{"id":57715,"date":"2022-02-01T14:42:17","date_gmt":"2022-02-01T09:12:17","guid":{"rendered":"https:\/\/www.mygreatlearning.com\/blog\/videos-sessions-are-of-great-use-to-understand-the-concepts\/"},"modified":"2024-11-05T16:58:51","modified_gmt":"2024-11-05T11:28:51","slug":"videos-sessions-are-of-great-use-to-understand-the-concepts","status":"publish","type":"post","link":"https:\/\/www.mygreatlearning.com\/blog\/videos-sessions-are-of-great-use-to-understand-the-concepts\/","title":{"rendered":"Great Learning videos & sessions are of great use to understand the concepts in-depth - Lenin Christopher, PGP AIML"},"content":{"rendered":"\n

I was a full-stack developer familiar with Angular, React & Java. I was a Java Architect before joining the PGP-AIML program.<\/a><\/p>\n\n\n\n

We need an AI web robot to assist doctors in APOLLO hospitals using speech to text for Inpatient services to summarize and take clinical notes. Natural language processing (NLP) software that uses machine learning to find insights and relationships in unstructured data.<\/p>\n\n\n\n

We need to pay a total of INR 12 Crore * 12 = 288 Crores every year, but with the recommended solution, it was completely replaced with the new solution.<\/p>\n\n\n\n

Technologies Used - Python, Pandas, Numpy, Scikit-Learn, TensorFlow, Keras, Apache Spark MLlib, Seaborn & Matplotlib<\/p>\n\n\n\n

Machine Learning Techniques Used: Logistic Regression, K - Means, Naive Bayes, Decision Tree, Random Forest, PCA, Support Vector Machine, Clustering, K-Fold Cross-Validation, Grid Search & Feature engineering.<\/p>\n\n\n\n

Developed an inhouse framework with the base of Apache cTAKES, an open-source NLP project & built 18 modules on top of it and modified the entire project, which currently competes with Amazon Comprehend as a better alternative & was well received by Dr. Prathap C. Reddy, Chairman - Apollo Hospitals Groups in 2019 press meet.<\/p>\n\n\n\n

1) Developed an AI web robot to assist doctors APOLLO using speech to text for Inpatient & Outpatient services by modifying the source code of Apache cTAKES, an open-source NLP JAVA project which currently competes with Amazon Comprehend as a better alternative.<\/p>\n\n\n\n

2) Text generation using Long short-term memory (LSTM) for Clinical notes generation.<\/p>\n\n\n\n

3) MEDICATION Module to extract data from the broken forms of sentences to identify fully specified brand\/item name -Drug Change Status, Strength Unit, Route Annotation, Frequency Unit, Date Annotation, Measurement Annotation, Fraction Strength, Form FSM.<\/p>\n\n\n\n

4) Named Entity Recognition to extract - Anatomical Site Identification, Disease Disorder, Sign Symptom, Procedure, Allergy, Request, Laterality, Allergy Severity, Allergy Reaction, Referral, Specialty, Discharge, FollowUp, Laterality Family History, Health Care Activity<\/p>\n\n\n\n

5) Prediction Algorithm - Auto correct any irrelevant medical context to predict meaningful context with the help of NLP Grammars, Sentiment Analysis, Negation, Laterality, Stop words, Body parts, Time Series, Medical Dictionary, English Synonyms, Acronyms, Positive words, Negative words & Word net data<\/p>\n\n\n\n

6) Time Predictor Algorithm to find out time-series related speech.<\/p>\n\n\n\n

7) Family Relation, Allergy Severity & Reaction Finder Algorithm<\/p>\n\n\n\n

8) Allergy identification based on the positive\/negative conversation<\/p>\n\n\n\n

9) Sentencify Clinical Data \u2013 Neural network trained with real consultations for efficient recommendations.<\/p>\n\n\n\n

10) Laterality & body site of an HRA automatic identification & linkage to corresponding pipers.<\/p>\n\n\n\n

11) Synonym Prediction Algorithm<\/p>\n\n\n\n

12) NLP to identify clinician shorthand text forms & resolve accordingly<\/p>\n\n\n\n

13) Body Site Identification & Tag in relationship extraction module<\/p>\n\n\n\n

14) API for User Dictionary, User Abbreviation & Global Synonym<\/p>\n\n\n\n

15) Medipad Template Recommendation<\/p>\n\n\n\n

16) Sentiment Analysis \u2013 Polarity of clinical sentence mapped with pipers.<\/p>\n\n\n\n

17) Pre-annotated machine learning-based test-value extraction component - Map to selected terminologies<\/p>\n\n\n\n

(ICD10, SNOMEDCT, LOINC, RxNorm) based on concept semantic types.<\/p>\n\n\n\n

The Impact, my recommendations generated at the organization, was COST SAVINGS:<\/p>\n\n\n\n

Total charge calculation for Amazon Comprehend, which is the only competitor in the market:<\/p>\n\n\n\n

Size of each request = 2,550 characters<\/em><\/p>\n\n\n\n

Number of units per request = 26 units [2,550 characters \u00f7 100 character per unit]<\/em><\/p>\n\n\n\n

Total Units: 1,000 (requests) x 26 (units per request) = 26,000<\/em><\/p>\n\n\n\n

Price per unit = $0.01<\/em><\/p>\n\n\n\n

Total cost = [No. of units] x [Cost per unit] = 26,000 x $0.01 = $260.00 or INR 18,535.27<\/em><\/p>\n\n\n\n

Total Per Year (USD) - $ 78,000 * 12 = $ 9,36,000<\/em><\/p>\n\n\n\n

Total Savings Per Year (INR) - INR 12 Crore * 12 = 288 Crores<\/em><\/p>\n\n\n\n

\u2022 Propelled 3-year revenue growth from $1.2B to $12.25B.<\/em><\/p>\n\n\n\n

\u2022 Boosted market share by 8% & customer satisfaction by 65%<\/em><\/p>\n\n\n\n

It has helped me to understand & apply machine learning concepts to solve real-world use cases.<\/p>\n\n\n\n

Great Learning videos & sessions are of great use to understand the concepts in depth.<\/p>\n\n\n\n

Resources That Shaped Lenin\u2019s Successful Learning<\/h2>\n\n\n\n