Jan Hering's Homepage

Welcome to my homepage. I am currently working at post-doc position in the Biomedical Imaging Algorithms (BIA) group which is led by Prof. Jan Kybic. The group is part of the Center for Machine Perception (CMP), at the Department of cybernetics, Faculty of electrical engineering of the Czech Technical University in Prague.

Contacts:

Dr. rer. nat. Jan Hering
Office G-104 (Building G, 1st floor)

Address:

Czech Technical University in Prague
Faculty of Electrical Engineering
Department of Cybernetics
Karlovo namesti 13
121 35 Prague 2
Czech Republic

I also have a ResearchGate profile.

News

11.02.2018 | Presentation at SPIE MI: Image Processing

Given a presentation of our work on generalized multiple-instance learning and its application in computer-aided diagnosis of Multiple Myeloma in low-dose CT images of femurs. Slides from the presentation are avaialble here
The article is available from here

21.12.2017 | Machine-learning Seminar, MFF

Given a talk at the ML Seminar (MFF, Charles Univ.) about Generalized MIL and application in biomedical imaging. The slides are available from here

Archive

Projects

Multiple Instance Learning (MIL) for bone cancer detection (current)

One of the characteristics of multiple myeloma is the propagation of the disease into the bone marrow tissue. In clinical routine, the extent of bone marrow tissue infiltration is used for staging purposes. Since a thorough and complete annotation of all infiltrations is hard to accomplish, the problem of automatic infiltration detection can be formulated as a multiple-instance learning task.

Generalized-MIL Problems

The standard MIL formulation is based on two (complementary) conditions -- the positive identifiability and negative exclusion. The later says, that in a negative bag, all instances are negative and conversely, for the first one, that a single positive instance already makes the whole bag positive.

The standard MIL formulation is not robust against a single false-positive instance, which can easily occur in noisy conditions. A generalized MIL formulation considers the number of positive instances, that decides wheter the global label is positive or negative is a parameter. We have shown (SPIE MI 2018), that this parameter can be efficiently learned during the training phase.

Related publications

SPIE Medical Imaging, 2018

Former projects

(Pre)processing of diffusion-weighted MR Images

One of the crucial tasks for robust analysis of diffusion-weighted data is the task of correcting for motion and acquisition distortion. To achieve a robust correction scheme I have focused on developing an image registration method that can consider and efficiently combine the information from available image metrics. In the novel, multi-objective correction scheme, a local optimization method by traditional pair-wise image registration is considered together with a global, particle-swarm optimization method.

Related publications

IEEE TMI Paper 2016
Doctoral Thesis

MITK Diffusion

During my doctoral studies, I have contributed to the development and release-process of the MITK Diffusion application, an open-source solution for the processing of diffusion-weighted MR images.

Related publications

Clinical Imaging 2016
JCARS 2014 (Goch et al.)

GPGPU

While finishing my mathematics studies at the Heidelberg University, I worked as student apprentice in the group of Dr. Susanne Kroemker, where my projects focused on parallel computation using graphics cards. Next to the project, i have applied this technique in my diploma-thesis in the context of electro-magnetic navigation in endoscopy

Publications

First Author

---

Journals:

Hering J, Wolf I, Maier-Hein KH, Multi-Objective Memetic Search for Robust Motion and Distortion Correction in Diffusion MRI. IEEE TMI, 2016 DOI
Hering J, ..., Bickelhaupt S., Applicability and discriminative value of a semi-automatic 3D-spherical-volume for the assessment of the apparent-diffusion-coefficient in suspicious breast lesions - feasibility study. Clinical Imaging, 2016 DOI

---

Conferences:

Hering J, Kybic J, Lambert L. Detecting multiple myeloma via generalized multiple-instance learning SPIE Medical Imaging 2018: Image Processing, Houston, TX
Hering J, Wolf I, Meinzer HP, Stieltjes B, Maier-Hein (Fritzsche) KH. A quantitative evaluation of errors induced by reduced field-of-view in diffusion tensor imaging. CDMRI, MICCAI 2013 Workshops, Japan
Hering J, Wolf I, Meinzer HP, Maier-Hein KH. Model-based motion correction of reduced field of view diffusion MRI data. In: SPIE Medical Imaging, 2014.
Hering J, Neher PF, Meinzer HP, Maier-Hein KH. Erzeugung von Referenzdaten für Kopfbewegungskorrektur in Diffusion-MRI. In: Bildverarbeitung für die Medizin 2014.
Hering J, Wolf I, Moher Alsady T, Meinzer HP, Maier-Hein KH. A Memetic Search Scheme for Robust Registration of Diffusion-Weighted MR Images. In: Bildverarbeitung für die Medizin 2015
Hering J, Neher PF, Stiejltes B, Maier-Hein KH. DTI Tractography Challenge MICCAI 2014 – MITK Global Tractography

---

Abstracts:

Hering J, Neher PF, Meinzer HP, Maier-Hein KH. Construction of ground-truth data for head motion correction in diffusion MRI. Proc. Annual Meeting ISMRM, Mailand 2014

Co-Authored

Hirjak D, Thomann PA, Wolf RC, Kubera MK, Goch CJ, Hering Jan, Maier-Hein KH. White matter microstructure variations contribute to neurological soft signs in healthy adults: White Matter Connectomics and Neurological Soft Signs. Human Brain Mapping, 2017 DOI