Introduction 

Good morning everyone thank you for giving me the opportunity to talk about IT and research. It goes without saying that this is a very important and interesting topic in today's context and this is a topic that is close to my heart as those who know my IT background would undoubtedly appreciate. 

That aside, since the time is limited, let’s jump right into it

What I am going to discuss is Ai and some tools for research. that’s it. No big deal.

At the end of the discussion today, you will learn a couple of tools you may want to try out on your own research journey.

However, the most important objective of my speech is not that.

I want to delve deep into the cognitive process of research, the thinking, the learning and the human intelligence part. how that is paramount in the research process. Artificial part comes after that. Its role is to complement human intelligence - NOT to replace it. 

Therefore in effect I am talking more about IA than Al.

IA stands for intelligence augmented. 

Outline

My talk will have the following outline

1. When does the need to research arise?

2. How does the research process naturally unfold?

3. What needs to be done before turning to the machine?

4. What tools can be used along the way?

5. What problems need to be addressed?

6. Final remarks

When does the need to research arise?

To answer this question unfortunately I have to talk a little bit about myself. I am a learner at heart as you would have gathered from the introduction, I have some experience and useful skills in learning and teaching. Currently I work full time in the private sector tertiary care level hospital, after working in most remotest parts of the island for 7 years. As I am a clinician primarily, my research needs are primarily clinical scenario oriented. At times I will have to do in-depth research on topics when I work on more academic topics and write journal articles. 

I think this requirement is familiar to you as well. Some of you are academic researchers, some of you are postgraduate students and research is a prerequisite for your masters degree, some are clinicians and you will encounter those clinical scenarios as well. 

The basic premise is we have a partially answered “problem” to solve.  

Research Process

As mentioned it starts with the “Need for knowledge”. It may be intrinsically motivated by your own needs or extrinsically motivated by an external requirement. 

Once idea generation happens a study is designed, regulatory approvals and funding if any are seeked. Study is carried out, results analyses and presented 

+---------------------+

|  Idea Generation    |

+---------------------+

           |

           v

+---------------------+

|    Study Design     |

+---------------------+

           |

           v

+---------------------+

| Regulatory Approval  |

+---------------------+

           |

           v

+---------------------+

|  Conducting the     |

|       Study         |

+---------------------+

           |

           v

+---------------------+

|   Data Analysis     |

+---------------------+

           |

           v

+---------------------+

|  Reporting Results   |

+---------------------+

First step is formulating an answerable question

Next we need an answerable question. If we cannot formulate a structured question our answer from the quest is likely to be underwhelming. However this notion is also being challenged by the semantic search capabilities of the current large language models. 

Enter the PICO format. 

P population/patient group

I : Intervention

C : comparison control group

O : outcome or result

E.g in diabetics who had a coronary angioplasty prior to an MI, i.e. identified by coronary CT, how long should dual anti platelets be administered, compared to otherwise, to prevent stent stenosis or acute MI.

This part of the process HAS to be yours. At least in the initial phase. Then we can enhance it with the help of Ai. 

Let’s get AI to enhance this.

Here is the prompt I used for that 

I want you to act as a seasoned medical researcher with 20 years of experience. I will paste a clinical research topic formatted in PICO format. I want you to improve it so that I can get valid, relevant and impactful research on the topic: in the diabetics who had a coronary angioplasty prior to an MI, i.e. identified by coronary CT, how long should dual anti platelets be administered, compared to otherwise, to prevent stent stenosis or acute MI.

This is the output I got from Claude Sonnet 3.5 

“As a seasoned medical researcher with 20 years of experience, I can help refine your PICO format to ensure you get valid, relevant, and impactful research on this topic. Let's break down and improve each component:

P (Population): Diabetic patients with a history of coronary angioplasty prior to myocardial infarction (MI), identified by coronary CT

I (Intervention): Extended duration of dual antiplatelet therapy (DAPT)

C (Comparison): Standard duration of DAPT

O (Outcome): Prevention of stent stenosis or recurrent acute MI

Improved PICO format:

P: Adult patients (age ≥ 18 years) with Type 2 diabetes mellitus who have undergone percutaneous coronary intervention (PCI) with stent placement prior to experiencing a myocardial infarction, as confirmed by coronary computed tomography angiography (CCTA)

I: Extended duration dual antiplatelet therapy (specific duration to be determined by the study, e.g., 18 months, 24 months, or 30 months)

C: Standard duration dual antiplatelet therapy (typically 6-12 months, as per current guidelines)

O: Primary outcomes:

1. Incidence of stent stenosis (≥50% diameter stenosis) at follow-up angiography

2. Recurrence of acute myocardial infarction

Secondary outcomes:

1. All-cause mortality

2. Cardiovascular mortality

3. Stent thrombosis

4. Major bleeding events (as defined by TIMI or BARC criteria)

5. Quality of life measures

Time frame: Follow-up period of at least 3 years post-PCI

This improved PICO format addresses several key points:

1. Specifies the type of diabetes (Type 2) and age group, which is important for generalisability.

2. Clarifies the intervention as PCI with stent placement, which is more specific than "coronary angioplasty."

3. Defines the extended DAPT duration more precisely, allowing for comparison of different time frames.

4. Includes both primary and secondary outcomes, providing a more comprehensive assessment of benefits and risks.

5. Adds a time frame for follow-up, which is crucial for assessing long-term outcomes.

This refined PICO format should help you find more relevant and impactful research on the topic, allowing for a more nuanced understanding of the optimal DAPT duration in this specific patient population.

You can use any AI, like chatGPT, gemini etc. It's important to prompt it properly. Here I have given the persona of a seasoned researcher, explained my objective and got the above as the result.

Next step, access literature

Armed with a properly formulated research question, we need to access the literature now.

Those who can remember we used to go to the library and search for a book called index medicus. There will be code and reference to a particular journal article where we will have to ask the librarian to search and provide.

Then came the INTERNET and PUBMED. 

PubMed was first released in January 1996 as an experimental database under the Entrez retrieval system, providing free access to MEDLINE, a premier bibliographic database maintained by the National Library of Medicine (NLM). The initiative aimed to democratize access to biomedical literature, which was previously accessible mainly through institutional facilities. By June 1997, PubMed was officially launched to the public, marking a significant shift towards private, free, home- and office-based MEDLINE search

We are blessed to remember the pre internet search era so that we can appreciate what we have now!

For us to achieve this goal we need some tools.

Tools

As this is an article on artificial intelligence, I will be concentrating on digital tools - both software and hardware. In today's rapidly evolving technological landscape, relying solely on a pen and paper approach for research would put one at a significant disadvantage. While traditional methods have their place, the complex nature of AI necessitates leveraging advanced digital tools to stay competitive and current in the field.

tools what do we need

1. We need a way to meaningfully search/browse the internet

2. We need to store all the details systematically so that we can access them whenever we want

3. We need to be able to synthesize connections and process information in our own minds, create mental schema and representations, I think this step is the most crucial of them all. Reading without comprehension is like consuming a meal without proper digestion. Just as the body needs time to extract nutrients from food, the mind requires focused attention to absorb and benefit from information. When we just collect details without processing, we risk letting valuable knowledge pass through our minds without retention or understanding, much like nutrients that aren't assimilated in a rushed digestive process.

4. We need to provide some output from that 

The tools I use for these are as follows:

For searching the knowledge, we will need a browser. I have used many and these days it’s Arc browser. Among others, following are the key reasons

1. AI integration

2. Spaces for various contexts

3. Easel feature to collect stuff 

4. Speed and great tab management

These features are really helpful but not mandatory. You can use practically any modern browser.

For Literature search process we have couple of options:

Pubmed

Semantic Scholar

Consensus

elicit 

Papers

Perplexity

And my favourite is SciSpace. Here is a screenshot of how the outcome of the previous query looks like.

So now with these details and background understanding, I can start collecting actual papers or hop on to research rabbit

Then we need to hold that information somewhere before I process them.

Enter Zotero, my chosen reference MANAGER. You may use Mendely or any other software. But you will need one. This is the “central repository”. So when in need, I know where the knowledge is stored. The collection of papers can be exported out of research rabbit to Zotero easily.

zotero also has an inbuilt PDF reader and I will do the annotations on the PDF within Zotero and then I collect all those details into my note taking app. 

My current note taking app is obsidian: A markdown editor. 

Advantages offered by obsidian include simplicity, portability, ability to synchronize between multiple devices , ability to link notes and thoughts, and as they call it - “future proofing”.

Once the processing has happened in Obsidian, if there is an output to happen like a paper or a presentation, I use Google Workspace to create. You can use Microsoft Office suite or Openoffice as well. 

This is the process I have refined over the years. But there is a (new) problem here.

Problem of pace of change

There is an interesting relationship between technological progress and human adaptation. Over time humans adapt and a somewhat linear pattern but the technological progress moves in a logarithmic fashion. Following is a graphical representation of this idea first put forward by Astro Teller CEO of Google X. Problem is how can we keep up with this rapid advancement

closing remarks

Life itself 

It’s pivotal to consider and craft your life itself to be a research orientation. This goes without saying but if you really consider the strains of life and career we have to endure today, if correct skills are not inculcated, we can’t even think of using artificial intelligence in research! Couple of things that come to my mind are as follows:

1. Adaptability and agility: you will not have all the necessary tools and the ideal environment set up for you all the time, however you must be ready to use what’s available to do the maximum possible.

2. Growth mindset You need to consider that you are not the best version of yourself YET. Therefore everything that happens to you is a welcome opportunity for you to learn and improve the current state.

3. learner mindset : you must be inquisitive and have an insatiable desire to learn. This is not cliche. This is a fact.

Study environment and Appropriate Tool Use

Remember all of us are different in our preferences. I work best on the move and in community learning environments. Someone may prefer the privacy and solitude in their own private study. find what works for you and keep on experimenting until you find yours. 

Custom GPTs and ai writers

There are ways to automate everything using agents, customGPTs and Ai writers. But my current understanding and recommendation is strong discouragement of that particular use case. The human cognitive process however slow and inefficient it may be is indispensable. After all, your own intelligence has to improve too. 

Critical thinking skills need to be honed. These intellectual exercises strengthen your mental muscle. You wouldn’t want your personal trainer / gym instructor to do your physical training, do you ?

Problems need addressing 

Ethical aspects

Another important aspect in using these tools is to consider ethical aspects of research. The important key element in research is discovery of truth. If there is any doubt about the veracity of the information used the near output of the research is also at this question.

We know artificial intelligence can make up study findings, including data Recent mass retraction of scientific publications is a testimony to this. Many prestigious institutions which we would never think would be involved in this fraudulent activities have been implicated. 

This is a grave of concern since the outcome of medical research has a direct and strong relationship with human medical care. Fraud at the point of origin of knowledge can not be good for humanity. 

Additionally, various biases, including cognitive biases, can result in ethical dilemmas for the quality of the research output

requirements by established journals 

Reputed journal have published their Ai requirements, You can find these online in their respective websites. 

https://youtu.be/KC1KPvBm51E?si=r5kmKXLNhI_xM38b

Deeper Problem

In future, when you see a research finding it would be rather difficult to determine whether this is artificially generated research or genuine research maybe augmented by artificial intelligence, but not fabricated. 

Human intelligence will shine

I believe in the future, one of the most important critical skills, necessary and valued, would be human intelligence itself.  We should teach and train people in critical thinking skills and in applying human intelligence to ascertain usefulness and usability of the output of artificial intelligence. 

Communication and critical thinking skills are going to be vital and will give an unfair advantage to those who possess it. Ability to articulate yourself clearly and accurately will be as important as any other clinical skill taught these days in medical faculties.

Thank you very much

Dr MTD Lakshan