With the increase in biological and medical data it has become necessary for medical and bioinformaticians to have some automated approaches to identify different patterns it their data, so as to predict or have some useful information. Many applications have been described above for genetic algorithm, along with these applications GA has been applied in protein structure prediction, RNA structure prediction and Motif finding. Basic steps of GA are almost same in many applications but it requires expertise, parameters and involves a huge number of randomness and can provide different results in outcomes.
Applications of Genetic Algorithm in medicine
Oncology
Screening tests suggests a valuable chance cancer detection at early stages, which when keep an eye on by proper handling could recover the patient’s survival rate. Developing a non-invasive procedure for the detection of cervical cancer, Duraipandian et al, using colposcopy developed Raman spectra of cervical area and the biomolecular evidence produced was investigated through a Genetic algorithm-partial least square-discriminant study scheme to distinguish between dysplastic and a normal cervix.
Partial least square is a statistical technique directing to discovery a model of linear regression between some predictor variables and dependent variable. This arrangement was capable to distinguish dysplasia from a ordinary cervix with a sensitivity of 72% and specificity of 90%. The beginning of DNA microarray technique has opened the mode for immense gene expression profiling, revolutionizing the arena of molecular prognosis and diagnostics. Conversely, producing large data sets of analytical and statistical challenges requires the need to know main analytical genes. Because of inherent capability of genetic algorithms for finding optimal solution among complex and large solutions by many synchronized interactions, they can be applied for analyzing microarray data from numerous cancer cell lines. A scientist, Dolled-Filhart with his team, produced microarray data through staining tissues from breast cancer with numerous specific antibodies for different markers for searching a set of minimal biomarkers with highest organization and prediction values in patients with breast cancer. Data examined using genetic algorithm exhibited three markers with existing antibodies could describe a group of patients having more than 95% survival rate. Tan e with his team, showed a study to explore the association between trace elements of soil and cervical cancer death reported in China. A blend of genetic algorithm and PLS, was used to select 5 out of 25 trace elements and a model of LSSVM (machine of least square support vector) was established. LSSVM is a technique used in train machine for inferring a function to search a pattern in respective training data. A combination of (genetic algorithm and PLS) and LSSVM have potential to predict cervical cancer which lies on trace elements. One of the significant revealing factors inducing the selection of an suitable healing method for cancer patients is detection of disease prediction. In a reflective study which carried out on more than 200 patients, performances of 4 diverse data mining approaches were compared to define the consequence of cancer patients after hospitalization not being at terminal stages. In contrast to other approaches, genetic algorithms had chosen the slightest number of descriptive variables to predict the consequence of patients.
Cardiology
Genetic algorithms have been used in diverse cardiovascular medicine fields. Atherosclerotic plaques are signs of many strokes and myocardial infarctions. Estimation of plaque mechanical features i.e elasticity would allow medical doctor to trace better and map susceptible or unsteady plaques. A system involving genetic algorithms has been used for factor assessment essential for precise elasticity quantification to regulate tissue elasticity. This system is gradient-based approach used for factor assessment of inadequacy of gradient-based methods for inhomogeneous solution having numerous local minima and prerequisite for considerable computer application time limits. Discovery of clinical and biomarker proteomics is promptly developing in diagnosis, prognosis, disease follow-up and medical. Recent tools i.e mass spectrometry can produce data of thousands of proteins from samples (patient). But, the complexity and cost of these approaches and statistical and computational approaches for analysis requires the selection of a limited related markers for developing medical assay. An improved form of the genetic algorithm which is maintained through local floating has been employed, enhanced procedure predicted the chances for major adverse cardiac event (abbreviated as MACE). This practice was competent to choose a section of seven proteins comprising myeloperoxidase to calculate the chances of MACE with an accuracy of 77%, which outclassed over numerous existing approaches.
Models of logistic regression have often used in disease detection. Because of outstanding performance, a genetic algorithm has been used to choose the superlative logistic regression system variables, directing to model the myocardial infarction existence in patients having chest pain. The genetic algorithm-based technique was greater in selecting variables for traditional approaches. Important fundamentals in the automatic analysis of ECG (electrocardiogram) is the recognition of QRS complexes allowing the calculation of variability in heart rate and other related analytical factors. Simple and effective genetic algorithm was introduced to identify QRS complexes after that, f-waves and p-waves were successfully mined from patient databases. These algorithms let a comprehensive research for electrocardiogram details.
Endocrinology
Hypoglycemia is the frequently occurring insulin therapy complication for patients having diabetes mellitus type 1 (T1DM). It can bring changes in EEGs patterns. Nguyen et al, joined ANNs, genetic algorithms, and Levenberg-Marquardt training methods for the detection of hypoglycemia on the basis of EEG signals. Association between EEG signals and blood glucose was modeled with the help of ANNs which was trained using the global search capability of genetic algorithms and for the local searching LM were joined. Data from four electrocardiogram factors resulting from two electrocardiogram channels were used to analyze system for hypoglycemia detection with a sensitivity of 75% and specificity of 60%. In another study, a genetic algorithm-based manifold regression using fuzzy interpretation system was established for the detection of nocturnal hypoglycemia in patients (children) with T1DM. Corrected QT interval and heart rate was used to detect hypoglycemia with 75% sensitivity and over 50% specificity.
Pediatrics
Cardiotocography is an inexpensive and non-invasive method to evaluate the fetus uterine contractions and heart rate to regulate fetal health. [17] a genetic algorithm has been applied to pick the ideal cardiotocogram recording features to maintain a vector machine (SVM) classifier. Outcomes of this study presented that the new structure categorized fetal health status as abnormal or normal with an accuracy of 99.3% to 100%, which was found to be higher than ANN algorithm intended for the same reason. Autism is a neurodevelopmental syndrome that can be seen in early childhood stages and can be characterized by weakened social working, non-verbal and verbal communications and behaving repetitively. To identify autism on the basis of data provided by microarray gene expression, a method was used in which genetic algorithms selected the most related genes linked with diseases. Most occurring chosen genes comprise TOP1, RMI1, NRIP1, CEP350, ZFHX3, PSENEN, NFYA, ANP32A, SP1 and SEMA4C. Such genes give an input for ensemble classifiers including random forest classifiers and SVM. System introduced predictable autism with sensitivity of 96% and specificity of 83%. [18]Acute lymphoblastic leukemia is typically found in children and has various subtypes. Examination of data derived from gene expression of tumorous cells can facilitate us in categorizing cancers. Because huge information size can be produced from profiling method of microarray gene expression, a genetic algorithm was used to choose the most related genes required to classify Acute lymphoblastic leukemia.
Silhouette statistics was used as a distinguishing function between six subtypes of Acute lymphoblastic leukemia. The suggested procedure reached a 100% taxonomy precision and used less selective genes related to other approaches. Aneuploidy is a state where a few or one chromosome in the cell nucleus are below or above the normal number of chromosomes in a species. Conservative chromosomal studies on amniocentesis models are achieved for certain identification of fetal aneuploidy, rather extensive required time for such procedures requires the advancement of faster analytical tests. The proteomic profile from amniotic fluid samples was recognized through mass spectrometry technique and the produced data was measured by a genetic algorithm. The suggested technique could perceive aneuploidy with sensitivity of 100%, specificity of 72%–96%, positive predictive value of 11%–50% and negative predictive value of 100% .
Surgery
Powerful mathematical algorithms, ANNs are able to predict system behaviors. Because of ANNs predictive value, ANN based on genetic algorithm was established to find the surgery results for patients having NSCLC (non-small cell lung-cancer). The genetic algorithm was used to facilitate optimization to prevent local minima. Model of GANN have potential to predict NSCLC outcomes from patients more precisely and significantly than models of logistic regression. Addition of tumor size in designs can meaningfully increase estimate outcomes. For example the number of aged patients requiring cardiac surgeries rises. Because of extraordinary occurrence of comorbid situations in elderly, proper predictions of postoperative mortality and morbidity would be helpful, preventing overestimation of denial and surgery risk for deserving patients, which could occur with some forecasting models. Applying a genetic algorithm, Lee et al, exhibited that a small length of break after cardiac operation was linked with younger age.
Pulmonology
In pulmonology common diagnostic techniques distinguish between lung infections and monitor the analytical method toward precise methods. For automation lung sound analysis, a hybrid genetic algorithm based ANN was planned. The genetic algorithm was useful to enhance the ANN training factors and decrease the calculation time. This system can categorize the lung sounds into wheeze, crackle and normal.
Calculating partial pressure for CO2 in the arterial blood, PaCO2 is key managing critically ill patients. Difficulties related with arterial blood sampling can be avoided by non-invasive procedures to identify PaCO2 i.e evaluation of breathe out CO2 at end-expiration (PetCO2) possibly will be useful in typical persons; though, their practice in ill patients might be influenced and less supportive . A genetic algorithm has been designed to calculate the PaCO2 by different 11 variables from capnography. The suggested system can enhance the accuracy and PaCO2 prediction biasness.
Infectious diseases
Tuberculosis is a fatal infectious disease for both developing and undeveloped contries after the occurrence of HIV (human immunodeficiency virus). To forecast the analysis (non-tuberculosis vs. tuberculosis patients), factors comprised of analysis factors and laboratory statistics were used to project an genetic algorithm ANN. The grouping precision of the classification was around 95%, which was greater than the consequences acquired by other procedures. HAART (Highly active antiretroviral therapy), a central part treating HIV, is comprised of a mixture of numerous antiretroviral treatments targeting to decline virus replication. Since long-lasting HAART cure requires patient agreement and might be linked with some adverse effects, controlled treatment disruption has been suggested to diminish not only adverse effects, but also the selecting stress on the virus that can direct towards the appearance of strong particles. Consequently, Castiglione et al, invented a genetic algorithm -based system to select the best HAART usage arranged to regulate HIV and aid the immune system to reconstruct. A simulated model of the invulnerable system was used to evaluate the properties of anti-HIV medicines on virtual patients and it could accomplish therapeutic consequences and defense against an adaptable contamination similar to a full-length cure.
Radiotherapy
IMRT or Intensity modulated radiotherapy was established to transmit precise radiation dose to a target i.e the brain, neck, head or prostate. Development of IMRT includes range of 5 to10 angles for projecting wavelet and defining the radioactivity dose. Using genetic algorithm can increase the variety of scaffold angles in a rational time frame. Genetic algorithm-based irradiation development has been used for patients having different types of cancer comprising pancreatic, brain tumors and rhabdomyosarcoma. Genetic algorithm have used to enhance the project of stereotactic radiotherapy treatment and radiosurgery plans.
Orthopedics
Biomedical engineering has presented countless results for orthopedic operation. THA (Total hip arthroplasty) has enhanced the execution of several spiking hip joint infections. Yet, disappointments of femoral stem of Total hip arthroplasty can cooperation the achievement of medication. Ishida et al, stated genetic algorithm in manipulating an improved femoral stem constituent geometry. Genetic algorithms have also been misused to choose the best strategy of tibial fastening screws to diminish the likelihood of screw fracture or slackening. In another study, a mixture of genetic algorithms and ANNs was applied to project backbone pedicle screws for repairing spinal ruptures.
Hybrid algorithm was capable to project screws with a greater exhaustion life and model pullout and flexible features. Scoliosis is a 3D irregularity of vertebral axis bends. The development of the infection, which only occurs in a small fraction of patients, is examined by sequential X-rays time. Meanwhile numerous X-rays exposures can rise the chances of having cancer, it is necessary to evaluate the infection using different approaches. Jaremko et al, established a genetic algorithm-based ANN algorithm to evaluation the angle of vertebral axis irregularity from indices of stem surface irregularity. The hybrid system was competent to define the angle irregularity within the accuracy of 5% in two third patients or more.
Neurology
Multiple sclerosis is neural system inflammatory disease described by the development of white matter blemishes. Computer-aided analysis has been useful for finding pathologic types in these patients. One study showed that, a genetic algorithm was established to distinguish the MS injuries of brain MRIs. The resemblance of injuries can be found by genetic algorithm and a radiologist was about 87%. A useful predictive approach is EEG to find the anomalous brain electrical discharges taking place during an attack. To project an automatic method for discovery of abnormal EEG indications, numerous learning algorithms such as Quickprop, LM, Momentum, Delta-bar delta, and Conjugate gradient were used for ANN training of EEG-based grouping of healthy versus epileptic persons. A genetic algorithm was used to discover the best factors and construction of the ANN. The consequences verified that the learning algorithms technique combined with the genetic algorithm was best procedure for ANN training got an overall success of almost 96.5% performance.
Numerous reports have recommended that mitochondrial dysfunction shows role in Parkinson’s disease. Mitochondrial genetics has peculiarities, a simple evaluation of mitochondrial mutations among disease and healthy situations may not be so revealing. Consequently, a genetic algorithm has been devised to identify naturally important patterns of mitochondrial mutations found in Parkinson’s patients. The planned scheme was capable to identify Parkinson’s disease with an accuracy of 100% founded on patterns of mitochondrial DNA mutations.
Comments
Post a Comment