Search
  • Sam Boone

PhD opportunity developing AI-powered automated digital microscopy techniques for the Earth Sciences

Updated: Nov 19, 2021

The Melbourne Thermochronology Research Group is seeking a highly motivated PhD candidate to pursue a project in the fields of Computer Science and Earth Science at the University of Melbourne, Australia, to begin in 2022.


As part of a cross-disciplinary collaboration between the MTRG, the School of Computer and Information Systems and members of the Melbourne Data Analytics Platform (MDAP), this PhD project aims to utilise machine learning techniques to develop artificial neural networks which are capable of identifying, quantifying and characterising microscopic crystallographic features in minerals.


Preliminary results for two image sets of fission tracks in apatite. 1st column shows transmitted light images; 2nd column, reflected light images; 3rd column portrays solution files with marked fission track openings; 4th column shows fission track openings predicted by the artificial neural network.


Digital microscopy is an emerging cornerstone technology for the Earth sciences, enabling scientists to quantify the optical, morphological and compositional characteristics of geological materials and constrain their provenance, chemical, pressure and thermal histories. Yet, current techniques require significant manual operation by expert analysts, rendering them time-consuming and prone to bias. The implementation of machine learning techniques into digital microscopy will enable unprecedented analytical automation, helping to remove user bias and making these powerful techniques more widely available to a wider non-specialist community.


The project will focus on the digital fission track methodology, a microscopy-based geological dating technique involving the identification and measurement of radiation damage zones in crystals resulting from spontaneous fission of 238U atoms. This is a technique pioneered by researchers in the MTRG who have developed the world’s leading technology in this area (Fission Track Studios). Utilising an existing database of more than 30,000 photomicrograph stacks of fission tracks in mineral grains, the PhD candidate will develop, train and test an artificial neural network for the purposes of automatically identifying and quantifying 3D crystallographic features under the supervision of MTRG and MDAP researchers. The applicability of the resulting image analysis algorithms to other digital microscopy techniques will also be explored.


Digital fission track analysis using the FastTracks software, developed by the Melbourne Thermochronology Research Group.


Eligible applicants should have a BSc in a related field, and an Honours or, preferably, MSc degree in Computer Science or Earth Science, as well as a strong record of oral and written communication skills in English. Applicants with an Earth Science educational background should have prior computational science experience and knowledge of the Python programming language. Additional programming skills (e.g. familiarity with Linux), machine learning experience, other programming languages (e.g. Java), and familiarity with microscopy and geological concepts are valued but not required.


Students who are accepted into the program are automatically entered for consideration for a Graduate Research Scholarship or Melbourne Research Scholarship, which are awarded to domestic and international students, respectively, based on academic merit to cover remission of fees and living expenses. A range of other more specific scholarships are also available (https://scholarships.unimelb.edu.au).


To enquire or apply, please contact Dr Samuel Boone (samuel.boone@unimelb.edu.au). Applicants should provide the following:

  • Cover letter (stating motivation, summarising scientific work and research interest)

  • CV (education, work and research record, publications, and other qualifying activity)

  • Names and contact details of 2-3 references (name, relation to candidate, and e-mail)

106 views0 comments