Machine learning applications in cancer prognosis and prediction pdf

Posted on Thursday, June 17, 2021 10:58:19 PM Posted by Lieraisifi - 18.06.2021 and pdf, english pdf 1 Comments

machine learning applications in cancer prognosis and prediction pdf

File Name: machine learning applications in cancer prognosis and prediction .zip

Size: 2051Kb

Published: 18.06.2021

Blood Adv ; 4 23 : — Machine learning ML is rapidly emerging in several fields of cancer research.

Applications of Machine Learning in Cancer Prediction and Prognosis

Refworks Account Login. Open Collections. UBC Theses and Dissertations. Featured Collection. Mira Keyes, Radiation Oncology Supervisory Committee Member iii Abstract In this thesis, we presented the design steps for developing new, reliable, and cost-effective diagnostic and prognostic tools for cancer using advanced Machine Learning ML techniques.

We proposed tools to improve the diagnostic, prognostic and detection accuracy of quantitative digital pathology by incorporating advanced image analysis, image processing, and classification methods. In this thesis we presented our ML image-based analytic approaches for three cancer types prostate, cervix, and kidney with different scale ranges from the sub-micron to multiple centimeters. In this thesis, we demonstrated the full workflow to design an automated prognostic and grading system specially designed for prostate cancer.

We started with demonstrating techniques for prostate glandular structures detection. Next, we introduced an automated cell segmentation method along with an interactive segmentation correction method requiring minimum user-interaction and finally, we introduced our ML classification algorithms.

Next, we studied renal carcinoma. We presented the workflow of renal carcinoma classification from image processing to feature selection and development of machine learning classification techniques. We extracted the features from renal vessel structures and demonstrated the design steps of machine learning classifiers to discriminate between different renal carcinoma subtypes using these features. We applied our techniques for cervical pre-cancer abnormality detection.

We showed the whole pipeline of designing an automated classification method starting from tissue imaging to the development of ML classifiers using both classical and deep-learning methods. Although we conducted these studies on specific cancer types, a modified version of our algorithm could be applied to other cancers and disease sites.

The number of patients diagnosed with cancer increases daily and despite advances in medical technology and extensive research, the incident rate of cancer is still very high. Over the past few decades, Machine Learning ML methods have been applied successfully in the healthcare domain. ML offers promising avenues for prediction of disease progression, extraction of clinically relevant information from the data, therapy planning, patient management and much more.

These methods, if used in clinical settings, could improve patients life quality and increase the cancer survival rate. Calum MacAulay. Chapter 2: A version of this material has been published in [C1]. Nilgoon was the lead investigator, responsible for all major areas of concept formation, and analysis, as well as manuscript composition.

Amir Bakhtiari helped with algorithm development, Paul Gallagher was responsible for image preprations using the Hyperspectral imaging system. The entire work was conducted under the supervision of Dr. Calum MacAulay and support and input from Dr. Pierre lane, and Dr. Mira Keys. Chapter 3 and Chapter 4: A version of this material has been published in [J1] and [J2]. Nilgoon was the research lead on this study. She was responsible for the study design, algorithm development and validation as well as manuscript composition.

Pierre Lane, Dr. Mira Keys and Dr. Martial Guillaud. Chapter 5: A version of this material has been published in [C2]. Nilgoon was responsible for the design and coding. Amir Bakhtiari helped with algorithm development. Chapter 6: The content and patient material of this chapter is based on a collaboration with Inria France, with Dr. Xavier Descombes' guidance. Dr Ambrosetti was responsible for data collection and image annotation.

Nilgoon was responsible for the literature review, feature implementation and data analysis. The development and construction of the Multispectral Imaging System developing the device and collecting the patient images was performed as a part of a NIH funded collaboration with US researchers Drs.

Cantor SB. Nilgoon was the research lead on the study. She was responsible for the literature review, designing and implementing the proposed algorithms, performing all experiments, analyzing the results and writing the manuscript. Pierre Lane, and Dr. Michele Follen. The thesis was written by Nilgoon Zarei, with editing assistance from Dr. Bioimaging , As we can see in this histogram, the number of publications in this domain increased over the years.

This histogram is normalized by the number the total number of publications on Machine Learning. This figure shows the pipeline of our study. The focus of this chapter is on the Glandular Segmentation.

In the later chapters, we will explain the classification and validation processes. This figure shows our hyperspectral image acquisition set-up on the left and a representation of hyperspectral image construction that could be acquired by this system on the right. The image in the red-box is the first three PCA components and is used for further analysis. The second image with orange background is second, third and fourth PCA components and the last image is the xviii third fourth and fifth PCA components.

No satisfactory result is obtained using the last two images. They are presented here only for illustration. At the last step, another K-means clustering sequence is applied to detect nuclei within each detected gland. GL stands for glass, S stands for stroma and N stands for nerve trunk.

The second row shows the gland detection results where G stands for gland; we showed an example of nerve trunk detection and removal in the third image. The third row shows the nuclei segmentation results. This is an example of a core with densely packed tumor glands and small luminal spaces. In this example, the tumor has glandular structures with slit-like luminal spaces. The gland image is not shown here.

A the PCA image with a zoom-out of a section of the core xix with glands. B shows the failure of the gland detection. In the red box, multiple glands were mislabeled and were not detected. The nuclei image is not shown here. June The superimposed rectangles show the locations with a high probability of the presence of glands mostly in the center. In the previous chapter, we discussed our algorithm for glandular segmentation. The focus of this chapter is on the cell segmentation.

In the later chapters, we will describe the classification and validation processes. The red circles enclose cells where the difference between the two thresholding methods is obvious. The red circles enclose areas where debris existed and where debris removed upon applying the proposed algorithm. Failure is due to clumped nuclei. This method finds the skeleton of a cluster and uses the endpoints as a possible location for the cell center.

As we can see in the rightmost figure, the initial four nuclei resulted in six potential nuclei centers, we will explain how to find the four nuclei out of these six spots Red dots represent detected nuclei This algorithm consists of two clustering methods K-means clustering and density-based clustering in the first run, pixels are selected if they belong to the same cluster generated xxii from two clustering methods , on the second run we feed the remaining pixels to the density-based clustering and choose those with large area.

The blue dot represents the original location selected by the segmentation algorithm and other dots are potential locations that the user could possibly select The focus of this chapter is on the classification step. The black line represents a random line. The purple line is the baseline using SVM. LSVM is used to classify between vs red line, 6 vs 7 shown with the blue line and 8 vs 9 shown with the green line.

In this image, many pathological features are illustrated, although, in this study, we are interested in angiogenesis hallmark, other pathological indications such as infiltration of inflammatory cells are worth mentioning.

Other features such as abundant clear cell carcinoma, although is not discussed in this chapter, were the topic of previous chapters where we showed the importance of nuclei features and how they can be used for classification purposes. The processes consist of imaging and pre-processing of the captured images.

This workflow starts with an RGB image, pathological image input and the resulted output is the skeleton of the vessels where all the nuclei and other structures are removed. This workflow starts with an RGB image, the pathological image of renal carcinoma input and ends with the output which is the skeleton of the vessels where all the nuclei and other structures are removed. In each node, a selection is made as shown in red. The final class selection would be based on the majority votes of all trees.

Boosted tree is an ensemble algorithm that uses the majority votes of some stump trees, in each step the weights for the next decision making would be updated as shown with the green arrows above. The processes mentioned in the blue boxes are completed and fully developed by our group. The processes illustrated in green boxes are the focus of this chapter. This figure shows the study design. Patients were consented for an IRB-approved study, images were acquired, the Loop Electrical Excision Procedure LEEP was carried out, the specimen was processed and annotated, and the dataset was subjected to analysis for training and testing classification

Machine Learning Applications in Breast Cancer Diagnosis

Overview DOI: With rapid advances in experimental instruments and protocols, imaging and sequencing data are being generated at an unprecedented rate contributing significantly to the current and coming big biomedical data. Meanwhile, unprecedented advances in. Meanwhile, unprecedented advances in computational infrastructure and analysis algorithms are realizing image-based digital diagnosis not only in radiology and cardiology but also oncology and other diseases. Machine learning methods, especially deep learning techniques, are already and broadly implemented in diverse technological and industrial sectors, but their applications in healthcare are just starting.

Machine learning applications in cancer prognosis and prediction

Providing prognostic information at the time of cancer diagnosis has important implications for treatment and monitoring. Although cancer staging, histopathological assessment, molecular features, and clinical variables can provide useful prognostic insights, improving risk stratification remains an active research area. We used a weakly-supervised approach without pixel-level annotations, and tested three different survival loss functions. The DLS was developed using 9, slides from 3, cases and evaluated using 3, slides from 1, cases. In multivariable Cox regression analysis of the combined cohort including all 10 cancers, the DLS was significantly associated with disease specific survival hazard ratio of 1.

This capacity is especially appropriate to restorative applications, particularly those that rely upon complex proteomic and genomic estimations. Therefore, AI is much of the time utilized in malignant growth determination and discovery. All the more as of late AI has been applied to malignant growth guess and forecast.

We apologize for the inconvenience...

Refworks Account Login. Open Collections. UBC Theses and Dissertations. Featured Collection.

Universitatea Lucian Blaga din Sibiu. An early diagnosis of breast cancer offers treatment for it; therefore, several experiments are in development establishing approaches for the early detection of breast cancer. High complexity models are associated with high accuracy and high variability. Early prediction of breast cancer will help with the survival of breast cancer patients.


The predictive models discussed here are based on various supervised ML techniques as well as on different input features and data samples.


breast cancer detection using machine learning pdf

Navigare articol

Breast cancer accounts for the largest number of cancer cases all around the world. These numbers are particularly high in developing countries. In United States US , breast cancer disease is the most common diagnosed cancer in women. It is ranked as second cause of cancer death in women. Early detection is the key to reduce the mortality rates.

S urgical site infection SSI following neurosurgical operations is a burdensome complication in the field. Such complications can impact morbidity, mortality, and economics. The financial burden caused by craniotomy infections is often compounded by the direct costs incurred by prolonged hospitalization of the patient, diagnostic tests, treatment, and reoperation. Machine learning ML is used for outcome prediction in the neurosurgical field. Several ML algorithms have been developed using complex mathematical models that can learn from clinical data from, for example, neuro-oncology, neurovascular surgery, traumatic brain injury, and epilepsy.

Either your web browser doesn't support Javascript or it is currently turned off. In the latter case, please turn on Javascript support in your web browser and reload this page. Today, despite the many advances in early detection of diseases, cancer patients have a poor prognosis and the survival rates in them are low. Recently, microarray technologies have been used for gathering thousands data about the gene expression level of cancer cells. These types of data are the main indicators in survival prediction of cancer. This study highlights the improvement of survival prediction based on gene expression data by using machine learning techniques in cancer patients.

Skip to search form Skip to main content You are currently offline. Some features of the site may not work correctly.

COMMENT 1

  • It is clear that the application of ML methods could improve the accuracy of cancer susceptibility, recurrence and survival prediction. Based on [3], the accuracy of cancer prediction outcome has significantly improved by 15%–20% the last years, with the application of ML techniques. Brooke S. - 28.06.2021 at 01:47

LEAVE A COMMENT