.82%
 ErrorSPD method 2 = 𝐷𝑚 − 𝐷𝑡 ∗ 100/𝐷𝑡 equation 4.28
= (1.0711-1.035) *100/1.035
=3.49%
57

58. Infusion Model 1 Model 2 Injected volume
Model 1
Injected volume
Model 2
under delivery 3(12.5%) 9(37.5%) 4.5595 unit 2.7001 unit
over delivery 15(62.5%0) 15(62.5%) 5.4336 unit 9.3392 unit
Mean value 1.0711 1.1056
58
correspondent error regarding the displacement of the syringe plunger

59. 59
Scatter Diagram for method two

60. 60
Scatter Diagram for method one

61.  Previous studies have documented that the T1D treatment with insulin pump
reduces the HA1C, the number of hospitalizations, and the hypoglycemia events.
 As we stated on our objective in the development of prototype, we emphasized
on how low-cost insulin pump can be built starting from the selection of
materials in parallel with keeping the performance of the pump relative to
commercial ones
61

62.  It has been promoted as being a safety feature against hypoglycemia, especially
during sleep or in patients who have hypoglycemia unawareness, as less time is
spent in the hypoglycemic glucose range.
 In addition to this whenever there is imminent risk of hyperglycemia it notifies the
user to use the correction dose.
62

63.  Although the low-cost insulin pump prototype presented in this thesis is under
development, the evaluation of the device based on value is similar to those described in
the literature for commercial insulin infusion pumps, ranging from ± 2 to ± 5%
 It is worth mentioning that there are no mandatory accuracy requirements or acceptance
criteria for insulin pumps; however, under the assumption that a mean total deviation from
the target of 5% is acceptable.
63

64.  Jahn et al.,[55] Freckmann et al.,[57] and Borot et al.[58] evaluated
durable pumps and patch pumps using an adaptation of IEC 60601-2-24 protocol.
 According to the results of these studies,
 20.9, 46.6 and 77.8% of obtained values were within the ± 5%.
 39.5, 71.2 and 94.4% of obtained values were within the ± 10%.
 54.0, 81.2 and 94.4% of obtained values were within the ± 15%.

64

65. 65
Jahn et
al.,[55]
Freckmann et
al.,[57]
Borot et
al.[58]
Method 1 Method 2
± 5% 20.9% 46.6% 77.8% 0% 37.5%
± 10% 39.5% 71.2% 94.4% 25% 100%
± 15% 54.0% 81.2% 94.4% 47% 100%

66.  let’s look at a patient with a correction factor of 1:100, that is, all things being equal;
1 unit of insulin would lower their blood glucose concentration by 100 mg/dl.
 An injection 0.1units of insulin should lower the blood glucose by 10 mg/dl.
Continuing this line of reasoning, an injection of 0.05 units should lower the blood
glucose of this insulin sensitive patient by 5 mg/dl.
 Say the pump delivered 0.06 units instead. This would result in lowering their blood
glucose by 6mg/dl or an overall difference of 1 mg/dl.
 That’s why we made our focus on prototyping an accurate model for insulin pump.
66

67.  In conclusion, results show that the developed miniaturized mechanical system
presented functionality, precision, and accuracy when coupled to the electronic
system, and responded well to repeatability tests based on the results obtained
by using displacement of syringe plunger methods.
67

68.  The next version of the prototype can be developed by implementing safety
control for fluid injection, stepper motor, system pressure, cartridge volume,
occlusion flux, and others
 The other recommendation is our prototype does not have an algorithm which
can estimate the calorie of meals. But with further studies on machine learning
an algorithm can be developed.
68

69. 69