Machine Learning Methods As a Good Solution for Cancer Diagnosis
Solutions using machine learning methods help with oncology diagnosis and sharing medical data across the medical community. We support a Cancer Center team in the development of software allowing us to analyze medical images by using machine learning methods (deep learning).
In order to be able to diagnose and mitigate the effects of the disease earlier than before, a number of advanced techniques of microscopic image analysis had to be developed.
To illustrate, it was necessary to made possible:
- segmenting of nuclei
- classifying of nuclei types
- making a statistical summary of samples
- finding regions of interests
- classifying tumor analysis of mitosis
Uniquely, a comprehensive approach to the problem resulted in the design of software available from both API and web platform.
Consultations with professionals allow to apply mechanisms of mathematical modeling, machine learning methods. Moreover data analysis and image processing.
Additional values: saving, upload and comment on images with other specialists from all over the world.
One of the biggest challenges in the development process was the lack of data for the learning process. To overcome these difficulties, we use the data augmentation technique in order to expand the data set. Moreover, it is worth mentioning that during the cooperation with Cancer Center, we could also provide business support. Low IT knowledge of physicians and difficulties in understanding caused difficulties in accessing to the end customer. To brighten up all the issues we conducted educational workshops using previously prepared MVP. Showing what we can achieve with machine learning technology allowed us to change the doctors’ attitude.
Finally, the development works were successfully completed. The Cancer Center software and communication tools for doctors and specialists enabled fast and effective prevention of cancer. In the full spectrum of functionalities. In addition, the Cancer Center team achieve considerable success in two MICCAI 2015 competitions in the field of “Imaging & Digital Pathology”. The team won 1 place in the Combined Radiology and Pathology Classification competition . Also they won 3. place in the Segmentation of testicles in a pathological image (cell segmentation in the histological image). Last of all The Cancer Center was selected from among the best and was able to present its solution to doctors and scientists. They came to the conference in Munich from all over the world.