Theoretical knowledge, often passed on through books, makes it possible to grasp the fundamental concepts of a subject, a profession or a science. However, for knowledge to be passed on in full, it is essential to combine this theory with methodology (how to reason about complex subjects) and know-how (gained from practical experience). These two forms of knowledge are difficult to convey effectively in the pages of a book because they are non-linear and require interaction and context. Writing knowledge in the form of decision trees makes this possible.
A complex task is broken down into a succession of simple tasks that follow a logical order. The answers at the previous step determine the next question and the possible answers. The required knowledge is offered at the right moment, as an image or explanatory text accompanying the current question. The user does not need to read an entire book to access the knowledge; it is immediately available in the right place.
Know-how is also conveyed in decision trees in three ways:
- Report templates: to describe a circumstance (for example, trauma) or a condition (for example, appendicitis), a specific path through the tree can be selected. The user thus follows a checklist adapted to the specific situation. A subject-matter expert can describe the most frequent situations.
- Precise options: if the answers to the previous question rule out some possible answers to the next question, it is easy to offer the appropriate question with only the remaining answers. This means writing the same question several times with different answers in each branch of the tree. Whereas this would be indigestible in a book, it becomes easier to read in a decision tree, where only one question is visible, with a choice of answers narrowed down according to the previous answers.
- Built-in logic: decision trees use logic nodes and calculation nodes. The latter make it possible to create Boolean logic and to perform mathematical calculations or score calculations. They act as switches within the tree, opening or closing a branch depending on the result obtained. They are also used to summarise the result of a piece of work in the conclusion of the generated report, incorporating professional expertise (for example, for a radiologist, classifying a lesion as benign or aggressive, generating a TNM classification or follow-up scores).
Using this knowledge in tree form allows everyone — novices, experts or machines — to easily carry out complex intellectual work, while writing a report that is reliable, explainable, reproducible and structured.
This approach is the method used by Doctreen: we break a complex problem down into a succession of simple questions, asked in the order that corresponds to best professional practice. The reasoning progresses answer by answer, ensuring the completeness, accuracy and standardisation of the process.
Doctreen trees can incorporate calculation and Boolean logic algorithms that make it possible to build the intelligence for reasoning and drawing conclusions. A conclusion is automatically proposed to the user, who is responsible for completing it and then validating it with their human judgement.
The trees are enriched with interactive illustrations, making it possible, for example, to select an area on an image to designate an anatomical region in a medical report, or a technical component of a machine in an industrial maintenance report.
The structured data produced in this way can then be put to use at a later stage, for research, continuous improvement or enhancing algorithms.
