Comparing the Architecture and Performance of AlexNet, Faster R-CNN, and YOLOv4 in the Multiclass Classification of Alzheimer Brain MRI Scans
Blog Author: Archita Khaire
Currently there is no cure for Parkinson's disease. It doesn’t always affect how long you live but has a severely negative impact on the quality of life of patients and their caregivers. Early detection of PD can improve the symptoms dramatically.
Image Credit: https://www.pcori.org/research-results/pcori-stories/improving-life-women-parkinsons-disease
Early Detection of PD using AI
There are multiple early signs that can indicate onset of Parkinson's disease. Health data collected around these signs is used to build machine learning algorithms that can predict the PD.
1. Speech impairments like dysphonia (defective use of the voice), hypophonia (reduced volume), monotone (reduced pitch range), and dysarthria (difficulty with articulation of sounds or syllables) are used to detect PD. UCI Machine Learning Repository: Parkinsons Telemonitoring Data Set was created by Athanasios Tsanas and Max Little of the University of Oxford, in collaboration with 10 medical centers in the US and Intel Corporation who developed the tele-monitoring device to record the speech signals.
Image Credit: DOI:10.1109/SIBGRAPI.2016.054
3. Difficulty in walking is another sign of growing Parkinson's disease. Deep learning algorithms are used in Gait analysis to classify the movement properties into two classes PD and non-PD using features such as imbalance, frequency of falls etc.
4. Researchers are attempting to detect PD using human olfactory data. In patients with OD, loss of smell starts many years early before onset of motor skill symptoms. There is potential to diagnose PD much early with this method.
Leveraging AI to treat PD
Parkinson’s is caused by the death of dopamine-producing nerve cells. Scientists at Stanford school of medicine identified Miro1, mitochondrial protein which resists removal of damaged cells. They analyzed millions of drugs using AI to identify a compound that could bind with Miro and enhance the cell mechanism to remove damaged nerve cells.
Researchers at John Hopkins have identified the glucagon-like peptide receptor, or GLP1 receptor as potential receptor for slowing down PD. They are leveraging to AI to check if diabetes drug could be used to prevent neurological cell deaths.
Blog Author: Archita Khaire
Common symptoms of Alzheimer's include:
Early identification of patients suffering from Alzheimer's is the biggest challenge as it is often confused with Dementia. While some forms of Dementia could be reversed, Alzheimer's is irreversible. Emerging blood-based biomarkers offer opportunities for screening patients before performing lumbar puncture tests to check for amyloid proteins in cerebrospinal fluid.
Image Credit: DOI: 10.7717/peerj.6543/fig-
Machine learning models such as SVM, SHMR (Sparse High-order Interaction Model with Rejection option) could be used to detect AD patients using inexpensive and easily accessible biomarkers (e.g., Plasma). Only those patients difficult to diagnosis are recommended for invasive and/or more expensive screening (e.g., CSF).
There are image based deep learning algorithms that can detect the progression of Alzheimer's disease based on MRI or positron emission tomography (PET) images of patient.
Image Credit: https://doi.org/10.3389/fninf.2018.00035
In the early stages of Alzheimer's disease, an MRI scan of the brain may be normal. In later stages, MRI may show a decrease in the size of different areas of the brain (mainly affecting the temporal and parietal lobes)
Image Credit: Journal of Nanobiotechnology volume 19, Article number: 72 (2021)
Researchers are now working on ultrasensitive non-invasive micro-biomarkers which in combination with deep learning algorithms could be used to detect Alzheimer's disease.
Scientists are using artificial intelligence to screen 80 FDA-approved drugs and reveal which could be used as Alzheimer’s treatments.
Image Credit: Nature Communications volume 12, Article number: 1033 (2021)
Image Credit: https://ec.europa.eu/programmes/horizon2020/en/news/neurotwin-proposes-novel-therapy-alzheimer%E2%80%99s-disease
Researchers are building computational framework to represent the mechanisms of interaction of electric fields with personalized brain networks and assimilate neuroimaging data in order to design personalized optimization strategies to treat Alzheimer’s disease.
Using neuroimaging data from Alzheimer's disease, scientists will build a model that recapitulates the networks and the dynamical landscape of the individual brain. The objective is to employ this model to design and test personalized neuromodulation protocols capable of restoring healthy dynamics.
Blog Author: Archita Khaire
Diabetes is one of today’s greatest global problems. As per International Diabetes Federation, currently there are more than 450 million people suffering from diabetes across the globe and count is projected to increase to 700 million by 2045. It is one of the fastest growing emergency causing 4 million deaths worldwide every year.
Image Credit: International Diabetes Federation
You may not feel any symptoms from diabetes at first. High blood sugar (hyperglycemia) and high levels of insulin (the hormone that manages blood sugar levels) start to damage your body silently, many years before you’re diagnosed with diabetes.
Glucose is the fuel that feeds your body’s cells, but to enter your cells it needs a key. Insulin is that key. People with type 1 diabetes (#T1D) don’t produce insulin. People with type 2 diabetes (#T2D) don’t respond to insulin as well as they should and later in the disease often don’t make enough insulin. The main causes of T2D include lifestyle, physical activity, dietary habits and heredity, whereas T1D is thought to be due to autoimmunological destruction of the Langerhans islets hosting pancreatic-β cells.
Numerous algorithms and different approaches have been applied, such as traditional machine learning algorithms, ensemble learning approaches and association rule learning in order to predict diabetes. Here are few examples:
SVM for diabetes prediction
ANT_FDCSM for diabetes detection
Image Credit: Critical-depth Raman spectroscopy enables home-use non-invasive glucose monitoring (plos.org)
With advances in smart wearables and edge computing, researchers are now building non-invasive devices for continuous glucose monitoring (#CGM). Scientists at MIT have shown that they can use Raman spectroscopy to directly measure glucose concentrations through the skin.
Type 2 diabetes is a complex and progressive disease characterized by various metabolic defects and affecting multiple organs.
Image Credit: https://doi.org/10.1016/j.amjmed.2013.06.007
Chronic diabetic complications include: heart failure, diabetic neuropathy, nephropathy, retinopathy, and diabetic foot.
Multiple machine learning algorithms have been developed for automatic detection of diabetic neuropathy.
Diabetic nephropathy (DN) is defined by elevated urine albumin excretion or reduced glomerular filtration rate (GFR), or both. This serious complication causes destructive scarring in diabetic patients.
Pathologists usually classify DN based on a visual assessment of glomerular pathology using immunofluorescence microscopy and electron microscopy. AI scientists are using image classification algorithms to enhance and fast track the detection of Diabetic nephropathy.
Image Credit: Appl. Sci. 2020, 10(18), 6185; https://doi.org/10.3390/app10186185
For retinopathy as well, researchers have developed image classification algorithms that can detect diabetes from retinal images.
Researchers are experimenting the effectiveness of e-nose technology, using machine learning classifiers, to predict single- and poly-microbial species targeted for diabetic foot infection.
Diabetes Medication and Care
FDA has approved artificial pancreas that automatically monitors blood glucose and provides appropriate insulin doses.
Image Credit: Artificial pancreas system upgraded with AI algorithm (medicalxpress.com)
AI methods have found their way into daily self-management of diabetes, mainly in the form of smartphone apps and wearables such as smart watches: For example, food recognition apps use machine learning methods and cloud-based image analyzing and may help people with diabetes to make more informed food choices and insulin treatment decisions.
Artificial Intelligence is already transforming diabetes care. It will continue to have transformational impact on the everyday lives of people with diabetes.
Blog Author: Kulani Melaku
Cardiovascular diseases(CVDs) are diseases that affect the heart and blood vessels of the body. This disease is the number one cause of death worldwide.
There are many types of CVDs ie: Arrhythmia, Valve Disease, Coronary Artery Disease, Heart Failure, Peripheral Artery Disease, Aortic Disease, Congenital Heart Disease, Pericardial Disease, Cerebrovascular Disease, and Deep vein thrombosis are the common ones.
AI can help cardiac care in 3 major areas.
AI Aided Diagnostic
Machine learning, deep learning, and cognitive computing are examples of AI approaches that have the potential to transform how cardiology and cardiovascular care are conducted (for example, how we generate knowledge, evaluate data, and make decisions), particularly in cardiovascular imaging.
Echocardiogram in the parasternal long-axis view, showing a measurement of the heart's left ventricle
#Echocardiography is a test that uses sound waves to produce live images of your heart. The image is called an echocardiogram. This test allows your doctor to monitor how your heart and its valves are functioning.
AI tools, in particular machine learning, provide new possibilities to enhance the accuracy of image interpretation in clinical echocardiography practice, especially between non-expert clinicians.
The algorithm that is integrated into the software is capable of automatically calculate the following in few seconds: the volumes of the left chambers, the systolic flow, and the ejection fraction of the LV from the data acquired with 3D echocardiographic techniques.
Researcher Sengupta developed a cognitive machine-learning algorithm, trained with speckle tracking echocardiographic(STE) data, to differentiate constrictive pericarditis from restrictive cardiomyopathy.
Narula also showed that supervised learning algorithms could differentiate athlete heart and hypertrophic cardiomyopathy, using STE data, more accurately than traditional measure systems. Another potential field of application of ML models in echocardiography is heart valve disease(HVD).
HVD is an increasingly common pathology that can benefit from cardiac imaging ML integration through early diagnosis, treatment, or surgery planning.
B. Magnetic Resonance Imaging (MRI) In cardiac MRI, ventricular segmentation is one of the fields with more potential for the application of ML models. It makes it possible to quantify the volumetry and improve the efficiency and reproducibility of clinical assessments.
Researcher Avendi used deep learning algorithms trained through cardiac MRI datasets, for the automatic detection and segmentation of the right ventricular chamber foreseeing the accuracy of these algorithms. Likewise, for left ventricular segmentation, several automated neural networks(ANNs) have been successfully developed, especially for cardiac cine MRI.
Another application of ML in cardiac MRI takes place in the detection of subacute or chronic myocardial scar.
C. Cardiac Computed Tomography (CT scan)
ML image analysis techniques in cardiac CT are increasingly used in the diagnosis and risk assessment of coronary artery disease and atherosclerosis.
Coronary computed tomographic angiography is a noninvasive modality to detect coronary artery disease. To characterize coronary plaque, automatic coronary artery calcium scoring in CCTA using ML models provides added clinical value by reducing false-positive and interobserver variability. Researcher Gonzàles used a convolutional neural network to calculate the Agatston score from CT without prior segmentation of coronary artery calcification.
An electrocardiogram (ECG) is commonly performed to test heart function in emergency room patients. The test can identify cardiac abnormalities but not failure.
Automated ECG Interpretation
Blood tests can be used to identify markers of cardiac failure but are unreliable. To help with diagnosis, Mayo Clinic researchers taught artificial intelligence to distinguish between ECG patterns of people ultimately diagnosed with heart dysfunction and those who were not. The AI-enhanced ECG performed better and faster at diagnosing dysfunction than standard blood tests.
Researchers used AI to identify a new biomarker for heart attack. Researchers first used fat biopsies from patients undergoing cardiac surgery to analyze genes associated with inflammation, scarring, and blood vessel formation. The expression of these genes was matched with coronary CT scan images to determine features indicating changes to the fat surrounding the heart. Researchers then used AI to develop the fat radiomic profile (FRP), which identifies red flags in the fat lining the blood supply to the heart. This new profile could be used to predict heart attacks in patients up to five years before they occur.
AI Aided Continuous Monitoring
An AI wearable can help prevent up to a third of heart failure repeat hospital admissions. The device is worn by heart failure patients for up to three months after being discharged from the hospital and performs continuous ECGs. The AI establishes normal heart rate, heart rhythm, respiratory rate, and walking and sleep patterns for each patient. It then analyses deviations from these norms for indications the patient’s heart condition is worsening. The system was able to predict rehospitalization with 85% specificity up to 10 days before patients were readmitted.
An AI hospital monitoring device has received approval for home use by patients with chronic obstructive pulmonary disease (COPD) and heart failure. The device monitors pulse, respiration, oxygen saturation, temperature, and mobility. The data is analyzed for warning signs and health care providers are automatically alerted.
A personal AI device is available for detecting atrial fibrillation (or AFib). The device is worn around the chest and performs continuous ECGs. AI interprets the data and alerts, wearers, to irregularities via mobile app. Individuals have control of their data and choose when to share it with their doctor.
Research shows patients who wear monitoring devices are more likely to receive a timely diagnosis than those who receive standard care.
Jeff Williams, Apple's chief operating officer, speaks about the Apple Watch Series 4 at the Steve Jobs Theater in Cupertino, California, U.S., on September 12, 2018. REUTERS/Stephen Lam/File Photo
AI is truly on the verge of redefining how cardiovascular care is delivered to patients. Companies and researchers have implemented AI in every step of the process, from continuous monitoring of basal heart rate for early warning signs to a quick and efficient noninvasive diagnosis of cardiac conditions. AI is also making later stages of the care pathway more efficient, such as real-time visualization of the cardiac anomaly and subsequent therapy selection. However, the question that remains to be answered is will this advanced technology, in the long run, be able to bring down the cost and time of cardiovascular care for patients.
Artificial intelligence's arrival in the cardiovascular profession is bringing with it a variety of new opportunities for providing innovative tailored care. The way we practice cardiology, particularly in the field of cardiac imaging, is changing, and physicians must be prepared. Tech companies are forging new links between patients and doctors, transforming healthcare from a passive to a pervasive activity. Physicians should not be terrified of AI's integration into cardiology; rather, they should embrace it, because their specialist knowledge will always be important.
Blog Author: Archita Khaire
Lung cancer is the second most common cancer in both men and women.
Image Credit: Cancer.org
The American Cancer Society’s estimates for lung cancer in the United States for 2021 are:
About 235,760 new cases of lung cancer (119,100 in men and 116,660 in women)
About 131,880 deaths from lung cancer (69,410 in men and 62,470 in women)
Image Credit: American Lung Association
There are 2 main types of lung cancer and they are treated very differently.
Small cell lung cancer (SCLC)
Small cell lung cancer is fast-growing lung cancer that develops in the tissues of the lungs. The types of small cell lung cancer are named for the kinds of cells found in the cancer and how the cells look when viewed under a microscope.
- Small cell carcinoma
- Combined small cell carcinoma
Image credit: https://doi.org/10.1016/j.omto.2021.02.004
Advances in genomics, machine learning and development of new mathematical models have facilitated insights into the intratumoral heterogeneity and specific genetic alterations of #SCLC
The SCLC subtypes defined by the relative expression of four key transcriptional regulators:
Non-small cell lung cancer (NSCLC)
Adenocarcinoma start in the cells that would normally secrete substances such as mucus.
Squamous cell carcinoma start in squamous cells, which are flat cells that line the inside of the airways in the lungs.
Large cell (undifferentiated) carcinoma can appear in any part of the lung. It tends to grow and spread quickly, which can make it harder to treat.
Lung cancer occurs when people breathe in dangerous, toxic substances. Smoking is the number one cause of lung cancer followed by radon exposure as the second most common cause. Small cell lung cancer is almost always associated with cigarette #smoking.
Image Credit: Translational Cancer Research, Vol 5, No 4 (August 2016) / Radiomics applied to lung cancer: a review
A deep learning model can detect patterns on chest computed tomography (CT) scans to identify smokers at high risk for Lung Cancer. This method is commonly known as Radiomics and proceeds in two phases—first a training or feature selection phase and then a second testing or application phase. During the training phase algorithms perform automatic extraction of quantitative features from medical images. In the testing phase, the #Radiomics are applied to a particular patient’s image, with the process being similar to the training phase but now the selected features are identified by the algorithm, extracted, and then used to classify the patient. Here is sample dataset if you want to build your own image classifier for chest CT scans.
Particle pollution is also considered one of the influential long-term factors of lung cancer. Particle pollution refers to a mix of tiny solid and liquid particles that are in the air we breathe. Many of the particles are so small as to be invisible, but when levels are high, the air becomes opaque.
Depending on type and stage of lung cancer doctors recommend combination of surgery (VATS), radiation therapy, chemotherapy, targeted drug therapy, Immunotherapy and Radiofrequency ablation to treat the lung cancer.
Researcher are working to augment Video-assisted thoracic surgery (VATS) with robots powered by artificial intelligence to make it minimally invasive. Da Vinci surgical robots the first FDA approved AI-powered robots used for thoracic surgery.
AI models are now used for real-time monitoring of cancer cell death during Radiofrequency ablation (#RFA)
Image Credit: doi: 10.1136/jitc-2020-001343
Immunotherapy uses drugs to help a patient’s immune system fight cancer. But some patients are at risk of paradoxical response called “hyperprogression. Researcher are now using AI models to identify the patients who are at harm by immunotherapy.
Blog Author: Archita Khaire
Breast cancer treatment depends on the stage and type of breast cancer.
In localized cancer, patients receive breast-conserving surgery and radiation therapy to clear the cancer cells. The further breast cancer progresses, the greater the combination of therapies the patient needs. These therapies may include surgery, chemotherapy, hormone therapy, and radiation therapy.
The doctor can recommend a different type of surgery based on the size & location of cancer, size of the breast.
Image Credit: https://medizzy.com/feed/6090735
Lumpectomy: The surgeon removes cancer tissue and leaves a small amount of healthy tissue behind. In the U.S. about 25 percent of women need additional surgery after a #lumpectomy.
Image Credit: Biomedical Optics Express Vol. 12, Issue 5, pp. 2647-2660 (2021)
Researchers are working to build a medical device that can perform optical coherence tomography (OCT) on excised breast tissue in the operation room and feed imaging data into an AI engine to detect cancer cells. The real-time feedback can help the surgeon determine if more tissue needs to be removed thus reducing the possibilities of additional surgeries in the future.
Mastectomy: The surgeon removes the breast tissue (including the skin and nipple) and the tissues that cover the chest muscles.
Modified radical mastectomy: Rare procedure to remove the entire breast, many lymph nodes under the arm, the lining of the chest muscles, and, in some cases, the chest wall muscles.
Some patients undergo breast reconstruction after mastectomy. Autologous breast reconstruction with deep inferior epigastric artery perforator (DIEP) flap has become an integral component of the holistic treatment of breast cancer patients. Combination of 3D printing and Artificial intelligence is being experimented to augment breast reconstruction surgery.
Sometimes results of breast reconstruction are disappointing. The CINDERELLA (Comparing patient’s decision on aesthetic outcome with the BCCT.core objective evaluation after controlled teaching in patients proposed for breast cancer locoregional treatment) project aims to create a gold standard method for the aesthetic evaluation by giving patients a better insight into the outcomes, allowing them to judge more objectively, also using inputs from both objective and subjective factors. Machine learning is also being used to determine patient satisfaction post reconstruction surgery.
Chemotherapy uses drugs that stop the growth of cancer cells. A person can receive chemotherapy drugs orally or via injection into the vein or muscle. Machine-learning algorithms are now used to determine the fewest, smallest doses that could shrink the cancer cells thus reducing toxic side effects of chemotherapy.
Image Credit: Khademhosseini Laboratory
Researchers are now using biomedical sensors combined with artificial intelligence to monitor and adjust the chemotherapy treatment.
Hormonal therapy blocks cancer cells from getting the #hormones they need to grow. AI is now effectively used to personalize hormone therapy using the chemical composition of the cancer tissue.
Image Credit: https://blog.einstein.ai/receptornet/
ReceptorNet uses advanced machine learning algorithms to determine estrogen receptor status without the need to use an IHC stain. It learns to assign high attention weights to tiles in the H&E image that are the most important for decision-making and to assign low attention weights to tiles that are insignificant for this task. The presence or absence of estrogen receptors would help doctors determine the course of therapy.
Image Credit: https://doi.org/10.6004/jnccn.2020.7554
Immunotherapy works with your body’s immune system to help it fight cancer cells or to control side effects from other cancer treatments. Machine learning has vastly improved the cancer immunotherapy.
Image Credit: https://doi.org/10.1016/j.apsb.2021.02.007
Artificial Intelligence is used in many ways in immunotherapy. ML classification algorithms are used to characterize the cancel cells which help determine accurate immunotherapy for the breast cancer patients. ML algorithm are used to predict immunotherapy scores thus increasing the chance of successful cancer immunotherapy.
Radiation therapy uses high-energy X-rays to kill cancer cells. External radiation therapy (EBRT) uses a machine to send radiation to the affected area of the body. Internal radiation therapy (brachytherapy) uses a radioactive substance in needles that surgeon places near the cancer tissue.
Image Credit: The Breast, Volume 49, February 2020, Pages 194-200
Artificial intelligence is being effectively used to individualize radiation therapy for breast cancer. Deep learning algorithms use neural networks to identify and generate contouring patterns to be combined with information from multiple sources, in order to generate volumes on CT images that can subsequently be validated for RT dose planning by the physician.
Big data and advancement in AI are poised to improve breast cancer treatment. Precision therapy and non-invasive surgeries are improving the patient survival rate and recovery after the treatment.
Blog Author: Archita Khaire
Breast cancer is the most common cancer worldwide. Last year 2.3 million women were diagnosed with breast cancer and 685,000 deaths were reported. The National Cancer Institute states that more than $500 million dollars are spent on breast cancer research yearly. Even with the large investment, approximately 43,000 women in the U.S. are expected to die in 2021 from breast cancer.
Based on the human epidermal growth factor receptor 2 (Her2), progesteron receptor (PR), and estrogen receptor (ER), breast cancer is categorized into four molecular subtypes:
Image Credit: Journal of Cancer
From early detection to survivorship, Artificial Iintelligence is helping doctors and researchers improve the survival rate for patients suffering from #breastcancer.
There are many established risk factors for breast cancer, including gene mutations, estrogen levels, and breast density. A primary example: studies have shown that women with the BRCA1/2 gene mutation have a high risk of breast cancer.
Image Credit: BMC Medical Genomics volume 14, Article number: 122 (2021)
Several deep learning algorithms are available to detect these mutations to predict break cancer risk. BRCA exchange provides data on thousands of BRCA variants to develop understanding of a patient's cancer risk.
Image credit: Nature Communications volume 11, Article number: 5727 (2020)
Researchers are using deep learning algorithms to classify estrogen receptors from Breast Cancer Metabolomics Data. Immunohistochemistry (IHC) is used most often to find out if cancer cells have estrogen and progesterone receptors. There is excellent open source library (histolab) available on GitHub that can be used to get started with quantitative image analysis of stained slides.
Image credit: MIT News
Mammography is considered the first line of defense in breast cancer detection. MIT has developed an image-based deep learning algorithm that can detect cancer in #mammographs up to 5 years in advance. Another important feature of the algorithm is that it is equally effective for various races; this is not the case with many other algorithms that mostly rely on mammographs of white women to train their ML models.
Recently, a thermal imaging based tool was approved by the European Union, and it uses cloud based AI analytics to detect breast lesions.
Image credit: IEEE Transactions on Medical Imaging ( Volume: 39, Issue: 4, April 2020)
A Breast ultrasound is often used as a follow-up test if a #mammography detects abnormalities. Deep neural networks are available to classify the ultrasound images as benign or malignant.
Dense breasts are one of the most common risk factors for breast cancer development. Approximately 40% of women have dense breasts. indicating a higher risk for breast cancer. However, a mammography may miss over 1/3 of cancers from patients with dense breasts. 3D ultrasound and 3D mammography (also known as tomosynthesis) are better alternatives to effectively identify tumors in dense breasts. Duke and IMAIOS have developed a free online app and ML algorithm to train the reading of breast #tomosynthesis images.
A Breast MRI is more invasive and only recommended for women at a higher than average risk of breast cancer. A Breast MRI has a higher rate of false-positive cases when compared to mammographies. Although AI is augmenting breast MRIs to improve accuracy, further research is needed to enhance their efficiencies. Recently, research presented at the 2019 Breast Imaging Symposium demonstrated the successful use of advanced algorithms on breast MRIs, which are traditionally used in cases of patients with particularly dense breast tissue.
Image Credit: Volume 34, Issue 5, 12 November 2018, Pages 840-851.e4
If mammographs and MRI fail to confirm the cancer, a #biopsy is recommended to diagnose breast cancer. Combined with machine learning, Laser Raman Spectroscopy (LRS) offers real time detection of cancer cells during needle biopsy.
With rapid advancements in Artificial Intelligence and robotic technology, the preliminary screening of breast cancer without physicians and the diagnosis of breast cancer without radiation are possible in the future!
Blog Author: Archita Khaire
Medulloblastoma is the most prevalent brain tumor type in children. It is also called a cerebellar primitive neuroectodermal tumor (PNET)—that starts in the region of the brain at the base of the skull, called the posterior fossa. In the United States, about 400 children are diagnosed with #medulloblastoma every year. The survival rate of Medulloblastoma is about 70%.
Image Credit: JNS Journal of Neurosurgery
Common symptoms of Medulloblastoma include Headaches, Nausea, Vomiting, Clumsiness or similar balance problems, and vision problems. Sometimes Medulloblastoma enters into the ventricles of the brain causing a buildup of cerebrospinal fluid which can result in symptoms of an enlarged head. Based on gene changes in the cancer cells they are grouped into 4 subtypes: WNT, SHH, Group III, Group IV. SHH (also know as Sonic HedgeHog) is the most common in adults.
MRI and CT scan and common methods to diagnose the Medulloblastoma. Machine learning algorithms have enabled neurosurgeons to detect gene mutations associated with cancer within the plasma. However, this type of early diagnosis does not work in pediatric brain cancers because these cancers often have very few mutations. Instead, certain critical genes are turned on or turned off by epigenetic changes that regulate their activity in cases of pediatric brain tumors. There is evidence that DNA methylation changes at cytosine-phosphate-guanine (CpG) sites are associated with cancer. Researchers are performing Quantitative epigenetics to detect biomarkers of Medulloblastoma.
Image Credit: The Lancet Oncology, https://doi.org/10.1016/S1470-2045(17)30243-7
Surgery is the most common treatment to remove the medulloblastoma tumor cells. Doctors are using enhanced imaging techniques (Cortical mapping using artificial intelligence) that enables the identification of areas of the brain that control the senses, language, and motor skills.
Image Credit: https://sciglow.com/developing-brain-maps-through-artificial-intelligence/
Researchers at John Hopkins are working on the possibility of using biodegradable, lab-engineered nanoparticles that could kill medulloblastoma tumor cells. The #nanoparticles are diffused within the tumor due to the increased permeability of the tumor vessels, however; the low clearance prohibits the formulations from getting out. This methodology is mostly used for siRNA but not for DNA plasmid delivery. It is a therapeutic strategy commonly known as Suicide Gene Therapy.
Artificial intelligence is opening a new era of nano medicine to treat cancers. To improve the efficacy of gene therapy, researchers are developing machine learning models that could predict the shapes and sizes of nanoparticle delivery systems, a form of drug delivery to cancer site, the most effective path of delivery.
Researchers are working to learn more about medulloblastoma, ways to diagnose it in early stages, non-invasive ways to treat it and prevent from reoccurrence. Proton therapy is one of the area where use of artificial intelligence is being experimented to improve accuracy and efficacy of the treatment.