An Undertaken Report For Heart Disease Prediction And Identification Using Machine Learning Methods
S.Ramkumar, M Sathyashree,
School Of Computing, Department Of Computer Applications,
Kalasalingam Academy Of Research And Education, Krishnankoil, Sirivilliputhur [email protected], [email protected]
Abstract: -
According To Recent Survey By Who (World Health Organization) 17.9 Million People Die Each Year Because Of Heart Related Diseases And It Is Increasing Rapidly. With The Increasing Population And Diseases, It Is Become A Challenge To Diagnosing Disease And Providing The Appropriate Treatment At The Right Time. But There Is A Light Hope That Recent Advances In Technology Have Accelerated The Public Health Sector By Developing Advanced Functional Bio Medical Solutions. This Aims At Analyzing The Various Data Mining Techniques Namely Naive Bayes, Random Forest Classification, Decision Tree And Support Vector Machine By Using A Qualified Dataset For Heart Diseases Prediction Which Is Consist Of Various Attributes Like Gender, Age, Chest Pain Type, Blood Pressure, Blood Sugar Etc. The Research Includes Finding The Correlations Between The Various Attributes Of The Dataset By Utilizing The Standard Data Mining Techniques And Hence Using The Attributes Suitably To Predict The Chances Of A Heart Disease. These Machine Learning Techniques Take Less Time For The Prediction Of The Disease With More Accuracy Which Will Reduce The Dispose Of Valuable Lives All Over The World.
Keywords; Naive Bayes, Support Vector Machine, Random Forest Classification, Decision Tree.
I. Introduction
Health Is One Among The World Challenges For Humanity. World Health Organization (Who) Has Mentioned That For An Individual Proper Health Is The Fundamental Right. So To Keep People Fit And Healthy Proper Health Care Services Should Be Provided. 31 Percentage Of All Deaths World Wide Are Because Of Heart Related Problems. Diagnosis And Treatment Of Heart Disease Is Very Complex, Particularly In Developing Countries, Due To The Lack Of Diagnostic Devices And A Shortage Of Physicians And Other Resources Affecting Proper Prediction And Treatment Of Cardiac Patients. With This Concern In The Recent Times Computer Technology And Machine Learning Techniques Are Being Used To Develop Software To Assist Doctors In Making Decision Of Heart Disease In The Preliminary Stage.
Early Stage Detection Of The Disease And Predicting The Probability Of A Person To Be At Risk Of Heart Disease Can Reduce The Death Rate. Medical Data Mining Techniques Are Used In Medical Data To Extract Meaningful Patterns And Knowledge. Medical Information Has Redundancy, Multi-Attribution, Incompleteness And A Close Relationship With Time. The Problem Of Using The Massive Volumes Of Data Effectively Becomes A Major Problem For The Health Sector. Data Mining Provides The Methodology And Technology To Convert These Data Mounds Into Useful Decision Making Information. This Prediction System For Heart Diseases Would Facilitate Cardiologists In Taking Quicker Decisions So That More Patients Can Receive Treatments Within A Shorter Period Of Time, Resulting In Saving Millions Of Life.
1.1 Methods
Data Mining Provides The Methodology And Technology To Convert Data Mounds Into Useful Decision-Making Information. In This Research The Comparison Of Different Machine Learning Techniques Like- Support Vector Machine, Decision Tree, Random Forest, Naive Bayes Are Implemented To Predict Heart Disease. Naïve Mathematician Used Probability For Predicating Heart Disease, Svm Used On Classification And Regression Technique, Random Forest Works With Varied Decision Tree. These Algorithms Show Different Accuracy. We Will
Try To Tuning Our Techniques To Obtain Better Accuracy Which Will Be Beneficial For More Accurate Prediction.
1.2 Objectives
The Main Objective Of This Study Is To Predict Whether A Patient Is Affected With Heart Disease Or Not Using Different Machine Learning Algorithms On A Qualified Dataset. Find Out The Correlations Between Different Attributes. Obtaining Clear Idea Of Our Proposed Data Mining Techniques And Analyze The Result And Comparing Between The Results Of Different Data Mining Techniques. We Will Analyze Our Techniques If There Is Any Possibility To Bring Improvement For Our Results.
1.3 Data Pre-Processing
Heart Disease Data Is Pre-Processed By Using Various Collections Of Records. The Dataset Contains A Total Of 303 Patient Records, Where 6records Are With Some Missing Values.
Those 6records Have Been Removed From The Dataset And The Remaining 297 Patient Records Are Used In Pre-Processing Shown In Fig.1.
Fig.1. Pre Processed Data Constraints
1.4 Feature Selection And Reduction
Among The 13 Attributes Of The Data Set, Two Attributes Pertaining To Age And Sex Are Used To Identify The Personal Information Of The Patient. The Remaining Attributes Are Considered Important As They Contain Vital Clinical Records. Clinical Records Are Vital To Diagnosis And Learning This Verity Of Heart Disease.
1.5 Classification Modelling
The Clustering Of Datasets Is Done On The Basis Of The Variables And Criteria Of Decision Tree (Dt) Features. Then, The Classifiers Are Applied To Each Clustered Dataset In Order To Estimate Its Performance. The Best Performing Models Are Identified From The Above Results Based On Their Low Rate Of Error.
Ii. Literature Review
It Is Another Type Of Fat In The Blood. High Lipoid Levels Typically Mean You Frequently Eat A Lot Of Calories Than You Burn. If It Go Too High It Can Increase Our Risk Of Heart Disease.
Ideally, Our Triglyceride Level Should Be Less Than 150 Mg/Dl. The American Heart Association (Aha) States That A Triglyceride Level Of 100 Mg/Dl Or Lower Is Considered
‘Optimal’. Non-Hdl Cholesterol. Non- High Density Compound Protein Cholesterol Is That The Distinction Between Total Cholesterol And Hdl Cholesterol (Hdl-C). Non-Hdl-C Contains Of Cholesterol In Lipoprotein Particles That Are Involved In Hardening Of The Arteries (Atherosclerosis). This Includes Beta-Lipoprotein (Ldl), Compound Protein, Intermediate- Density Compound Protein And Very-Low-Density Compound Protein. In Some Cases Non- Hdl-C Fraction Can Be Considered A Better Marker Of Risk Than Ldl Cholesterol. Now Let Us Talk A Little Bit Further About High-Sensitivity Creactive Protein. C-Reactive Protein Otherwise Known As Crp Is A Protein Your Liver Produces As Part Of Your Body’s Response To Injury Or Infection (Inflammatory Response). Crp Is A Sign Of Inflammation Somewhere In The Body. But High Sensitivity Crp Tests Cannot Figure Out Where Exactly In The Body This May Be Happening Or Why It Is Happening. Inflammation Plays A Central Role In The Process Of Atherosclerosis Where Fatty Deposits Clog Our Arteries. Now Measuring Crp Alone Will Not Tell Our Doctor Our Risk Of Heart Disease. But If We Factor In Hs-Crp Then Test Results With Other Blood Test Results Ad Risk Factors For Heart Disease Will Help To Create An Overall Picture Of Our Heart Health. Because These Is A Variability In Crp Levels That The Test Should Be Done Twice And Two Weeks Apart In Order To Get A Good And Acceptable Result.
An Hs-Crp Level Above 2.0 Mg/L Indicates A Higher Risk Of Heart Disease. This Test Screening Is Not Currently Recommended For People Without Symptoms Or A Known Risk Of Heart Disease. Cholesterol- Lowering Statin Medications May Reduce Crp Levels And Decrease Your Heart Disease Risk. Next Thing We Are Going To Discuss Will Be The Lipoprotein.
Lipoprotein, Or Lp, Is A Type Of Ldl Cholesterol. Our Lp Level Is Determined By Our Genes And Is Not Generally Affected By Lifestyle. However High Level Of Lp May Be A Sign Of Increased Risk Of Heart Disease, Though It Is Not Clear How Much Risk. Our Doctor Might Order And Lp Test If We Already Have Atherosclerosis Or Heart Disease But Appear To Have Otherwise Normal Cholesterol Levels. Lp Is Often Tested If We Have A Family History Of Early-Onset Heart Disease Or Sudden Death. Here One Thing We Should Also Add Is Drugs Are In Development To Lower Lp, But It Is Not Yet Clear What Effect Lowering Lp Will Have On Heart Disease Risk. People With High Lp Are Generally Advised To Keep A Low Ldl Cholesterol Level [8]. Up Next We Are Going To Talk About Plasma Ceramides. This Is A New Type Of Test That Measures Levels Of Ceramides In The Blood. Ceramides Are Produced By All Of Our Cells And Play A Significant Role In The Here In This Chapter We Will Be Discussing About Various Machine Learning Classifiers And Previous Work On The Heart Disease. In Machine Learning We Can Use Different Algorithms Otherwise Known As Classifiers To Help Us Predict For Our Project. Here In Our Project We Are Looking Forward To Predict The Number Of Patient That Have Heart Disease And The Number Of Patient That Do Not Have Heart Disease Running Four Algorithm To Our Data Set. The Reason We Are Going To Use Four Is That It Will Allow Us To Get Better And More Reliable Prediction.
Because If We Are Using One Algorithm Or Classifier And Do Not Have Anything Else To Compare It With Then We Cannot Say That It A Reliable Prediction Because It Might Be Giving Us A Very Good Accuracy But This Algorithm Might Not Be The Best Or More Appropriate One To Use For Our Scenario. Whereas If We Use More Than One Algorithm Or Classifier In Our Case Four Of Them, We Can Compare Them With One Another And If We Find One
Classifier Is Giving Us Accuracy That Is Not Even In The Ball Park Of The Other Algorithm Provided Accuracy We Can Understand That Something Is Going Wrong. It Can Be That The Algorithm Itself Is Not Suitable For The Job Or We Made A Mistake In Our Coding. So Using More Than One Algorithm Is Essential For Any Prediction Based System.
Now The Algorithms That We Have Chosen To Use In Our Project Are: 1. Decision Tree, 2.Naïve Bayes, 3.Svm (Support Vector Machine) And Lastly 4. Random Forest. We Will Be Discussing Each Of Those Algorithms Below. And Finally We Are Also Going To Discuss About The Previous Work That Has Been Done And Show How It Improved Over Time And What Improvements We Were Able To Bring In Our Project. Now We Are Going To Talk About Some Of The Work Done Before On Heart Disease Prediction. Many Research Has Been Done On Blood Test In Order To Predict Heart Disease. Our Blood Offer Us With Many Clues About Our Heart Condition. For Example, If Our Cholesterol In Our Blood Is High That Is A Clear Sign That We Are At The Increased Risk Of Having A Heart Attack. Other Substances In Our Blood Can Also Help Our Doctor To Determine If We Have Heart Failure Or Area Trick Of Developing Plaque Deposits In Our Arteries Also Known As Atherosclerosis. So It Is Very Important To Remember That One Blood Test Alone Is Not Enough To Determine Our Risk Of Heart Disease. The Most Vital Risk Factors For Cardiopathy Square Measure Smoking, High Blood Pressure High Cholesterol And Diabetes. Now Let Us Look At Some Of The Blood Test That We Can Do To Diagnosis And Manage Heart Disease. First Of All We Can Do The Cholesterol Test. A Cholesterol Test Also Known As Lipid Panel Or Lipid Profile, Measures The Flats (Lipids) In Our Blood. The Measurements Can Indicate Our Risk Of Having A Heart Attack Or Other Heart Disease.
The Test Is Typically Including Measurements Of Total Cholesterol. This Is A Sum Of Our Blood Cholesterol Content. If It Is High Than It Puts Us At A High Level Risk Of Having A Heart Attack. In An Ideal State, The Total Cholesterol Should Be Below 200 Mg Per Deciliter (Mg/Dl) Or 5.2 Mill Moles Per Liter (Mmol/L).Low-Density Lipoprotein (Ldl) Cholesterol. This Is Sometimes Called The ‘Bad’ Cholesterol. Too Much Of It In Blood Causes The Accumulation Of Fatty Deposits In Our Arteries, Which Reduces Blood Flow. These Plaque Deposits Typically Rupture And Cause Major Heart And Tube Issues. Our Ldl Cholesterol Level Should Be Less Than 130 Mg/Dl In Order For Us The Stay Fit. More Desirable Level Should Be Under 100 Mg/Dl, Especially If We Have Diabetes Or A History Of Heart Attack, Heart Stents, Heart Bypass Surgery Or Other Heart/Vascular Conditions. High-Density Lipoprotein (Hdl) Cholesterol. This Is Typically Referred To As The ‘Good’ Cholesterin As A Result Of It Helps Take Away Cholesterin, Keeping Arteries Open And Your Blood Flowing More Freely. Ideally, Your Hdl Cholesterol Level Should Be Over 40 Mg/Dl For A Man, And Over 50 Mg/Dl For A Woman. Triglycerides. Growth, Function And Ultimately Death Of Many Types Of Tissue.
Ceramides Square Measure Transported Through The Blood By Lipoproteins And Square Measure Related To Coronary Artery Disease. Three Specific Ceramides Have Been Liked To Plaque Buildup In The Arteries And Insulin Resistance. This Elevates The Level Of These Ceramides In The Blood Indicates A Higher Risk Of Cardiovascular Disease Within On To Five Years [8]. And Now Finally We Are Going To Talk About Natriuretic Peptides. Before Talking About Natriuretic Peptides We Need To Realize That Brain Natriuretic Peptide, Also Called B- Type Natriuretic Peptide(Bnp),Isaproteinthatourheartandbloodvesselsproduce. Bnpcan Help Us By Eliminating Our Body Fluids And Relaxing Our Blood Vessels And Funnels Sodium Into
Our Urine. When Our Heart Is Damage Do Ourbody Secretes High Levels Of Bnp In To Our Blood Stream To Try To Ease The Strain On Our Heart.
Bnp Levels May Also Rise If We Have A New Or Increased Chest Pain (Unstable Angina) Or After A Heart Attack. Now Our Bnp Level Can Also Help In The Diagnosis And Evaluation Of Heart Failure And Other Heart Conditions. Normal Levels Do Vary According To Age And Gender And Whether We Are Overweight. One Of The Foremost Vital Uses Of Bnp Is To Do To Map Out Whether Or Not Shortness Of Breath Is Because Of Failure. Now For People Who Have Heart Failure, Establishing Abase Line Bnp Can Be Helpful And Future Tests Can Be Used To Help Gauge How Well Our Treatment Works. In Addition To That A Variation Of Bnp Called N-Terminal Bnp Also Is Useful In Diagnosing Heart Failure And In Some Laboratories Is Used Instead Of Bnp N- Terminal Bnp May Also Be Useful In Evaluating Our Risk Of A Heart Attack And Other Problems If We Already Have Heart Disease. Finally A High Level Of Bnp Alone Is Not Enough To Diagnose A Heart Problem In That Case Our Doctor Will Also Consider Our Risk Factors And Other Blood Test Results. So As We Can See Due To The Demand Of Having A System That Can Predict Heart Disease Prediction Many Studies Have Taken Place.
Iii. Framework Design 3.1 Existing System
In This System, The Input Details Are Obtained From The Patient. Then From The User Inputs, Using Ml Techniques Heart Disease Is Analyzed. Now, The Obtained Results Are Compared With The Results Of Existing Models With In The Same Domain And Found To Be Improved.
The Data Of Heart Disease Patients Collected From The Uci Laboratory Is Used To Discover Patterns With Nn,Dt, Support Vector Machines Svm And Naive Bayes. The Results Are Compared For Performance And Accuracy With These Algorithms. The Proposed Hybrid Method Returns Results Of 87% For F-Measure, Competing With The Other Existing Methods.
Disadvantages
1.Prediction Of Cardio Vascular Disease Results Is Not Accurate.
2.Data Mining Techniques Does Not Help To Provide Effective Decision Making.
3.Cannot Handle Enormous Datasets For Patient Records.
3.2 Proposing System
After Evaluating The Results From The Existing Methodologies, We Have Used Python And Pandas Operations To Perform Heart Disease Classification For The Data Obtained From The Uci Repository. It Provides An Easy To Use Visual Representation Of The Dataset, Working Environment And Building The Predictive Analytics. Ml Process Starts From A Pre-Processing Data Phase Followed By Feature Selection Based On Data Cleaning, Classification Of Modeling Performance Evaluation. Random Forest Technique Is Used To Improve The Accuracy Of The Result.
Advantages
1. Increased Accuracy For Effective Heart Disease Diagnosis.
2. Handles Roughest (Enormous) Amount Of Data Using Random Forest Algorithm And Feature Selection.
3. Reduce The Time Complexity Of Doctors.
4. Cost Effective For Patients.
Iv. Result Analysis
The Results Have Been Obtained By Applying Different Classification Algorithms. In Our First Experiment We Used The Whole Dataset With All Features And Applied Support Vector Machine, Decision Tree, Random Forest, Gaussian Naive Bayes Contains Accuracy Of The Different Algorithms That We Applied On Our Dataset.
1. A Lot Of Features Can Affect The Accuracy Of The Algorithm. So Working With The Features Is Very Important. There Are Few Reasons For Which Some May Want To Work With Some Selected Features.
2. Choosing Less Features Helps Us To Train Faster.
3. By Picking Up The Most Important Features, We Can Use Interactions Between Them As New Features. Sometimes This Gives Surprising Improvement.
4. Some Features Are Linearly Related To Others. This Might Put A Strain On The Model.
5. Feature Selection Means To Select Only The Important Features In-Order To Improve The Accuracy Of The Algorithm.
6. It Reduces Training Time And Reduces Over Fitting.
4.1. Feature Importance
A Very Basic Question That We Might Ask Of A Model Is What Features Have The Biggest Impact On Predictions? This Concept Is Called Feature Importance. In Dataset There May Be Some Attributes Which Don’t Effect The Prediction That Much. In Some Cases, Few Attributes May Decrease The Accuracy Level Of A Model. So, It Is Important To Work With The Correct Attributes. So Far We Have Worked With All The Features Of The Dataset And Listed The Accuracy Of Different Models. Now, We Want To See The Change Of Accuracies Of Different Classifiers After Selecting A Subset Of The Attributes. We Can See The Importance Of A Feature Via Decision Tree And Random Forest.
4.2. Crossvalidation
Cross Validation Is An Essential Step In Model Training. It Tells Us Whether Our Model Is At High Risk Of Over Fitting. In Many Competitions, Public Lb Scores Are Not Very Reliable.
Often When We Improve The Model And Get A Better Local Cv Score, The Lb Score Becomes Worse. It Is Widely Believed That We Should Trust Our Cv Scores Under Such Situation.
Ideally We Would Want Cv Scores Obtained By Different Approaches To Improve In Sync With Each Other And With The Lb Score, But This Is Not Always Possible. Usually 5-Fold Cv Is Good Enough. If We Use More Folds, The Cv Score Would Become More Reliable, But The Training Takes Longer To Finish As Well. However, We Shouldn’t Use Too Many Folds If Our Training Data Is Limited. Otherwise We Would Have Too Few Samples In Each Fold To Guarantee Statistical Significance. Many Times The Data Is Imbalanced, I.E. There May Be A High Number Of Class1 Instances But Less Number Of Other Class Instances. Thus We Should Train And Test Our1 Algorithm On Each And Every Instance Of The Dataset. Then We Can Take An Average Of All The Noted Accuracies Over The Dataset.
1. The K-Fold Cross Validation Works By First Dividing The Dataset Into K-Subsets.
2. Let’s Say We Divide The Dataset Into (K=10) Parts. We Reserve 1 Part For Testing And Train The Algorithm Over The Other 9 Parts.
3. We Continue The Process By Changing The Testing Part In Each Iteration And Training The Algorithm Over The Other Parts. The Accuracies And Errors Are Then Averaged To Get An Average Accuracy Of The Algorithm.
4. An Algorithm May Under Fit Over A Dataset For Some Training Data And Sometimes Also Over Fit The Data For Other Training Set. Thus With Cross-Validation, We Can Achieve A Generalized Model.
4.3 Analysis
From All The Differential Algorithms Performed Better Depending Upon The Situation Whether Cross Validation And Feature Selection Is Used Or Not. Every Algorithm Has Its Intrinsic Capacity To Out-Perform Other Algorithm Depending Upon The Situation. For Example, Random Forest Performs Much Better With A Large Number Of Datasets Than When Data Is Small. While Support Vector Machine Performs Better With A Smaller Number Of Datasets. In Case Of Decision Tree Missing Values Play An Important Role. Even After Imputing It Can’t Give The Result Which It Can With A Perfect Dataset. Gaussian Naive Bayes Is The Best Classifier On This Dataset Which Were Shown In The Fig.2 To Fig.7. The Reason Of Its Pre- Assumption That All The Attributes Are Independent. If There Was A Dependency Between The Attributes In The Dataset It Would Have Given Less Accuracy.
Fig.2. Gaussian Naïve Bayes
Fig.3. Gradient Boosting
Fig.4. Support Vector Machine
Fig.5. Random Forest
Fig.6. Logistic Regression
Fig.7. Gaussian Mixture Model
V. Conclusion
In This Research We Have Tried To Compare Different Machine Learning Algorithms And Predict If A Certain Person, Given Various Personal Characteristics And Symptoms, Will Get Heart Disease Or Not. The Main Motive Of Our Report Was To Comparing The Accuracy And Analyzing The Reasons Behind The Variation Of Different Algorithms. We Have Used Cleveland Dataset For Heart Diseases Which Contains 303 Instances And Used 10-Fold Cross Validation To Divide The Data Into Two Sections Which Are Training And Testing Datasets.
We Have Considered 13 Attributes And Implemented Four Different Algorithms To Analyze The Accuracy. By The End Of The Implementation Part,We Have Found Gaussian Naïve Bayes And Random Forest Are Giving The Maximum Accuracy Level In Our Dataset Which Is 91.21 Percent And Decision Tree Is Performing The Lowest Level Of Accuracy Which Is 84.62 Percent. Probably For Other Instances And Other Datasets Other Algorithm May Work In Better Way But In Our Case We Have Found This Result. Moreover, If We Increase The Attributes, Maybe We Can Found More Accurate Result But It Will Take More Time To Process And The System Will Be Slower Than Now As It Will Be Little More Complex And Will Be Handling More Data’s. So Considering These Possible Things We Took A Decision Which Is Better For Us To Work With.
Vi.Future Enhancement
This Work Will Be Useful In Identifying The Possible Patients Who May Suffer From Heart Disease In The Next 10 Year. This May Help In Taking Preventive Measures And Hence Try To Avoid The Possibility Of Heart Diseases For The Patients. So When A Patient Is Predicted As Positive For Heart Disease, Then The Medical Data For The Patient Can Be Closely Analysed By The Doctors. An Example Would Be- Suppose The Patient Has Diabetes Which May Be The Cause For Heart Disease In Future And Then The Patients Can Be Given Treatment To Have Diabetes In Control Which In Turn May Prevent The Heart Disease. The Heart Disease Prediction Can Be Done Using Other Machine Learning Algorithms. Logistic Regression Can Also Perform Well In Case Of Binary Classification Problems Such As Heart Disease Prediction.
Random Forests Can Perform Well Than Decision Tree. Also, The Ensemble Methods And Artificial Neutral Networks Can Be Applied To The Data Set. The Results Can Be Compared And Improvised.
References
[1] Canfield, J., Hansen, M. V., And Rackner, V. (2005). Heart Disease. Health Communications. [3] Carney, R. M. And Freedland, K. E. (2010). Psychotherapies For Depression In People With Heart Disease. Depression And Heart Disease, Page 145–168.
[2] Gold, J. C. And Cutler, D. J. (2000). Heart Disease. Enslow Publishers.
[3] Healey, J. (2005). Heart Disease. Spinney Press.
[4] Hook, S. V. (2001). Heart Disease. Smart Apple Media.
[5] Jiang,W.Andxiong,G.L.(2010). Epidemiologyofthecomorbiditybetweendepression And Heart Disease. Depression And Heart Disease, Page 1–37.
[6] Johansson, P. (1998). Heart Disease. Enslow Publishers.
[7] Klapholz, M. (2003). Heart Failure In The Elderly. Heart Disease, 5(4):241–243.
[8] Mensah, G. A. (2009). The Burden Of Valvular Heart Disease. Valvular Heart Disease: A Companion To Braunwalds Heart Disease, Page 1–18.
[9] Moodie, D. (2016). The Journal Congenital Heart Disease-2016. Congenital Heart Disease, 11(1):5–6.
[10] Morris,P.,Warriner,D.,Morton,A.,Andmayhew,P.(2016). Heartdisease. Jpmedical Ltd.
[11] Otto, C. M. (2009). Evaluation Of Valvular Heart Disease By Echocardiography. Valvular Heart Disease: A Companion To Braunwalds Heart Disease, Page 62–84.
[12] Rajamannan, N. M. (2009). Cellular, Molecular, And Genetic Mechanisms Of Valvular Heart Disease. Valvular Heart Disease: A Companion To Braunwalds Heart Disease, Page 39–
54.
[13] Roberts, W. C. And Ko, J. M. (2009). Clinical Pathology Of Valvular Heart Disease.
Valvular Heart Disease: A Companion To Braunwalds Heart Disease, Page 19–38.
[14] Shavelle, D. M. (2009). Evaluation Of Valvular Heart Disease By Cardiac Catheterization And Angiocardiography. Valvular Heart Disease: A Companion To Braunwalds Heart Disease, Page 85–100.
[15] Sheen, B. (2004). Heart Disease. Thomson/Gale.
[16] Silverstein, A., Silverstein, V. B., And Nunn, L. S. (2006). Heart Disease. Lerner.
[17] Stout, K. (2009). Valvular Heart Disease In Pregnancy. Valvular Heart Disease: A Companion To Braunwalds Heart Disease, Page 424–436.
[18] Thomas, R. (2002). Heart Disease. Vega. [25] Tiger, S. And Reingold, M. (1986). Heart Disease. J. Messner.
