Table 2 Analysis by means of the framework (see Table 1) in the first column, of the disciplinary perspectives of four disciplines (I – IV) involved in the development of diffusion MRI for the diagnosis of kidney cancer. Next to the general descriptions of each disciplinary perspectives, each aspects is specified for the case at hand and relative to each discipline

[1] What is the overarching goal of the (disciplinary) professional or research practice?

The goal of this practice is to diagnose and treat patients.

In the case at hand, the goal is to diagnose and treat patients with kidney tumours according to their specific tumour type.

The goal of medical biology is to understand and describe (by visualizations, simple equations or words) mechanisms of human physiology and pathology.

In the case at hand, the goal is to determine which tissue properties can be used to distinguish between kidney tumour tissue types (see I).

The goal of MRI physics is to develop machines that induce and manipulate the nuclear magnetic resonance (NMR) signal and to produce accurate representations of the object of imaging based on this signal. An example of the latter is visualizing microstructural properties of tissues by probing the diffusion signal.

In the case at hand, the goal is to find those diffusion properties that correlate to tissue properties that can be used to distinguish between kidney tumour tissue types (see II).

The goal of signal and image processing is to turn NMR signals into visualizations and to process images so as to correct for artefacts and to derive relevant parameters (e.g., the diffusion coefficient).

In the case at hand, the goal is to derive those parameters that reflect relevant diffusion properties to distinguish between kidney tumour tissue types (see III).

[2] What are the kinds of phenomena the discipline is typically interested in?

The phenomenon of interest is the disease, which includes how it can be diagnosed and treated.

In the case at hand, it is the ‘kidney tumour’ as a clinical phenomenon, as something that affects a patient’s health. It includes how the tumour can be diagnosed, and whether it can be diagnosed by means of diffusion MRI (in addition to the standard clinical imaging and blood tests).

The phenomena of interest in the biomedical perspective on kidneys concern the functions and (dis)functioning of human physiology, divided in organs, systems of organs, organ components, cells, cell organelles, etc.

In the case at hand, the organ of interest is the kidney, the tissue types that enables the kidney’s functioning and kidney tumour formation.

The phenomena that MRI physics deals with are magnetic fields, proton spins, radiofrequency pulses, relaxation times and coils.

The phenomenon of interest in signal and image processing is images as arrays of measurements, where each measurement is represented by a number. In other words, signal and image processing entail processing of arrays of numbers. This includes fitting mathematical equations, extrapolation, histogram analysis, noise suppression, etc.

[3] What is the objective of research or investigation in the discipline (i.e., the objective of using and producing knowledge) in this discipline?

The objective of investigation by a clinician is to diagnose the cause of a disease, and to develop a treatment plan.

In the case at hand, the objective is to diagnose kidney tumours which involves determining its presence, type and severity of the disease and determining a suitable treatment plan.

The objective of research in medical biology is to describe and understand the mechanisms by which organs etc. fulfil their functions. The phenomena of interest (e.g., the functions or organs) are explained in terms of (micro) architecture, shapes (of organs or cells), molecules (e.g. proteins, hormones and Na+/K + pump) and interactions.

In the case at hand, different tumour tissue types are described in terms of the presentation of their cells under the microscope (i.e., how they ‘look’) and in terms of their behaviour (i.e., how rapidly they grow).

The objective of research in this field involves developing new hardware (e.g., coils or other components of the machine) to improve image quality or to enable imaging of specific body areas, developing new acquisition sequences that allow new signal contrasts and hence visualizing new tissue aspects, and software development.

In the case at hand, the objectives of investigation, and of technological advancement in diffusion MRI, firstly relate to developing models that accurately represent tissue physiology, and secondly, to apply diffusion-weighted imaging to different pathologies to investigate the feasibility of detecting changes in tissue structure due to these pathologies.

The objective of research in this field is to improve mathematical manipulations of the signals by developing better (e.g., more accurate to the imaged tissue) methods for fitting, extrapolation or analysis of the signal.

In the case at hand, the objective is to develop algorithms that produce a combination of parameters that can be used to accurately characterize different kidney tumour types.

[4] What are the kinds of (mental or scientific) models or ‘pictures’ build to represent the knowledge about the phenomenon of interest?

Clinicians construct a ‘picture’ of the specific patient and his/ her disease by gathering relevant but heterogeneous information, and fitting it into a coherent whole (the picture, or model of the patient, see Van Baalen & Boon 2015).

The ‘picture’ of the patient enables further reasoning by the clinician about the patient, for instance, in order to get to a diagnoses, or to come up with a treatment plan that is suitable to the patient.

In the case at hand, clinicians use their ‘picture’ of patients with different types of kidney tumours to explain how information that they can derive from what they see in a diffusion MRI scan will affect their clinical decisions regarding these patients.

The way of modelling in this disciplinary perspective is causal-mechanistic, that is, the phenomena of interest (functions) are described and explained by models that represent the causal-mechanisms.

These models enable causal-mechanistic reasoning about the (dis)functioning of organs etc., in order to find causal explanations of the phenomena of interest.

In the case at hand, these models enable understanding the mechanisms that can be used to explain the different tissue properties that can be used to distinguish between tumour types.

The behaviour of the phenomena of interest and their interactions are modelled through mathematical equations, some of which stem from basic physics (e.g., electricity and magnetism), while others are more specific to MRI, (e.g., related to relaxation times, and formulas describing the relationship between the acquired signal and the image, and equations relating the acquired diffusion NMR signal to diffusion properties of the tissue).

These models enable reasoning, for instance, about how different tissue properties (e.g., tumour types) will result in different NMR signals.

The theoretical basis of this disciplinary perspective are mathematical models, such as the exponential model that is used to obtain the diffusion coefficient from the diffusion signal.

These models are used to find the most appropriate (set of) diffusion parameters to characterize different kidney tumour tissue properties.

[5] Which theories and concepts (about the phenomena of interest) are typically used in this discipline?

In crafting a picture of a patient (or class of patients), clinicians use medical concepts and theories, including knowledge that concerns the diagnosis and treatment of a disease. The latter is mostly developed in clinical trials, studying the effectiveness and efficacy of treatments to specific patient populations, or the sensitivity and specificity of diagnostic tools.

In the case at hand, clinicians specialized in kidney tumours are familiar with imaging of the kidney as a diagnostic tool to visualise the size and location of the kidney tumour.

Theories and concepts include theories that describe and explain the functions and mechanism of an organ or a system of organs.

In the case at hand, an example is the mechanism with which blood is filtered in the glomeruli.

Theories and concepts that are fundamental to this field are classical electrodynamics (referring to (electro)magnetic fields, flux, induction, currents and coils), nuclear magnetic resonance (describing the magnetic moment of nuclei caused by resonance due to a magnetic field), relaxation and Fourier transform (fundamental to transforming the measured signal into images).

Subfields are arranged according to acquisition method (e.g., diffusion MRI, functional MRI, etc.), body area or organ (e.g., abdomen, limbs, head, brain, heart) or pathology (e.g., tumours), or, focusing on technological advancement by developing new hardware or software development.

Theories and concepts in computer sciences (including mathematics) and information technology.

[6] Which methodology and (technological) instruments (to explain or investigate the phenomena of interest) are typically used in this discipline?

The methodology by which clinicians construct a picture of the patient (or a class of patients) is through gathering relevant information, on the one hand, by means of applying diagnostic techniques to the patient, and on the other hand, based on general medical knowledge and theory.

Methodology in medical biology include methods common in molecular biology, such as studying the cellular architecture of tissues and how molecules (proteins) are produced, excreted, metabolized and taken up by cells, in a laboratory environment (e.g., by looking at the organs themselves or the tissue in the microscope).

Methods of investigation in this field are, phantom experiments (to visualize or quantify known and controllable processes in MRI), experiments with healthy volunteers (to investigate the most adequate acquisition and modelling protocol for certain body areas), (clinical) trials with patient populations (to investigate the clinical value of a certain imaging protocol), and computer simulations (to simulate the effect of varying parameters).

The central methodology of image processing involves mathematical operations and computer programming, while experts in this discipline think of images as arrays and matrixes of numbers, rather than a literal representation of shapes and structures inside the body.

[7] What are the practical constraints regarding the overarching goal of the practice?

Practical constraints in constructing a picture of individual or classes of patients and in coming up with a treatment plan, are due to the workings of everyday clinical practice: doctors aim to diagnose and treat patients within the scarcity of time, knowledge and resources. In addition, treatment options are usually limited and associated with side effects. This implies that practical constraints are important to how clinicians assess diagnosis and treatment for specific (classes of) patients.

In the case at hand, these practical constraints will play a role in how clinicians participating in the development of an MRI tool assess its diagnostic quality and usefulness.

Practical constraints in this field relate to the availability of instruments to observe or measure features of interest. The mechanisms medical biologists are interested in are not directly observable since they take place in a living body that cannot be taken apart to look at them. Hence, researchers have to come up with (indirect) measures and methods to manipulate cells, proteins and organs.

Practical constraints in this field concern physical aspects of the MRI apparatus (e.g., bore size, magnetic field strength, gradient strength), which limit the kinds of objects that can be imaged as well as the imaging parameters that can be obtained (e.g., field of view, signal to noise ratio, resolution). Another practical constraint is the computing power needed to transform the data into images.

An important aspect of image processing is the available software. Custom-made algorithms need to be programmed in general programming languages such as MATLAB and C++.

Other practical constraints are primarily formed by the computing power of the computer with which data is transformed into images, but also by available software and scripts

[8] Which are the epistemic and pragmatic criteria that the discipline aims to meet in using and producing (novel) knowledge about the phenomena of interest?

Epistemic criteria playing a role in how a clinician uses and produces knowledge (i.e., how she produces the picture of a patient and how she uses it to reason about the patient and his disease) are relevance and reliability of information for the patient population and individual patients.

In the case at hand, for example, the clinician takes into account knowledge about the incidence of different types of tumours related to age, gender, lifestyle, etc. Additionally, in assessing the new MRI tool, the clinician will take into account the relevance and reliability of images produced by it.

Epistemic criteria for knowledge production in medical biology is, firstly, that the mechanisms can predict and/or explain (i.e., have predictive and explanatory value) the phenomena of interest; secondly, that these predictions and/or explanations are adequate; and thirdly, that these mechanisms are consistent with other knowledge.

The epistemic criteria for knowledge production in this field are consistency between the theory, predictions and the phenomenon (the MRI signal) that is produced and measured; reliability of production and manipulation of the MRI signal, which is often established in terms of reproducibility (i.e., similar results are obtained for similar cases) and the validity of the images produced as representation for imaged object (i.e., a part of the human body).

Important epistemic criteria concern the accuracy of the models to the imaged tissue types, and the validity of the processed signal to the imaged object.