Machine Learning Enhances Ability To Predict Survival From Brain Tumors

MedicalResearch.com Interview with:

Lee Cooper, Ph.D. Assistant Professor of Biomedical Informatics Assistant Professor of Biomedical Engineering Emory University School of Medicine - Georgia Institute of Technology

Dr. Cooper

Lee Cooper, Ph.D.
Assistant Professor of Biomedical Informatics
Assistant Professor of Biomedical Engineering
Emory University School of Medicine – Georgia Institute of Technology

MedicalResearch.com: What is the background for this study? What are the main findings? 

Response: Gliomas are a form of brain tumor that are often ultimately fatal, but patients diagnosed with glioma may survive as few as 6 months to 10 or more years. Prognosis is an important determinant in selecting treatment, that can range from simply monitoring the disease to surgical removal followed by radiation treatment and chemotherapy. Recent genomic studies have significantly improved our ability to predict how rapidly a patient’s disease will progress, however a significant part of this determination still relies on the visual microscopic evaluation of the tissues by a neuropathologist. The neuropathologist assigns a grade that is used to further refine the prognosis determined by genomic testing.

We developed a predictive algorithm to perform accurate and repeatable microscopic evaluation of glioma brain tumors. This algorithm learns the relationships between visual patterns presented in the brain tumor tissue removed from a patient brain and the duration of that patient’s survival beyond diagnosis. The algorithm was demonstrated to accurately predict survival, and when combining images of histology with genomics into a single predictive framework, the algorithm was slightly more accurate than models based on the predictions of human pathologists. We were also able to identify that the algorithm learns to recognize some of the same tissue features used by pathologists in evaluating brain tumors, and to appreciate their prognostic relevance. Continue reading

Machines Can Be Taught Natural Language Processing To Read Radiology Reports

MedicalResearch.com Interview with:

Eric Karl Oermann, MD Instructor Department of Neurosurgery Mount Sinai Health System New York, New York 10029 

Dr. Oermann

Eric Karl Oermann, MD
Instructor
Department of Neurosurgery
Mount Sinai Health System
New York, New York 10029 

MedicalResearch.com: What is the background for this study? What are the main findings? 

Response: Supervised machine learning requires data consisting of features and labels. In order to do machine learning with medical imaging, we need ways of obtaining labels, and one promising means of doing so is by utilizing natural language processing (NLP) to extract labels from physician’s descriptions of the images (typically contained in reports).

Our main finding was that (1) the language employed in Radiology reports is simpler than normal day-to-day language, and (2) that we can build NLP models that obtain excellent results at extracting labels when compared to manually extracted labels from physicians.  Continue reading

Deep Learning System Can Screen For Diabetic Retinopathy, Glaucoma and Macular Degeneration

MedicalResearch.com Interview with:

Blausen.com staff (2014). "Medical gallery of Blausen Medical 2014". WikiJournal of Medicine 1 (2). DOI:10.15347/wjm/2014.010. ISSN 2002-4436. Illustration depicting diabetic retinopathy

Illustration depicting diabetic retinopathy

Dr. Tien Yin Wong MD PhD
Singapore Eye Research Institute, Singapore National Eye Center,
Duke-NUS Medical School, National University of Singapore
Singapore

MedicalResearch.com: What is the background for this study? What are the main findings?

Response: Currently, annual screening for diabetic retinopathy (DR) is a universally accepted practice and recommended by American Diabetes Association and the International Council of Ophthalmology (ICO) to prevent vision loss. However, implementation of diabetic retinopathy screening programs across the world require human assessors (ophthalmologists, optometrists or professional technicians trained to read retinal photographs). Such screening programs are thus challenged by issues related to a need for significant human resources and long-term financial sustainability.

To address these challenges, we developed an AI-based software using a deep learning, a new machine learning technology. This deep learning system (DLS) utilizes representation-learning methods to process large data and extract meaningful patterns. In our study, we developed and validated this using about 500,000 retinal images in a “real world screening program” and 10 external datasets from global populations. The results suggest excellent accuracy of the deep learning system with sensitivity of 90.5% and specificity of 91.6%, for detecting referable levels of DR and 100% sensitivity and 91.1% specificity for vision-threatening levels of DR (which require urgent referral and should not be missed). In addition, the performance of the deep learning system was also high for detecting referable glaucoma suspects and referable age-related macular degeneration (which also require referral if detected).

The deep learning system was tested in 10 external datasets comprising different ethnic groups: Caucasian whites, African-Americans, Hispanics, Chinese, Indians and Malaysians

Continue reading

Machine Learning Applied To Predicting High-Risk Breast Lesions May Reduce Unnecessary Surgeries

MedicalResearch.com Interview with:

Manisha Bahl, MD, MPH Director, Breast Imaging Fellowship Program, Massachusetts General Hospital Assistant Professor of Radiology, Harvard Medical School

Dr. Bahl

Manisha Bahl, MD, MPH
Director, Breast Imaging Fellowship Program,
Massachusetts General Hospital
Assistant Professor of Radiology,
Harvard Medical School

MedicalResearch.com: What is the background for this study? What are the main findings?

Response: Image-guided biopsies that we perform based on suspicious findings on mammography can yield one of three pathology results: cancer, high-risk, or benign. Most high-risk breast lesions are noncancerous, but surgical excision is typically recommended because some high-risk lesions can be upgraded to cancer at surgery. Currently, there are no imaging or other features that reliably allow us to distinguish between high-risk lesions that warrant surgery from those that can be safely followed, which has led to unnecessary surgery of high-risk lesions that are not associated with cancer.

We decided to apply machine learning algorithms to help us with this challenging clinical scenario: to distinguish between high-risk lesions that warrant surgery from those that can be safely followed. Machine learning allows us to incorporate the full spectrum of diverse and complex data that we have available, such as patient risk factors and imaging features, in order to predict which high-risk lesions are likely to be upgraded to cancer and, ultimately, to help our patients make more informed decisions about surgery versus surveillance.

We developed the machine learning model with almost 700 high-risk lesions, then tested it with more than 300 high-risk lesions. Instead of surgical excision of all high-risk lesions, if those categorized with the model to be at low risk for upgrade were surveilled and the remainder were excised, then 97.4% malignancies would have been diagnosed at surgery, and 30.6% of surgeries of benign lesions could have been avoided.

Continue reading