Color Theme

Fernando Yepes Calderon

Fernando Yepes

  • Ph.D In biomedicine
  • Web developer
  • Data mining expert
"Coding is traveling through the unlimited space of creativity"

About Me

I am an electronics engineer with a Ph.D. in biomedicine and a very passionate person about giving solution to any tech challenge or improving the operation of almost anything through technology. I have focused my efforts in web implementations and services using PHP and more recently Python. In this learning journey, I have worked in a myriad of fields helping lawyers, marketing experts, software developers among others and I have visited several countries as well. I also ventured research, where medical imaging and data mining are my daily routine. Let me know what your technological need is and let's put hands on it

Projects

Some of the projects I have executed are even serving on-line or have multimedia records. Some others, despite very interesting have been developed to companies that kept the right to use and show, so I am not allowed to put that material here.

Skills

The list of skills is organized from general tools (top) to specialized platforms (bottom).

95% Python Big data
90% Python Art. Intelligence
90% PHP Web Dev.
80% Python Firmware
86% Magnetic Resonance Diffusion
90% Magnetic Resonance Anatomical
90% Hospital PACS Tools Dev
95% Python Imaging

Education

2016
Ph.D in Biomedicine

Universidad de Barcelona/Children's Hospital Los Angeles, Spain/USA

2011
Master in Bioengineering

Universidad de Barcelona (Scholarship requirement), Spain

2008
Master Parcours Sytèmes et Images

INSA-Lyon, France

2007
Master in Biomedical Sciences

Universidad de los Andes, Colombia

2003
CISCO CCNA

Networking Specialization at Universidad del Bosque, Colombia

2002
Electronics Engineer

Universidad del Valle, Colombia

Experience

Feb 2015 - Today
Children Hospital LA - Neurosurgery. LA-USA

Creating automatic tools to speed up diagnosis

Sept 2014 - Feb 2015
Keck School of Medicine. LA-USA

Developing algorithms to detect cancer in early stages

Jan 2009 - Jul 2014
GYM GROUP SA. Cali-Col

Leader of web platform and operations

Aug 2007 - Dec 2008
Logicciel Sarl. Lyon-FR

Web developer

Feb 2007 - Dec 2007
Sandgate Comunity Centre. Brisbane-AU

Networking expert

Feb 2003 - Jan 2007
Several companies, Bogota-Col

Tech advisor

Feb 2003 - Jan 2007
Several Universities, Bogota-Col

Lecturer

Feb 2000 - Nov 2002
LPEE - Universidad del Valle. Cali - Col

Laboratory technician. Quality in electrical measurements

Feb 1996 Aug 1997
Deportivo Cali. Cali-Colombia

Professional Soccer Player

Languages

I use these languages in a regular bases for conversations, presentations and also for reporting and documenting

100% Spanish Native Language
90% English After living/working in USA and Australia
85% French After living/working in France

Scientific Contributions

Not that I do not believe in collaborations, but here I just post what I proposed, executed and published as a project leader

The use of artificial intelligence in medicine will take us to another dimension of diagnosing and running medical treatments. Here, we use a classifier that goes voxel by voxel telling us whether each visited unit of information belongs to the target structure. This implementation is fully automatic and can work inside the legally exigent clinical environments where instruments should be HIPAA compliant. In the end, the code is capable of separating the lateral ventricles and calculate their volumes, providing the neurosurgeons with numbers even when facing abnormality. Knowing the lateral ventricle volumes is relevant in diagnosing and treating hydrocephalus/microcephalus patients.

Go to publication
The use of artificial intelligence in medicine will take us to another dimension of diagnosing and running medical treatments. Here, we use a classifier that goes voxel by voxel telling us whether each visited unit of information belongs to the target structure. This implementation is fully automatic and can work inside the legally exigent clinical environments where instruments should be HIPAA compliant. In the end, the code is capable of separating the lateral ventricles and calculate their volumes, providing the neurosurgeons with numbers even when facing abnormality. Knowing the lateral ventricle volumes is relevant in diagnosing and treating hydrocephalus/microcephalus patients.

Go to publication
A python implementation ready to be connected to existing tools in hospitals. This code library uses the masks that separate the tumors in medical images to quantify the edge spiculation. Edge irregularities are common in aggressive cancer, and this implementation shows numbers in an easy-to-interpret histogram. The evolution of this idea will be able to detect cancer from images and define its status with high accuracy, so diagnosis is accomplished faster, more accurately and without the need of the often painful histologies.

Go to publication
In this development we let the machine to solve a difficult problem for humans: to define when the patient is healthy, has a mild cognitive impairment or is currently suffering from Alzheimer disease. To accomplish this, we feed the computer with geometrical features extracted from cortical structures in the brain. Additionally, we present to the medical community, the regions of the brain that show a higher probability of being affected by the malady.

Go to publication
The tractography is Magnetic Resonance based technique that tracks the displacement of water molecules. The trace of water has a preference in regions where myelin -- a lipidic substance that covers neurons -- is in place. As myelination is a process that parallels maturation, it is possible to monitor maturation through tractography. We executed this technique in a cohort of 49 patients, 24 them born prematurely. Among other interesting findings, we managed to capture the catch up effect that allows premature babies to be equally capable than born term children.

Go to publication
This is hardware effort to increase the quality of MRI images in small animal imaging sessions. The idea is to create MRI receptors, make them smaller, stick them together and thus, recover the field of view lost with the smaller devices. The problem is that mutual induction appears when you put the small receptors close to each other. We proposed and successfully tested alternatives to eliminate the mutual induction in intimate coils and reduce the cross talking among 2nd and 3rd order neighboring.

Go to publication
This is not an evolution of the EdgeRunner algorithm. This is a new approach that verifies the existence of spiculations in the masses by correlating the findings on two planes (axial and sagittal). The implementation still runs along the edge, but the peak detection mechanism is based on derivatives and filters. The results are presented in the same histogram fashion as in the original EdgeRunner algorithm.

Go to publication

Keep in Touch

+1 818 930 0521

www.fernandoyepesc.com

fernandoyepesc@gmail.com

Contact

Send me a message for any question or invitation, I will be writing back to you in the act.