Tampere University Medical Physics and Biomedical Instrumentation Master's Thesis Guide: 30 ECTS, BBT.MJS.111 and Trepo

Quick answer: what is the MPBI thesis route?

Tampere University’s Medical Physics and Biomedical Instrumentation (MPBI) is a 120 ECTS Master of Science (Technology) study option in Biomedical Sciences and Engineering. The current degree page states that the master’s thesis carries 30 ECTS. The thesis therefore follows Tampere’s Technology thesis process and is graded 0-5. The current field object BSEM.MPBI is labelled at least 120 credits; that is a programme-level object, not 120 credits to add on top of the degree. Thesis preparation uses BBT.MJS.111 Master’s Seminar, Biomedical Sciences and Engineering, 2 ECTS, pass/fail. Keep these objects separate when planning credits.

1. Understand the two MPBI paths

MPBI has two current paths. Biomedical Instrumentation concentrates more on design and development of biomedical instrumentation. Medical Physics emphasizes the theoretical physics background needed for later medical-physicist specialist studies, including radiation physics, radiation oncology physics and medical imaging. A thesis can sit primarily in one path or at their interface. Do not assume that every MPBI thesis needs patient data, ionizing radiation, electronics or machine learning. Start from the actual research question, supervisor expertise, approved infrastructure and evidence level you can realistically produce.

2. Do not add the BSEM.MPBI credit label twice

The current Student’s Guide exposes BSEM.MPBI Medical Physics and Biomedical Instrumentation, at least 120 cr. The degree programme itself is also 120 ECTS and separately states that the master’s thesis is 30 ECTS. Therefore BSEM.MPBI must not be interpreted as 120 ECTS of taught specialization courses plus another 30 ECTS thesis. Use your own Sisu plan to see how the programme object is decomposed for your cohort. The stable current thesis facts are the 30 ECTS thesis and the Faculty of Medicine and Health Technology seminar route.

3. BBT.MJS.111 is the current thesis seminar

BBT.MJS.111 is a current 2 ECTS, pass/fail advanced-studies seminar. It develops understanding of the thesis process and scope, good writing practices, peer presentation and familiarity with other master’s theses in biomedical fields. The seminar is not the thesis and its pass/fail result is not the 0-5 thesis grade. This is also why CSEE’s ITC.CEE.800 should not be imported into MPBI: the programme already has a Faculty of Medicine and Health Technology thesis-seminar route.

4. Turn a broad health-tech idea into a bounded research claim

“Build a better monitor”, “improve medical imaging”, or “make diagnosis more accurate” is not yet a defensible thesis question. Define the physical or physiological quantity, device or algorithm, comparator, operating conditions, population or phantom, and the exact performance claim. A useful structure is: under specified conditions, does method/device A estimate or improve quantity B relative to reference C according to metric D? This forces the Methods section to contain the evidence needed for the conclusion and prevents a laboratory prototype from being described as a clinically proven medical device.

5. Separate the evidence layers

MPBI projects often move through several evidence layers: theoretical derivation or simulation; engineering implementation; bench or phantom testing; healthy-volunteer or participant measurement; patient or clinical data; and finally clinical or regulatory interpretation. These layers are not interchangeable. A correct simulation may support a physical model. A phantom study may establish technical performance. Participant data may establish measurement behavior under a defined protocol. None of those automatically prove patient benefit, diagnostic effectiveness, treatment safety or regulatory conformity. Label the level of every major conclusion.

6. Biomedical instrumentation: document the measurement chain

Current BBT.037 Biomedical Instrumentation grounds MPBI in noise/interference control, biopotential electrodes, biomedical measurement circuits, electrical safety, PCB design, assembly and characterization. For an instrumentation thesis, document the complete measurement chain: sensor or electrode interface, analog front end, gain, filtering, sampling, ADC resolution where relevant, reference/ground strategy, shielding, power and communication path, firmware/software version and final signal processing. A block diagram is useful, but the text must explain which choices materially affect the reported performance.

7. Calibration and metrological thinking

A measurement claim needs a reference. State what was calibrated, against which reference quantity or instrument, over what range, how often, and with how many repeats. Record environmental or setup conditions that alter calibration. Report an error metric that matches the claim, not merely a high correlation coefficient. For time-varying physiological signals, synchronization and time alignment can be as important as amplitude calibration. If the reference itself has uncertainty, acknowledge it. A thesis should allow the examiner to understand whether apparent accuracy came from the device, preprocessing, a narrow test range or the reference setup.

8. Noise, drift, interference and failure modes

Medical measurements are vulnerable to motion artifact, electrode contact, electromagnetic interference, temperature, baseline drift, saturation and device-specific failure. Do not clean these cases away silently. Define quality checks and exclusion rules before final analysis where possible, and report how many recordings or segments were excluded and why. If the device fails under a particular condition, that can be a valuable engineering result because it defines the operating envelope. Negative evidence should narrow the claim rather than disappear from the thesis.

9. Physiological measurements need biological context

Current BBT.HTI.506 Measurements of Physiological Systems covers physical, optical and bioelectric measurement, wearables, wireless techniques and cardiovascular, neural, muscular and pulmonary monitoring. A physiological signal contains both biology and instrumentation. Record participant posture/activity, sensor placement, protocol timing and other context that can change the signal. Distinguish between biological variability, instrument noise and preprocessing effects. Multiple windows cut from one person are not multiple independent participants. Your statistical unit must match the design.

10. Wearables and ambulatory monitoring

Wearable studies should report attachment and placement, motion context, missing data, battery or connectivity limitations, sampling frequency and artifact handling. If you compare a wearable with a reference system, explain whether both systems observed the same physiological interval and how clocks were synchronized. Avoid claiming general everyday accuracy from a short controlled-lab test unless that is the actual study domain. Conversely, free-living data can be realistic but may make the reference signal less controlled. The thesis should explain this trade-off rather than hide it.

11. Medical imaging: keep acquisition, reconstruction and analysis separate

A medical-imaging thesis may include acquisition physics, reconstruction, post-processing, registration, segmentation and automated analysis. Current BBT.HTI.502 explicitly covers tomographic reconstruction, image quality, PET performance measures, registration, fusion and segmentation. Do not describe all of these as one “image-processing” step. State which image representation enters each stage, what parameters are fixed or learned, and which stage the hypothesis concerns. This separation is essential when an apparent improvement in a downstream metric may actually come from changed acquisition or reconstruction.

12. Image quality needs a task and a metric

“Sharper” or “better-looking” is not enough. Define the image-quality property relevant to the question: spatial resolution, contrast, noise, artifact level, quantitative accuracy, reconstruction error or a task-specific measure. If human readers are involved, define reader selection, blinding, rating scale and the unit of analysis. If a numerical metric is used, explain why it represents the intended clinical or technical task. A technically improved metric does not automatically translate into improved diagnosis.

13. Segmentation, registration and algorithm evaluation

For segmentation, define the reference annotation process and report metrics appropriate to the structure and task. For registration, state the transformation family, reference or landmark definition and error measure. For predictive or learned models, split data at the participant level when multiple images from one person exist; otherwise train/test leakage can create falsely optimistic performance. Preprocessing fitted using the full dataset can also leak information. Keep a locked final test set or an equivalent independent evaluation strategy when the claim is generalization.

14. Radiation imaging and therapy: state the physical quantity

Current BBT.MPBI.601 covers radiation physics, x-ray imaging, CT, biological effects of ionizing radiation, external radiotherapy, brachytherapy and dose planning. A radiation-related thesis must be explicit about what is simulated, planned, measured or clinically delivered. State the relevant exposure or dose quantity and units; do not use “dose” generically when different quantities are involved. Operational radiation safety belongs to approved institutional and legal procedures, not to improvised thesis instructions. The thesis documents the approved research method; it does not replace radiation-safety governance.

15. Dose-planning and treatment studies need a clinical boundary

A dose-calculation or treatment-planning algorithm can be evaluated mathematically or against reference calculations without becoming a clinical treatment decision. Separate algorithmic agreement, phantom measurement, retrospective plan analysis and actual patient-treatment use. If clinical data are used retrospectively, state the permission and data-protection route. Avoid wording that implies a new method is safe for treatment merely because it improved a planning metric. Clinical adoption requires evidence and governance beyond a master’s thesis unless the project is explicitly embedded in an approved clinical framework.

16. Ultrasound imaging and therapy

Current BBT.MPBI.603 covers the physical principles and instrumentation of ultrasonic imaging and therapy. An ultrasound thesis should specify transducer configuration, transmit settings, frequency/bandwidth, geometry, propagation assumptions, reconstruction or beamforming, and the measured endpoint when these are relevant. Phantom, simulation and biological measurements provide different evidence. For therapeutic ultrasound, an imaging-quality result is not proof of therapeutic efficacy or safety. Keep acoustic/physical measurements, model predictions and biological outcomes separated and state which level each conclusion addresses.

17. Prototype performance is not regulatory conformity

Current BBT.MJS.143 covers EU medical-device regulation, safety/performance requirements, quality management, conformity assessment, clinical evaluation, vigilance and post-market surveillance. Use this knowledge to frame your engineering work, but do not convert an academic prototype into a regulatory claim. A thesis can show that a design was developed with particular safety or regulatory requirements in mind; it can document verification against selected requirements. It should not imply CE marking, conformity, clinical authorization or market readiness unless those statuses independently exist and are correctly documented.

18. Requirements, verification and validation are different

Write a small requirements table early: requirement, rationale, verification method and acceptance criterion. Verification asks whether the implementation meets specified design requirements. Validation asks whether the resulting solution is appropriate for its intended use or user need. In a thesis, you may complete extensive technical verification while only performing limited validation. Say so. This vocabulary is especially useful for instrumentation and clinical-device projects because it prevents a successful bench test from being overstated as proof that the device solves the full clinical problem.

19. Human-participant and clinical data

If the project involves people, patient records, medical images, physiological recordings or clinical workflows, establish the applicable ethics, permissions and data-governance route before collecting or accessing data. Tampere’s research-ethics and student data-protection guidance governs these questions. The fact that a supervisor, hospital or research group has data does not automatically mean a student may use it for a thesis. Record the approved data source, permitted variables and processing environment. Do not expose participant information in notebooks, screenshots, AI prompts or the public thesis.

20. Pseudonymised data are not automatically anonymous

A code replacing a name does not necessarily make data anonymous. If someone can reconnect the code to a person, the dataset remains pseudonymised personal data. Medical images and physiological signals can also carry identifying or linkable information. Minimise variables, restrict access and define retention/deletion responsibilities according to the approved plan. For figures, check overlays, headers, filenames and metadata. The public thesis should contain only what can legitimately be disclosed.

21. Participant numbers and repeated measurements

Define the experimental unit before calculating sample size or statistics. Repeated beats, image slices, windows, trials or sessions from one participant do not automatically create independent participants. Likewise, repeated readings from one physical device do not prove between-device reproducibility. If the design has participant, session and measurement levels, preserve that hierarchy in the analysis. This protects against pseudoreplication and allows the examiner to see what population or device class the result actually supports.

22. Choose metrics that match the claim

Different MPBI questions require different metrics. Measurement systems may need bias, precision, repeatability or agreement. Diagnostic algorithms may need sensitivity, specificity, discrimination or calibration. Imaging may need spatial/contrast/noise or task-specific metrics. Classification accuracy alone can be misleading under imbalance; correlation can be misleading for method agreement. State a primary metric and justify it from the intended claim. Report uncertainty or confidence intervals where appropriate and distinguish statistical from clinical importance.

23. Missing data, artifacts and exclusions

Physiological and clinical data are rarely complete. Explain why data are missing: participant dropout, sensor detachment, communication loss, corrupted file, unusable image or protocol deviation. State whether missingness is likely related to the condition being measured. Define exclusion rules transparently. If a preprocessing pipeline automatically rejects segments, report its parameters and rejection rate. A method that only works after removing the most difficult cases may have a narrower practical claim than its average metric suggests.

24. Software, hardware and configuration traceability

Keep an audit trail from hardware/device ID and configuration to raw files, calibration data, preprocessing, analysis script and final figure. Record software, firmware and model versions when updates change output. For imaging, preserve acquisition/reconstruction settings. For instrumentation, preserve schematic/layout revision and component substitutions that affect performance. For hospital systems, record relevant configuration without copying confidential details into the public thesis. Reproducibility is not only code sharing; it is the ability to explain exactly which system produced each result.

25. AI use in an MPBI thesis

Tampere permits AI use within its current study guidance, but responsibility stays with the student. Agree thesis-specific use with the supervisor and acknowledge it where required. Never upload protected patient, participant, hospital, unpublished company or restricted research data to an external AI service without an approved basis. Verify generated references, equations, code and physical explanations. In signal/image analysis, AI-generated code can silently alter preprocessing, units or split logic, so validate outputs against known cases and keep the final analysis reproducible without relying on an unrecorded chat session.

26. Literature review: separate physics, engineering, clinical and regulation

Organise the literature review by evidence role. Physics papers explain mechanisms and limits; engineering papers establish design and technical performance; clinical studies address patient or workflow outcomes; regulatory/standards material defines formal requirements. Do not cite an engineering prototype paper as proof of clinical effectiveness or a regulatory overview as proof that your own device complies. A useful literature matrix records population/model, device/modality, comparator, metric, evidence level and limitations. This makes the eventual discussion much easier to discipline.

27. Supervision plan and critical-path dependencies

Use Tampere’s Thesis Supervision Plan to make scope, meetings, responsibilities and milestones explicit. MPBI projects often depend on hospital access, participant recruitment, specialist equipment, calibration references, ethics review, data permissions or shared research infrastructure. Identify these dependencies before committing to a question. Agree a fallback: simulation instead of unavailable measurement, retrospective data instead of prospective recruitment, a narrower technical claim instead of clinical validation, or a different device endpoint if equipment access fails.

28. Write while measurements are running

Draft Methods as soon as the protocol stabilises. Maintain a result ledger linking each research question to planned figure/table, raw source and analysis script. Record deviations when they happen instead of reconstructing them months later. For hardware projects, photograph or diagram the setup and record device/configuration identifiers. For imaging or physiological data, preserve acquisition and preprocessing parameters. Early writing exposes missing controls and ambiguous units while there is still time to fix them.

29. Results and discussion: do not hide engineering limits

A strong MPBI Results chapter reports the operating region, not only the best case. Include drift, artifact sensitivity, calibration failure, reconstruction limits, exclusion rate and negative results when they are relevant. In Discussion, separate what the study directly showed from what might be possible after further development. If performance is strong only in a phantom, controlled posture or one device revision, say so. This makes the thesis more credible and gives future developers a precise starting point.

30. Maturity, Turnitin, Trepo and publicity

After the scientific work is ready, the institutional thesis process still matters. Complete the applicable maturity test route. Use Turnitin according to Tampere’s originality-check instructions and address issues with the supervisor before final submission. The assessed thesis is a public document, so confidential company, hospital or participant material must be separated appropriately. Submit the final thesis through Trepo in the required archival format and allow time for examination and graduation processing rather than treating upload day as the graduation date.

31. Final MPBI checklist

Before submission, verify the current 30 ECTS thesis and BBT.MJS.111 route in your Sisu plan; confirm supervisor and approvals; freeze hardware/software versions; trace every figure to raw data; retain calibration and reference information; prevent participant-level leakage; state the unit of analysis; justify primary metrics; disclose exclusions; keep radiation/clinical work within approved procedures; separate prototype verification from clinical/regulatory claims; protect personal and confidential data; document AI use; complete maturity and originality checks; produce the required archive-ready file; and reserve time for examination. A defensible MPBI thesis is defined by a transparent evidence chain, not by how medically ambitious the title sounds.

Radiation and patient-data boundary audit. For radiation projects, verify whether every reported value is simulated, planned, measured or clinically delivered, and name the radiation quantity and unit correctly. For patient data, verify patient-level splitting, permissions, pseudonymisation, access controls and public-disclosure boundaries consistently across Methods, Results and Discussion. Patient data should not be treated as a generic engineering dataset when its clinical provenance changes interpretation. Calibration evidence, radiation evidence, patient data and regulatory status each require an explicit final boundary check.

Measurement uncertainty and repeatability. For quantitative instrumentation work, separate calibration error, within-session repeatability, between-session reproducibility and between-device variation. A single aggregate RMSE can conceal these sources. Where feasible, show error across the operating range and inspect heteroscedasticity rather than assuming constant error. If a reference system has non-negligible uncertainty, the thesis should acknowledge that the observed disagreement belongs to the comparison, not automatically to the prototype alone. For physiological measurements, repeatability can also depend on sensor repositioning, participant state and operator technique. State which of these sources your design actually tests.

Clinical workflow and usability boundary. A technically accurate device can still fail in clinical or home use because setup time, alarms, placement, cleaning, data transfer or interpretation are impractical. If workflow or usability is part of the research question, define users, tasks and setting explicitly and evaluate them with a method appropriate to that question. If workflow was not evaluated, do not imply that technical performance proves clinical usability. Similarly, a retrospective dataset cannot establish prospective workflow performance. Keep engineering verification, human-factors evidence and clinical effectiveness as separate layers in the final argument.

Risk-based discussion. When the thesis concerns a medical-device concept, use risk thinking to explain why particular failure modes matter. Identify hazardous situations relevant to the research scope, connect them to design controls or tests, and report residual limitations without claiming that a complete regulatory risk-management file has been produced. For radiation, therapeutic or life-support-adjacent applications, be especially conservative: an academic experiment can characterize a component or algorithm without authorizing clinical operation. This distinction protects both scientific accuracy and the public reader.